fix(test): increase hook test timeouts for Windows in taskcancel tests

Add WINDOWS_HOOK_TEST_TIMEOUT_MS (15s) to taskcancel hook tests to
account for the slower PowerShell bridge that spawns a child Node
process on Windows CI. The double-process startup is variable and can
exceed Mocha's default 2s timeout, causing flaky failures.

Mirrors the approach already used in taskresume, hook-factory, and
user-prompt-submit tests.

.clinerules/debug-harness.md

@@ -1,128 +0,0 @@
-# Debug Harness
-
-HTTP-controlled debugger for the VSCode extension at `src/dev/debug-harness/server.ts`.
-
-## Quick start
-
-```bash
-# Build extension first if needed (protos + esbuild):
-npm run protos && IS_DEV=true node esbuild.mjs
-
-# Launch (skip-build if already built):
-npx tsx src/dev/debug-harness/server.ts --skip-build --auto-launch
-
-# In another terminal:
-curl localhost:19229/api -d '{"method":"status"}'
-```
-
-## Data Isolation
-
-The debugee runs with `CLINE_DIR=~/.cline2` by default, separate from your real `~/.cline`.
-This prevents the debugee's logout from logging out the debugger, and vice versa.
-Override with `--cline-dir /tmp/test-dir`. Check with `status()` → `clineDir`.
-
-## Browser Capture & OAuth
-
-The debugee runs with `CLINE_CAPTURE_BROWSER=1`, which intercepts `openExternal()` in
-`src/utils/env.ts`. URLs are captured instead of opening a real browser:
-
-- Logged to `$CLINE_DIR/data/debug-captured-urls.jsonl`
-- POSTed in real-time to `/captured-url` on the harness server
-- Queryable via `oauth.captured_urls`
-
-### OAuth API
-
-- **`oauth.captured_urls`** `{clear?}` — URLs the debugee tried to open
-- **`oauth.read_stored_token`** — Check auth token presence in secrets.json
-- **`oauth.simulate_callback`** `{path, code?, state?, provider?, token?}` — Build vscode:// callback URI
-- **`oauth.read_captured_urls_file`** — Read on-disk JSONL of captured URLs
-
-### OAuth testing flow
-
-For **Cline OAuth** (SDK local callback): The SDK starts a local HTTP server, the auth URL
-is captured. To complete: open the captured URL in a real browser (it redirects back to the
-SDK's callback server), OR extract the callback port and `curl http://127.0.0.1:PORT/callback?code=...`.
-
-For **MCP/Provider OAuth** (vscode:// URI): The redirect goes to a vscode:// URI.
-`oauth.simulate_callback` only *builds* the URI — it does not deliver it, and the ESM
-extension host can't `require()` the handler. To actually deliver the callback, call the
-debug-only hook via `ext.evaluate` (with `awaitPromise: true`):
-`globalThis.__clineHandleUri("vscode://saoudrizwan.claude-dev/...?code=...&state=...")`.
-It runs the same `SharedUriHandler.handleUri` as VSCode's real URI handler and exists only
-when `CLINE_CAPTURE_BROWSER` is set (the harness always sets it; never ships in prod).
-For end-to-end MCP OAuth, get a real `code` from the local MCP OAuth test server
-(`npm run dev:mcp-oauth-test-server`).
-
-## Navigating Views — Use Commands, Not Clicks
-
-Don't try to find/click small sidebar icons. Use VSCode commands via command palette.
-Registered in `src/registry.ts`:
-
-| Command | View |
-|---------|------|
-| `cline.accountButtonClicked` | Account / sign-in |
-| `cline.historyButtonClicked` | Task history |
-| `cline.settingsButtonClicked` | Settings |
-| `cline.mcpButtonClicked` | MCP servers |
-| `cline.plusButtonClicked` | New task (chat) |
-| `cline.worktreesButtonClicked` | Worktrees |
-
-```bash
-curl localhost:19229/api -d '{"method":"ui.command_palette","params":{"command":"cline.accountButtonClicked"}}'
-```
-
-## Key commands
-
-All via `POST localhost:19229/api` with `{"method":"...", "params":{...}}`:
-
-- **`launch`** / **`shutdown`** — lifecycle
-- **`ui.screenshot`** — screenshot to `/tmp/cline-debug/`; returns `{path}` — **use `read_file` on the path to examine, do NOT `open` the file** (Preview.app covers the VSCode window)
-- **`ui.open_sidebar`** — open the Cline sidebar
-- **`ext.set_breakpoint`** `{file, line, condition?}` — breakpoint by source file (sourcemap-resolved)
-- **`ext.evaluate`** `{expression, callFrameId?}` — eval in extension host
-- **`ext.resume`** / **`ext.step_over`** / **`ext.step_into`** — stepping
-- **`ext.call_stack`** — inspect when paused
-- **`web.evaluate`** `{expression}` — eval in webview
-- **`web.post_message`** `{message}` — send postMessage to extension host via exposed vsCodeApi
-- **`wait_for_pause`** `{timeout?}` — block until breakpoint hit
-- **`ui.locator`** `{role?, testId?, text?, frame?}` — Playwright locator (auto-retries on stale sidebar frame)
-- **`ui.react_input`** `{text, selector?, clear?, submit?}` — set React textarea value via `execCommand('insertText')`; works reliably across multiple tasks
-- **`ui.send_message`** `{text, images?, files?, responseType?}` — send chat message bypassing the textarea entirely (via gRPC postMessage)
-- **`ui.command_palette`** `{command}` — run VSCode command
-
-## Typical Session
-
-```bash
-# 1. Launch
-curl localhost:19229/api -d '{"method":"launch","params":{"skipBuild":true}}'
-
-# 2. Open sidebar + dismiss overlays (ALWAYS do this first)
-curl localhost:19229/api -d '{"method":"ui.open_sidebar"}'
-curl localhost:19229/api -d '{"method":"web.evaluate","params":{"expression":"document.querySelectorAll(\".sr-only\").forEach(el => el.parentElement?.click())"}}'
-
-# 3. Navigate to view
-curl localhost:19229/api -d '{"method":"ui.command_palette","params":{"command":"cline.accountButtonClicked"}}'
-
-# 4. Check captured OAuth URLs if testing auth
-curl localhost:19229/api -d '{"method":"oauth.captured_urls"}'
-
-# 5. Verify
-curl localhost:19229/api -d '{"method":"ui.screenshot"}'
-```
-
-## Caveats
-
-- **⚠️ Dismiss promotional overlays FIRST**: On fresh launches, full-screen promo overlays block the sidebar. **Dismiss immediately after `ui.open_sidebar`**, before any other interaction or screenshot. May need to run twice:
-  ```bash
-  curl localhost:19229/api -d '{"method": "ui.open_sidebar"}'
-  curl localhost:19229/api -d '{"method": "web.evaluate", "params": {"expression": "document.querySelectorAll(\".sr-only\").forEach(el => el.parentElement?.click())"}}'
-  ```
-- **Screenshots — don't open the file**: `ui.screenshot` and `ui.sidebar_screenshot` save PNGs to `/tmp/cline-debug/` and return the `{path}`. Use `read_file` on that path to examine screenshots. Running `open <path>` launches Preview.app on macOS which covers the VSCode window.
-- **Scripts count = 0 after launch**: CDP connects after extension host starts, so scripts parsed during startup aren't tracked. Breakpoints still work via sourcemap resolution.
-- **Port 9230**: Extension host inspector. If another VSCode instance uses this port, the harness will fail to connect. Kill other debug instances first.
-- **macOS only** for now (Playwright Electron launch behavior).
-- **Webview CDP**: `connect_webview` may fail depending on Electron version. `web.evaluate` still works via Playwright's `frame.evaluate()` fallback.
-- **Sourcemap paths**: esbuild outputs relative paths like `../src/extension.ts` in the sourcemap. The resolver handles this, but if a file isn't found, use `ext.source_files` to see exact paths.
-- **OAuth with fake codes**: Browser capture intercepts the URL but doesn't provide a valid auth code. For real OAuth testing, open the captured URL in a browser. For unit testing, mock the token exchange.
-
-See `src/dev/debug-harness/README.md` for full API reference.

.clinerules/general.md

@@ -13,55 +13,11 @@ This file is the secret sauce for working effectively in this codebase. It captu
 **What NOT to add:** Stuff you can figure out from reading a few files, obvious patterns, or standard practices. This file should be high-signal, not comprehensive.
 
 ## Miscellaneous
-- Avoid provider-specific string matching / hardcoded provider branches when fixing provider/config plumbing. Prefer provider metadata, shared catalog/defaults, explicit protocol/client capabilities, or centralized normalization utilities that apply by data shape rather than `providerId === "..."`. If a provider exception seems necessary, stop and explain why instead of adding ad-hoc string matching.
 - This is a VS Code extension—check `package.json` for available scripts before trying to verify builds (e.g., `npm run compile`, not `npm run build`).
 - When creating PRs, contributors should not create changelog-entry files. Maintainers handle release versioning and changelog curation during the release process.
 - When adding new feature flags, see this PR as a reference https://github.com/cline/cline/pull/7566
 - Additional instructions about making requests: @.clinerules/network.md
 
-## Searching the Codebase — Avoiding Build Output
-
-Several directories contain build output or generated code that produces
-noisy or unusable results with `search_files` / `grep`:
-
-| Directory | What it is | Why it's a problem |
-|-----------|-----------|-------------------|
-| `out/` | esbuild bundle output | Mirrors `src/` structure as minified JS — every search gets duplicate hits on single-line files |
-| `dist/` | Packaged extension | Entire extension bundled into one minified `extension.js` (~1 long line) |
-| `dist-standalone/` | Standalone build output | Same minification issue |
-| `src/generated/` | Generated protobuf code | Auto-generated from `proto/`; not the source of truth |
-| `src/shared/proto/` | Generated proto type defs | Auto-generated from `proto/`; not the source of truth |
-| `node_modules/` | Dependencies | Huge, not project source |
-
-### How to skip build output
-
-**`search_files`** — Point at `src/` (not the project root) and use `file_pattern`:
-```
-search_files(path="src/core", regex="myFunction", file_pattern="*.ts")
-```
-The `file_pattern` parameter is the most effective filter — e.g. `"*.ts"`,
-`"*.tsx"`, `"*.proto"`.
-
-**`grep` directly** — Exclude build dirs and restrict to source extensions:
-```bash
-grep -rn "myFunction" src/ --include="*.ts" --exclude-dir={out,dist,node_modules,generated}
-```
-
-### When you must search minified files
-
-Sometimes you need to verify what got bundled (e.g., checking if a change
-made it into the build). Minified files are typically one long line, so
-normal `grep` shows the entire file as context. Use these approaches:
-
-- **`grep -oP`** to extract just the match with limited surrounding context:
-  ```bash
-  grep -oP '.{0,40}myFunction.{0,40}' dist/extension.js
-  ```
-- **`read_file`** on files in `out/src/` — these have source maps and are
-  more readable than `dist/extension.js` (which is the fully bundled output).
-- **Source maps** — `out/src/*.js.map` and `dist/extension.js.map` can be
-  used to trace minified output back to original source locations.
-
 ## gRPC/Protobuf Communication
 The extension and webview communicate via gRPC-like protocol over VS Code message passing.
 
@@ -92,6 +48,93 @@ The extension and webview communicate via gRPC-like protocol over VS Code messag
 - `src/core/controller/task/explainChanges.ts` - Handler implementation
 - `webview-ui/src/components/chat/ChatRow.tsx` - UI rendering
 
+## Adding a New API Provider
+When adding a new provider (e.g., "openai-codex"), you must update the proto conversion layer in THREE places or the provider will silently reset to Anthropic:
+
+1. `proto/cline/models.proto` - Add to the `ApiProvider` enum (e.g., `OPENAI_CODEX = 40;`)
+2. `convertApiProviderToProto()` in `src/shared/proto-conversions/models/api-configuration-conversion.ts` - Add case mapping string to proto enum
+3. `convertProtoToApiProvider()` in the same file - Add case mapping proto enum back to string
+
+**Why this matters:** Without these, the provider string hits the `default` case and returns `ANTHROPIC`. The webview, provider list, and handler all work fine, but the state silently resets when it round-trips through proto serialization. No error is thrown.
+
+**Other files to update when adding a provider:**
+- `src/shared/api.ts` - Add to `ApiProvider` union type, define models
+- `src/shared/providers/providers.json` - Add to provider list for dropdown
+- `src/core/api/index.ts` - Register handler in `createHandlerForProvider()`
+- `webview-ui/src/components/settings/utils/providerUtils.ts` - Add cases in `getModelsForProvider()` and `normalizeApiConfiguration()`
+- `webview-ui/src/utils/validate.ts` - Add validation case
+- `webview-ui/src/components/settings/ApiOptions.tsx` - Render provider component
+
+## Responses API Providers (OpenAI Codex, OpenAI Native)
+Providers using OpenAI's Responses API require native tool calling. XML tools don't work with the Responses API.
+
+**Symptoms of broken native tool calling:**
+- Tools get called multiple times (e.g., `ask_followup_question` asks the same question twice)
+- Tool arguments get duplicated or malformed
+- The model responds but tools aren't recognized
+
+**Root causes to check:**
+1. **Provider missing from `isNextGenModelProvider()`** in `src/utils/model-utils.ts`. The native variant matchers (e.g., `native-gpt-5/config.ts`) call this function. If your provider isn't in the list, the matcher returns false and falls back to XML tools.
+
+2. **Model missing `apiFormat: ApiFormat.OPENAI_RESPONSES`** in its model info (`src/shared/api.ts`). This property signals that the model requires native tool calling. The task runner in `src/core/task/index.ts` checks this and forces `enableNativeToolCalls: true` regardless of user settings.
+
+**When adding a new Responses API provider:**
+1. Add provider to `isNextGenModelProvider()` list in `src/utils/model-utils.ts`
+2. Set `apiFormat: ApiFormat.OPENAI_RESPONSES` on all models that use the Responses API
+3. The variant matcher and task runner will handle the rest automatically
+
+## Adding Tools to System Prompt
+This is tricky—multiple prompt variants and configs. **Always search for existing similar tools first and follow their pattern.** Look at the full chain from prompt definition → variant configs → handler → UI before implementing.
+
+1. **Add to `ClineDefaultTool` enum** in `src/shared/tools.ts`
+2. **Tool definition** in `src/core/prompts/system-prompt/tools/` (create file like `generate_explanation.ts`)
+   - Define variants for each `ModelFamily` (generic, next-gen, xs, etc.)
+   - Export variants array (e.g., `export const my_tool_variants = [GENERIC, NATIVE_NEXT_GEN, XS]`)
+   - **Fallback behavior**: If a variant isn't defined for a model family, `ClineToolSet.getToolByNameWithFallback()` automatically falls back to GENERIC. So you only need to export `[GENERIC]` unless the tool needs model-specific behavior.
+3. **Register in `src/core/prompts/system-prompt/tools/init.ts`** - Import and spread into `allToolVariants`
+4. **Add to variant configs** - Each model family has its own config in `src/core/prompts/system-prompt/variants/*/config.ts`. Add your tool's enum to the `.tools()` list:
+   - `generic/config.ts`, `next-gen/config.ts`, `gpt-5/config.ts`, `native-gpt-5/config.ts`, `native-gpt-5-1/config.ts`, `native-next-gen/config.ts`, `gemini-3/config.ts`, `glm/config.ts`, `hermes/config.ts`, `xs/config.ts`
+   - **Important**: If you add to a variant's config, make sure the tool spec exports a variant for that ModelFamily (or relies on GENERIC fallback)
+5. **Create handler** in `src/core/task/tools/handlers/`
+6. **Wire up in `ToolExecutor.ts`** if needed for execution flow
+7. **Add to tool parsing** in `src/core/assistant-message/index.ts` if needed
+8. **If tool has UI feedback**: add `ClineSay` enum in proto, update `src/shared/ExtensionMessage.ts`, update `src/shared/proto-conversions/cline-message.ts`, update `webview-ui/src/components/chat/ChatRow.tsx`
+
+## Modifying System Prompt
+**Read these first:** `src/core/prompts/system-prompt/README.md`, `tools/README.md`, `__tests__/README.md`
+
+System prompt is modular: **components** (reusable sections) + **variants** (model-specific configs) + **templates** (with `{{PLACEHOLDER}}` resolution).
+
+**Key directories:**
+- `components/` - Shared sections: `rules.ts`, `capabilities.ts`, `editing_files.ts`, etc.
+- `variants/` - Model-specific: `generic/`, `next-gen/`, `xs/`, `gpt-5/`, `gemini-3/`, `hermes/`, `glm/`, etc.
+- `templates/` - Template engine and placeholder definitions
+
+**Variant tiers (ask user which to modify):**
+- **Next-gen** (Claude 4, GPT-5, Gemini 2.5): `next-gen/`, `native-next-gen/`, `native-gpt-5/`, `native-gpt-5-1/`, `gemini-3/`, `gpt-5/`
+- **Standard** (default fallback): `generic/`
+- **Local/small models**: `xs/`, `hermes/`, `glm/`
+
+**How overrides work:** Variants can override components via `componentOverrides` in their `config.ts`, or provide a custom template in `template.ts` (e.g., `next-gen/template.ts` exports `rules_template`). If no override, the shared component from `components/` is used.
+
+**Example: Adding a rule to RULES section**
+1. Check if variant overrides rules: look for `rules_template` in `variants/*/template.ts` or `componentOverrides.RULES` in `config.ts`
+2. If shared: modify `components/rules.ts`
+3. If overridden: modify that variant's template
+4. XS variant is special—has heavily condensed inline content in `template.ts`
+
+**After any changes, regenerate snapshots:**
+```bash
+UPDATE_SNAPSHOTS=true npm run test:unit
+```
+Snapshots live in `__tests__/__snapshots__/`. Tests validate across model families and context variations (browser, MCP, focus chain).
+
+## Modifying Default Slash Commands
+Three places need updates:
+- `src/core/slash-commands/index.ts` - Command definitions
+- `src/core/prompts/commands.ts` - System prompt integration
+- `webview-ui/src/utils/slash-commands.ts` - Webview autocomplete
+
 ## Adding New Global State Keys
 Adding a new key to global state requires updates in multiple places. Missing any step causes silent failures.
 
@@ -160,48 +203,3 @@ const isGenerating = explanationInfo.status === "generating" && !wasCancelled
 **See also:** `BrowserSessionRow.tsx` uses similar pattern with `isLastApiReqInterrupted` and `isLastMessageResume`.
 
 **Backend side:** When streaming is cancelled, clean up properly (close tabs, clear comments, etc.) by checking `taskState.abort` after the streaming function returns.
-
-## Debug Harness: clear inherited VSCode/Electron env vars before launching
-
-The debug harness (`apps/vscode/src/dev/debug-harness/server.ts`) launches a child
-VSCode via Playwright's `_electron.launch({ env: { ...process.env, ... } })`. If you
-run the harness from a process that was itself spawned by VSCode (e.g. the Cline
-extension host, an integrated terminal, or an agent running inside VSCode), the
-parent's VSCode/Electron env vars leak into the child and break the launch.
-
-The fatal one is **`ELECTRON_RUN_AS_NODE=1`**: it makes the child VSCode binary run
-as plain Node, so it rejects every VSCode CLI flag. Symptom:
-
-```
-.../Visual Studio Code.app/Contents/MacOS/Code: bad option: --extensionDevelopmentPath=...
-Error: Process failed to launch!   (Playwright _electron.launch)
-```
-
-This is NOT the macOS Playwright flakiness mentioned in the harness README — it's
-env inheritance. Fix: strip the inherited vars before starting the harness:
-
-```bash
-env -u ELECTRON_RUN_AS_NODE -u ELECTRON_NO_ATTACH_CONSOLE \
-    -u VSCODE_CLI -u VSCODE_CODE_CACHE_PATH -u VSCODE_CRASH_REPORTER_PROCESS_TYPE \
-    -u VSCODE_CWD -u VSCODE_ESM_ENTRYPOINT -u VSCODE_HANDLES_UNCAUGHT_ERRORS \
-    -u VSCODE_IPC_HOOK -u VSCODE_NLS_CONFIG -u VSCODE_PID -u VSCODE_L10N_BUNDLE_LOCATION \
-    npx tsx src/dev/debug-harness/server.ts --auto-launch --skip-build
-```
-
-Check your own env with `env | grep -iE 'electron|vscode_'` first; `ELECTRON_RUN_AS_NODE=1`
-present means you must scrub before launching.
-
-Other harness notes confirmed in practice:
-- The extension host is **ESM** (`VSCODE_ESM_ENTRYPOINT`), so `ext.evaluate` has no
-  `require` and module-internal functions aren't reachable as globals. To inspect
-  internal builders (e.g. `buildBedrockProviderConfig`), set a breakpoint with
-  `ext.set_breakpoint` and read locals via `ext.evaluate` with the paused `callFrameId`
-  — don't try to `require()` the bundle.
-- `web.evaluate` wraps the expression as a single returned expression; multi-statement
-  snippets must be an IIFE `(() => { ...; return x; })()`, otherwise you get
-  `SyntaxError: Unexpected token ';'`.
-- Webview settings inputs are `vscode-text-field` web components with debounced React
-  onChange. Setting `.value` + dispatching events via `web.evaluate` is unreliable for
-  some fields; focus the inner shadow `input` then use real keystrokes (`ui.type` +
-  `ui.press Tab`, or click the dropdown option) to make the value persist.
-

.clinerules/sdk-migration.md

@@ -1,26 +0,0 @@
-# SDK Adapter
-
-The VSCode extension runs on the Cline SDK (`@cline/core`, `@cline/llms`,
-`@cline/shared`) through an adapter layer in `apps/vscode/src/sdk/`. The
-webview still talks gRPC; the adapter translates between gRPC handlers and SDK
-calls. See `apps/vscode/src/dev/debug-harness/README.md` for the debug harness.
-
-## Conventions
-
-1. **Look up SDK APIs, don't guess.** Use `kb_search(name="sdk", query="...")`
-   before implementing against an SDK surface.
-2. **Reference the pre-SDK implementation when replacing a module.** Add a
-   `// Replaces classic src/core/... (see origin/main)` header and use
-   `kb_search(name="cline", commit="origin/main")` or
-   `git show origin/main:path` to consult the prior implementation.
-3. **Single entry point.** There is one codepath — the SDK adapter. No
-   `CLINE_SDK` env flag.
-4. **Use `{appBaseUrl}`**, never hardcode `app.cline.bot`.
-5. **Avoid `as` casts.** Use explicit conversion functions with tests. The
-   branded types in `apps/vscode/src/sdk/model-catalog/contracts.ts` exist so
-   casts are unnecessary outside parse/compute boundaries.
-
-## Debug harness
-
-- **Dismiss the Kanban/promo overlay** before any debug harness interaction.
-- **Use the command palette** to navigate tabs in the debug harness.

.github/workflows/cli-publish.yml

@@ -33,7 +33,7 @@ permissions:
 
 defaults:
   run:
-    working-directory: sdk
+    working-directory: .
 
 jobs:
   publish-main:
@@ -102,7 +102,7 @@ jobs:
           fi
 
           VERSION="${TAG#cli-v}"
-          PACKAGE_VERSION=$(node -p "require('./apps/cli/package.json').version")
+          PACKAGE_VERSION=$(node -p "require('./sdk/apps/cli/package.json').version")
 
           if [ "$PACKAGE_VERSION" != "$VERSION" ]; then
             echo "sdk/apps/cli/package.json version ${PACKAGE_VERSION} does not match ${TAG}"
@@ -172,7 +172,7 @@ jobs:
           )
 
           for package_name in "${EXPECTED[@]}"; do
-            dir="apps/cli/dist/${package_name#@cline/}"
+            dir="sdk/apps/cli/dist/${package_name#@cline/}"
             if [ ! -f "$dir/package.json" ]; then
               echo "Missing package manifest: $dir/package.json"
               exit 1
@@ -207,8 +207,8 @@ jobs:
       - name: Get Changelog Entry
         id: changelog
         run: |
-          # Grab content between the first "## " header and the next one in apps/cli/CHANGELOG.md
-          CONTENT=$(awk '/^## [0-9]/{if(found) exit; found=1; next} found{print}' apps/cli/CHANGELOG.md)
+          # Grab content between the first "## " header and the next one in sdk/apps/cli/CHANGELOG.md
+          CONTENT=$(awk '/^## [0-9]/{if(found) exit; found=1; next} found{print}' sdk/apps/cli/CHANGELOG.md)
           echo "content<<EOF" >> $GITHUB_OUTPUT
           echo "$CONTENT" >> $GITHUB_OUTPUT
           echo "EOF" >> $GITHUB_OUTPUT
@@ -349,7 +349,7 @@ jobs:
         if: steps.check_commits.outputs.skip != 'true'
         id: version
         run: |
-          BASE_VERSION=$(node -p "require('./apps/cli/package.json').version")
+          BASE_VERSION=$(node -p "require('./sdk/apps/cli/package.json').version")
           TIMESTAMP=$(date +%s)
           VERSION="${BASE_VERSION}-nightly.${TIMESTAMP}"
 
@@ -365,12 +365,12 @@ jobs:
         run: |
           node -e '
             const fs = require("node:fs");
-            const path = "apps/cli/package.json";
+            const path = "sdk/apps/cli/package.json";
             const pkg = JSON.parse(fs.readFileSync(path, "utf8"));
             pkg.version = process.env.VERSION;
             fs.writeFileSync(path, `${JSON.stringify(pkg, null, "\t")}\n`);
           '
-          cat apps/cli/package.json | grep '"version"'
+          cat sdk/apps/cli/package.json | grep '"version"'
 
       - name: Build platform binaries
         if: steps.check_commits.outputs.skip != 'true'
@@ -401,7 +401,7 @@ jobs:
           )
 
           for package_name in "${EXPECTED[@]}"; do
-            dir="apps/cli/dist/${package_name#@cline/}"
+            dir="sdk/apps/cli/dist/${package_name#@cline/}"
             if [ ! -f "$dir/package.json" ]; then
               echo "Missing package manifest: $dir/package.json"
               exit 1

.github/workflows/ext-vscode-publish-nightly.yml

@@ -31,12 +31,6 @@ jobs:
     if: github.repository == 'cline/cline' && (github.ref == 'refs/heads/main' || github.ref == 'refs/heads/dpc/sdk-migration-simpler-login')
     runs-on: ubuntu-latest
     environment: PublishNightly
-    # The VS Code extension's package.json and lockfiles live under apps/vscode/
-    # (the repo root has no package.json). Mirror ext-vscode-test.yml so install
-    # and publish steps run in the correct workspace.
-    defaults:
-      run:
-        working-directory: apps/vscode
 
     steps:
       - name: Checkout selected branch
@@ -59,10 +53,6 @@ jobs:
           # Keep publish environment aligned with test workflow/tooling lockfile expectations.
           # Newer LTS (Node 24 / npm 11) can make `npm list` fail with ELSPROBLEMS during vsce packaging.
           node-version: 22
-          cache: 'npm'
-          cache-dependency-path: |
-            apps/vscode/package-lock.json
-            apps/vscode/webview-ui/package-lock.json
 
       - name: Install root dependencies
         run: npm ci --include=optional

.github/workflows/ext-vscode-test.yml

@@ -92,12 +92,12 @@ jobs:
                       apps/vscode/webview-ui/package-lock.json
 
             - name: Install root dependencies
-              run: npm ci --include=optional
+              run: npm ci
               env:
                   GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
 
             - name: Install webview-ui dependencies
-              run: cd webview-ui && npm ci --include=optional
+              run: cd webview-ui && npm ci
 
             - name: Run Quality Checks (Parallel)
               run: npm run ci:check-all
@@ -131,12 +131,12 @@ jobs:
                       apps/vscode/webview-ui/package-lock.json
 
             - name: Install root dependencies
-              run: npm ci --include=optional
+              run: npm ci
               env:
                   GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
 
             - name: Install webview-ui dependencies
-              run: cd webview-ui && npm ci --include=optional
+              run: cd webview-ui && npm ci
 
             - name: Set up NPM on Windows
               if: runner.os == 'Windows'
@@ -155,11 +155,6 @@ jobs:
               id: build_step
               run: npm run ci:build
 
-            - name: Vitest Suites (SDK adapter + model catalog)
-              id: vitest_tests
-              if: ${{ !cancelled() && steps.build_step.outcome == 'success' }}
-              run: npm run test:vitest
-
             - name: Unit Tests with coverage - Linux
               id: unit_tests_linux
               if: ${{ !cancelled() && steps.build_step.outcome == 'success' && runner.os == 'Linux' }}
@@ -235,12 +230,12 @@ jobs:
                       apps/vscode/testing-platform/package-lock.json
 
             - name: Install root dependencies
-              run: npm ci --include=optional
+              run: npm ci
               env:
                   GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
 
             - name: Install webview-ui dependencies
-              run: cd webview-ui && npm ci --include=optional
+              run: cd webview-ui && npm ci
 
             - name: Download ripgrep binaries
               run: npm run download-ripgrep
@@ -249,7 +244,7 @@ jobs:
               run: npm run compile-standalone
 
             - name: Install testing platform dependencies
-              run: cd testing-platform && npm ci --include=optional
+              run: cd testing-platform && npm ci
 
             - name: Running testing platform integration spec tests
               timeout-minutes: 7

.github/workflows/sdk-publish.yml

@@ -26,7 +26,7 @@ on:
 
 defaults:
   run:
-    working-directory: sdk
+    working-directory: .
 
 jobs:
   test:
@@ -148,7 +148,7 @@ jobs:
         env:
           CHANNEL: ${{ steps.channel.outputs.channel }}
         run: |
-          BASE_VERSION=$(node -p "require('./packages/llms/package.json').version")
+          BASE_VERSION=$(node -p "require('./sdk/packages/llms/package.json').version")
 
           if [ "$CHANNEL" = "nightly" ]; then
             TIMESTAMP=$(date +%s)
@@ -166,11 +166,11 @@ jobs:
         if: steps.check_commits.outputs.skip != 'true'
         env:
           VERSION: ${{ steps.version.outputs.version }}
-        run: bun scripts/version.ts "$VERSION"
+        run: bun sdk/scripts/version.ts "$VERSION"
 
       - name: Verify publishability
         if: steps.check_commits.outputs.skip != 'true'
-        run: bun scripts/check-publish.ts
+        run: bun sdk/scripts/check-publish.ts
 
       - name: Prepare package tarball directory
         if: steps.check_commits.outputs.skip != 'true'
@@ -187,7 +187,7 @@ jobs:
           VERSION: ${{ steps.version.outputs.version }}
         run: |
           echo "Publishing @cline/shared@${VERSION} with tag '${CHANNEL}'..."
-          cd packages/shared
+          cd sdk/packages/shared
           TARBALL=$(bun pm pack --destination "$RUNNER_TEMP/sdk-npm-packs" --quiet)
           npm publish "$RUNNER_TEMP/sdk-npm-packs/$(basename "$TARBALL")" --tag "$CHANNEL" --access public
 
@@ -199,7 +199,7 @@ jobs:
           VERSION: ${{ steps.version.outputs.version }}
         run: |
           echo "Publishing @cline/llms@${VERSION} with tag '${CHANNEL}'..."
-          cd packages/llms
+          cd sdk/packages/llms
           TARBALL=$(bun pm pack --destination "$RUNNER_TEMP/sdk-npm-packs" --quiet)
           npm publish "$RUNNER_TEMP/sdk-npm-packs/$(basename "$TARBALL")" --tag "$CHANNEL" --access public
 
@@ -211,7 +211,7 @@ jobs:
           VERSION: ${{ steps.version.outputs.version }}
         run: |
           echo "Publishing @cline/agents@${VERSION} with tag '${CHANNEL}'..."
-          cd packages/agents
+          cd sdk/packages/agents
           TARBALL=$(bun pm pack --destination "$RUNNER_TEMP/sdk-npm-packs" --quiet)
           npm publish "$RUNNER_TEMP/sdk-npm-packs/$(basename "$TARBALL")" --tag "$CHANNEL" --access public
 
@@ -223,7 +223,7 @@ jobs:
           VERSION: ${{ steps.version.outputs.version }}
         run: |
           echo "Publishing @cline/core@${VERSION} with tag '${CHANNEL}'..."
-          cd packages/core
+          cd sdk/packages/core
           TARBALL=$(bun pm pack --destination "$RUNNER_TEMP/sdk-npm-packs" --quiet)
           npm publish "$RUNNER_TEMP/sdk-npm-packs/$(basename "$TARBALL")" --tag "$CHANNEL" --access public
 
@@ -235,7 +235,7 @@ jobs:
           VERSION: ${{ steps.version.outputs.version }}
         run: |
           echo "Publishing @cline/sdk@${VERSION} with tag '${CHANNEL}'..."
-          cd packages/sdk
+          cd sdk/packages/sdk
           TARBALL=$(bun pm pack --destination "$RUNNER_TEMP/sdk-npm-packs" --quiet)
           npm publish "$RUNNER_TEMP/sdk-npm-packs/$(basename "$TARBALL")" --tag "$CHANNEL" --access public
 

.github/workflows/sdk-test.yml

@@ -21,7 +21,7 @@ permissions:
 
 defaults:
     run:
-        working-directory: sdk
+        working-directory: .
 
 jobs:
     quality-checks:
@@ -96,12 +96,12 @@ jobs:
 
             - name: Run SDK Tests (Windows)
               if: ${{ !cancelled() && steps.build_sdk_step.outcome == 'success' && steps.build_cli_step.outcome == 'success' && matrix.os == 'windows-latest' }}
-              run: bun -F './packages/**' test
+              run: bun -F './sdk/packages/**' test
 
             - name: Smoke test SQLite under Node
               if: ${{ !cancelled() && steps.build_sdk_step.outcome == 'success' && matrix.os != 'windows-latest' }}
               timeout-minutes: 10
-              run: bun scripts/ci-node-smoke.ts
+              run: bun sdk/scripts/ci-node-smoke.ts
 
             - name: Run TUI e2e tests
               if: ${{ !cancelled() && steps.build_sdk_step.outcome == 'success' && steps.build_cli_step.outcome == 'success' && matrix.os == 'ubuntu-latest' && matrix.node-version == '24.x' }}
@@ -109,4 +109,4 @@ jobs:
 
             - name: Verify packages are publishable
               if: ${{ !cancelled() && steps.build_sdk_step.outcome == 'success' && steps.build_cli_step.outcome == 'success' && matrix.os == 'ubuntu-latest' && matrix.node-version == '24.x' }}
-              run: bun scripts/check-publish.ts
+              run: bun sdk/scripts/check-publish.ts

.gitignore

@@ -13,9 +13,6 @@ tmp
 pnpm-lock.yaml
 
 .clineignore
-.cline/enterprise
-.cline/remote-config
-**/.cline/remote-config
 .venv
 .actrc
 CLAUDE.local.md

.husky/pre-commit

@@ -1 +1,10 @@
-cd apps/vscode && lint-staged
+if ! command -v gitleaks >/dev/null 2>&1; then
+	echo "gitleaks is required for the pre-commit secret scan."
+	echo "Install it with: brew install gitleaks"
+	echo "Other install options: https://github.com/gitleaks/gitleaks#installing"
+	exit 1
+fi
+
+gitleaks git --pre-commit --redact --staged --verbose || exit 1
+
+lint-staged

.vscode/tasks.json

@@ -5,8 +5,8 @@
 	"tasks": [
 		{
 			"label": "compile-standalone",
-			"type": "npm",
-			"script": "compile-standalone",
+			"type": "shell",
+			"command": "npm run compile-standalone",
 			"group": "build",
 			"problemMatcher": [],
 			"presentation": {
@@ -18,8 +18,8 @@
 		},
 		{
 			"label": "npm: protos",
-			"type": "npm",
-			"script": "protos",
+			"type": "shell",
+			"command": "npm run protos",
 			"problemMatcher": [],
 			"isBackground": false,
 			"presentation": {
@@ -64,8 +64,8 @@
 			"group": "build"
 		},
 		{
-			"type": "npm",
-			"script": "build:webview",
+			"type": "shell",
+			"command": "npm run build:webview",
 			"group": "build",
 			"problemMatcher": [],
 			"isBackground": true,
@@ -85,8 +85,8 @@
 			}
 		},
 		{
-			"type": "npm",
-			"script": "build:webview:test",
+			"type": "shell",
+			"command": "npm run build:webview:test",
 			"group": "build",
 			"problemMatcher": [],
 			"isBackground": true,
@@ -107,8 +107,8 @@
 			}
 		},
 		{
-			"type": "npm",
-			"script": "dev:webview",
+			"type": "shell",
+			"command": "npm run dev:webview",
 			"group": "build",
 			"problemMatcher": [
 				{
@@ -144,8 +144,8 @@
 			}
 		},
 		{
-			"type": "npm",
-			"script": "watch:esbuild",
+			"type": "shell",
+			"command": "npm run watch:esbuild",
 			"group": "build",
 			"problemMatcher": {
 				"pattern": [
@@ -183,8 +183,8 @@
 			}
 		},
 		{
-			"type": "npm",
-			"script": "watch:esbuild:test",
+			"type": "shell",
+			"command": "npm run watch:esbuild:test",
 			"group": "build",
 			"problemMatcher": {
 				"pattern": [
@@ -223,8 +223,8 @@
 			}
 		},
 		{
-			"type": "npm",
-			"script": "watch:tsc",
+			"type": "shell",
+			"command": "npm run watch:tsc",
 			"group": "build",
 			"problemMatcher": "$tsc-watch",
 			"isBackground": true,
@@ -241,8 +241,9 @@
 			}
 		},
 		{
-			"type": "npm",
-			"script": "watch-tests",
+			"type": "shell",
+			"command": "npm run watch-tests",
+			"label": "npm: watch-tests",
 			"problemMatcher": "$tsc-watch",
 			"isBackground": true,
 			"dependsOn": [
@@ -280,8 +281,8 @@
 			"command": "rm -rf ${workspaceFolder}/apps/vscode/dist/tmp/user && mkdir -p ${workspaceFolder}/apps/vscode/dist/tmp/user"
 		},
 		{
-			"type": "npm",
-			"script": "storybook",
+			"type": "shell",
+			"command": "npm run storybook",
 			"group": "build",
 			"problemMatcher": [],
 			"isBackground": false,
@@ -308,7 +309,7 @@
 				"$tsc"
 			],
 			"options": {
-				"cwd": "${workspaceFolder}/sdk"
+				"cwd": "${workspaceFolder}"
 			}
 		}
 	],

CLAUDE.md

@@ -1 +1,2 @@
 @.clinerules/general.md
+@.clinerules/network.md

apps/vscode/.mocharc.json

@@ -6,13 +6,6 @@
 		"src/**/__tests__/*.ts",
 		"src/test/services/**/*.test.ts"
 	],
-	"ignore": [
-		"src/core/controller/models/__tests__/providerCatalogHandlers.test.ts",
-		"src/core/controller/models/__tests__/providerCatalogSmoke.test.ts",
-		"src/core/controller/models/__tests__/providerSwitchNormalization.test.ts",
-		"src/core/controller/models/__tests__/resolveModelInfo.test.ts",
-		"src/core/controller/models/__tests__/refreshClineRecommendedModels.test.ts"
-	],
 	"require": [
 		"ts-node/register",
 		"source-map-support/register",

apps/vscode/.vscode-test.mjs

@@ -1,13 +1,9 @@
 import { defineConfig } from "@vscode/test-cli"
 import path from "path"
-
 const vscodeTestVersion = process.env.VSCODE_TEST_VERSION ?? "stable"
 
 export default defineConfig({
-	files: [
-		"out/src/{core,test,utils,shared,integrations,hosts,services}/**/*.test.js",
-		"src/{core,test,utils,shared,integrations,hosts,services}/**/*.test.js",
-	],
+	files: "{out/**/*.test.js,src/**/*.test.js,!src/test/e2e/**/*.test.js,!out/src/test/e2e/**/*.test.js}",
 	mocha: {
 		ui: "bdd",
 		timeout: 20000, // Maximum time (in ms) that a test can run before failing

apps/vscode/esbuild.mjs

@@ -85,6 +85,44 @@ const esbuildProblemMatcherPlugin = {
 	},
 }
 
+const copyWasmFiles = {
+	name: "copy-wasm-files",
+	setup(build) {
+		build.onEnd(() => {
+			// tree sitter
+			const sourceDir = path.join(__dirname, "node_modules", "web-tree-sitter")
+			const targetDir = path.join(__dirname, destDir)
+
+			// Copy tree-sitter.wasm
+			fs.copyFileSync(path.join(sourceDir, "tree-sitter.wasm"), path.join(targetDir, "tree-sitter.wasm"))
+
+			// Copy language-specific WASM files
+			const languageWasmDir = path.join(__dirname, "node_modules", "tree-sitter-wasms", "out")
+			const languages = [
+				"typescript",
+				"tsx",
+				"python",
+				"rust",
+				"javascript",
+				"go",
+				"cpp",
+				"c",
+				"c_sharp",
+				"ruby",
+				"java",
+				"php",
+				"swift",
+				"kotlin",
+			]
+
+			languages.forEach((lang) => {
+				const filename = `tree-sitter-${lang}.wasm`
+				fs.copyFileSync(path.join(languageWasmDir, filename), path.join(targetDir, filename))
+			})
+		})
+	},
+}
+
 const buildEnvVars = {
 	"import.meta.url": "_importMetaUrl",
 	"process.env.IS_STANDALONE": JSON.stringify(standalone ? "true" : "false"),
@@ -138,6 +176,7 @@ const baseConfig = {
 	define: buildEnvVars,
 	tsconfig: path.resolve(__dirname, "tsconfig.json"),
 	plugins: [
+		copyWasmFiles,
 		aliasResolverPlugin,
 		/* add to the end of plugins array */
 		esbuildProblemMatcherPlugin,

apps/vscode/package.json

@@ -410,12 +410,9 @@
 		"test": "npx npm-run-all test:unit test:integration",
 		"test:integration": "vscode-test",
 		"test:unit": "cross-env TS_NODE_PROJECT=./tsconfig.unit-test.json mocha",
-		"test:vitest": "vitest run --config vitest.config.ts",
-		"test:vitest:watch": "vitest --config vitest.config.ts",
 		"test:coverage": "vscode-test --coverage",
 		"test:sca-server": "npx tsx watch scripts/test-standalone-core-api-server.ts",
 		"test:tp-orchestrator": "npx tsx scripts/testing-platform-orchestrator.ts",
-		"dev:mcp-oauth-test-server": "npx tsx src/dev/mcp-oauth-test-server/server.ts",
 		"e2e": "playwright test -c playwright.config.ts",
 		"test:e2e:build": "vsce package --allow-package-secrets sendgrid --out dist/e2e.vsix",
 		"test:e2e": "playwright install && npm run test:e2e:build && node src/test/e2e/utils/build.mjs && playwright test",
@@ -485,24 +482,25 @@
 		"rimraf": "^6.0.1",
 		"should": "^13.2.3",
 		"sinon": "^21.0.3",
-		"tar": "^7.5.2",
 		"tree-kill": "^1.2.2",
 		"ts-node": "^10.9.2",
 		"ts-proto": "^2.6.1",
 		"tsconfig-paths": "^4.2.0",
-		"typescript": "^5.4.5",
-		"vitest": "^4.0.17"
+		"typescript": "^5.4.5"
 	},
 	"dependencies": {
 		"@anthropic-ai/sdk": "^0.37.0",
+		"@anthropic-ai/vertex-sdk": "^0.6.4",
+		"@aws-sdk/client-bedrock-runtime": "^3.922.0",
+		"@aws-sdk/credential-providers": "^3.922.0",
+		"@azure/identity": "^4.13.0",
 		"@bufbuild/protobuf": "^2.2.5",
-		"@cline/agents": "^0.0.42",
-		"@cline/core": "^0.0.42",
-		"@cline/llms": "^0.0.42",
-		"@cline/shared": "^0.0.42",
+		"@cerebras/cerebras_cloud_sdk": "^1.35.0",
+		"@google-cloud/vertexai": "^1.9.3",
 		"@google/genai": "^1.30.0",
 		"@grpc/grpc-js": "^1.9.15",
 		"@grpc/reflection": "^1.0.4",
+		"@mistralai/mistralai": "^1.5.0",
 		"@modelcontextprotocol/sdk": "^1.25.1",
 		"@opentelemetry/api": "^1.9.0",
 		"@opentelemetry/core": "^2.1.0",
@@ -524,6 +522,9 @@
 		"@opentelemetry/sdk-trace-node": "^1.30.1",
 		"@opentelemetry/semantic-conventions": "^1.37.0",
 		"@playwright/test": "^1.55.1",
+		"@sap-ai-sdk/ai-api": "^2.7.0",
+		"@sap-ai-sdk/orchestration": "^2.7.0",
+		"@sap-cloud-sdk/connectivity": "^4.6.0",
 		"@streamparser/json": "^0.0.22",
 		"@tailwindcss/vite": "^4.1.14",
 		"@types/uuid": "^10.0.0",
@@ -549,14 +550,13 @@
 		"ignore": "^7.0.3",
 		"image-size": "^2.0.2",
 		"isbinaryfile": "^5.0.2",
-		"js-yaml": "^4.1.1",
 		"jschardet": "^3.1.4",
 		"json5": "^2.2.3",
-		"jwt-decode": "^4.0.0",
 		"mammoth": "^1.11.0",
 		"nanoid": "^5.1.6",
 		"nice-grpc": "^2.1.12",
 		"node-machine-id": "^1.1.12",
+		"ollama": "^0.5.13",
 		"open": "^10.1.2",
 		"open-graph-scraper": "^6.9.0",
 		"openai": "^6.21.0",
@@ -575,13 +575,15 @@
 		"simple-git": "3.36.0",
 		"strip-ansi": "^7.1.2",
 		"tailwindcss": "^4.1.14",
+		"tree-sitter-wasms": "^0.1.11",
 		"ts-morph": "^25.0.1",
 		"turndown": "^7.2.0",
 		"ulid": "^2.4.0",
 		"undici": "^7.26.0",
 		"uuid": "^11.1.0",
 		"vscode-uri": "^3.1.0",
-		"zod": "^4.3.6"
+		"web-tree-sitter": "^0.22.6",
+		"zod": "^3.24.2"
 	},
 	"overrides": {
 		"tar-fs": ">=3.1.1",

apps/vscode/proto/cline/models.proto

@@ -21,6 +21,8 @@ service ModelsService {
   rpc refreshOpenRouterModelsRpc(EmptyRequest) returns (OpenRouterCompatibleModelInfo);
   // Refreshes and returns recommended and free Cline models
   rpc refreshClineRecommendedModelsRpc(EmptyRequest) returns (ClineRecommendedModelsResponse);
+  // Refreshes and returns Cline provider models
+  rpc refreshClineModelsRpc(EmptyRequest) returns (OpenRouterCompatibleModelInfo);
   // Refreshes and returns Hugging Face models
   rpc refreshHuggingFaceModels(EmptyRequest) returns (OpenRouterCompatibleModelInfo);
   // Refreshes and returns OpenAI models
@@ -53,18 +55,6 @@ service ModelsService {
   rpc getAihubmixModels(EmptyRequest) returns (OpenRouterCompatibleModelInfo);
   // Refreshes and returns Vercel AI Gateway models
   rpc refreshVercelAiGatewayModelsRpc(EmptyRequest) returns (OpenRouterCompatibleModelInfo);
-  // Lists providers available from the unified SDK-backed catalog
-  rpc listProviders(Empty) returns (ProviderListingsResponse);
-  // Resolves model metadata for a provider through the unified SDK-backed catalog
-  rpc resolveProviderModels(ResolveProviderModelsRequest) returns (ProviderModelsResponse);
-  // Resolves model metadata for a provider/model without refreshing model lists
-  rpc resolveModelInfo(ResolveModelInfoRequest) returns (ResolveModelInfoResponse);
-  // Reads redacted effective provider configuration
-  rpc readProviderConfig(StringRequest) returns (ProviderConfigResponse);
-  // Writes provider configuration fields and returns redacted effective configuration
-  rpc writeProviderConfig(WriteProviderConfigRequest) returns (ProviderConfigResponse);
-  // Commits a mode-specific model selection atomically with its model metadata
-  rpc commitModelSelection(CommitModelSelectionRequest) returns (Empty);
 }
 
 // List of VS Code LM models
@@ -127,117 +117,6 @@ message OpenRouterCompatibleModelInfo {
   map<string, OpenRouterModelInfo> models = 1;
 }
 
-// Lightweight provider entry for the top-level model/provider picker.
-// Does not include the full model list; use resolveProviderModels for models.
-message ProviderListing {
-  string id = 1;
-  string name = 2;
-  optional string default_model_id = 3;
-  optional string family = 4;
-  optional string protocol = 5;
-  optional string auth_description = 6;
-  optional string base_url_description = 7;
-  bool allows_custom_model_ids = 8;
-
-  // SDK-driven hint for cost display. Values: "show" (default) or "hide".
-  // Sourced from `resolveProviderUsageCostDisplay(provider.metadata)` in
-  // `@cline/llms`. When "hide", consumers must suppress per-token pricing
-  // and total cost displays (matches the CLI's `shouldShowCliUsageCost`).
-  string usage_cost_display = 11;
-}
-
-message ProviderListingsResponse {
-  repeated ProviderListing providers = 1;
-}
-
-message ResolveProviderModelsRequest {
-  string provider_id = 1;
-  bool force_refresh = 2;
-  optional string request_id = 3;
-}
-
-message CatalogErrorInfo {
-  string kind = 1;
-  string message = 2;
-  optional string code = 3;
-  optional bool retryable = 4;
-}
-
-message ProviderModelsResponse {
-  string provider_id = 1;
-  string request_id = 2;
-  string config_fingerprint = 3;
-  int64 fetched_at = 4;
-  bool ok = 5;
-  map<string, OpenRouterModelInfo> models = 6;
-  optional string default_model_id = 7;
-  optional string source = 8;
-  optional CatalogErrorInfo error = 9;
-}
-
-message ResolveModelInfoRequest {
-  string provider_id = 1;
-  optional string model_id = 2;
-}
-
-message ResolveModelInfoResponse {
-  string provider_id = 1;
-  string model_id = 2;
-  optional OpenRouterModelInfo model_info = 3;
-  string source = 4;
-}
-
-message ProviderConfigResponse {
-  string provider_id = 1;
-  optional string base_url = 2;
-  optional string api_line = 3;
-  map<string, string> headers = 4;
-  optional string region = 5;
-  int64 api_key_length = 6;
-  bool has_access_token = 7;
-  bool has_refresh_token = 8;
-  optional string account_id = 9;
-  optional CommittedModelSelection plan_selection = 10;
-  optional CommittedModelSelection act_selection = 11;
-}
-
-message CommittedModelSelection {
-  string provider_id = 1;
-  string model_id = 2;
-  OpenRouterModelInfo model_info = 3;
-}
-
-message ProviderReasoningPatch {
-  optional bool enabled = 1;
-  optional string effort = 2;   // "none" | "low" | "medium" | "high" | "xhigh"
-  optional int32 budget_tokens = 3;
-}
-
-message WriteProviderConfigPatch {
-  optional string api_key = 1;
-  optional string base_url = 2;
-  map<string, string> headers = 3;
-  optional string region = 4;
-  optional string api_line = 5;
-  optional string access_token = 6;
-  optional string refresh_token = 7;
-  optional string account_id = 8;
-  optional ProviderReasoningPatch reasoning = 9;
-  optional bool clear_headers = 10;
-}
-
-message WriteProviderConfigRequest {
-  string provider_id = 1;
-  WriteProviderConfigPatch patch = 2;
-}
-
-message CommitModelSelectionRequest {
-  string provider_id = 1;
-  string mode = 2;
-  string model_id = 3;
-  OpenRouterModelInfo model_info = 4;
-}
-
 message ClineRecommendedModel {
   string id = 1;
   string name = 2;

apps/vscode/proto/cline/remote_config.proto

@@ -1,32 +0,0 @@
-syntax = "proto3";
-
-package cline;
-
-import "cline/common.proto";
-
-option go_package = "github.com/cline/grpc-go/cline";
-option java_multiple_files = true;
-option java_package = "bot.cline.proto";
-
-enum RemoteConfigType {
-  RULE = 0;
-  WORKFLOW = 1;
-  SKILL = 2;
-}
-
-message RemoteConfigSetting {
-  RemoteConfigType type = 1;
-  string name = 2;
-  string content = 3;
-  bool enabled = 4;
-  bool locked = 5;
-}
-
-message RemoteConfigSettingsResponse {
-  repeated RemoteConfigSetting settings = 1;
-}
-
-service RemoteConfigService {
-  rpc getRemoteConfigSettings(Empty) returns (RemoteConfigSettingsResponse);
-  rpc toggleRemoteConfigSetting(StringRequest) returns (RemoteConfigSetting);
-}

apps/vscode/proto/cline/state.proto

@@ -250,6 +250,7 @@ message Settings {
   optional string default_terminal_profile = 137;
   optional int32 terminal_output_line_limit = 138;
   optional int32 max_consecutive_mistakes = 139;
+  optional bool strict_plan_mode_enabled = 141;
   optional bool yolo_mode_toggled = 142;
   optional bool use_auto_condense = 143;
   optional bool cline_web_tools_enabled = 144;
@@ -285,6 +286,7 @@ message Settings {
   optional string act_mode_cline_model_id = 180;
   optional OpenRouterModelInfo act_mode_cline_model_info = 181;
   optional bool show_feature_tips = 182;
+  optional bool lazy_teammate_mode_enabled = 183;
 }
 
 message State {
@@ -389,7 +391,6 @@ message UpdateSettingsRequest {
   reserved 15; // was openai_reasoning_effort (moved to mode-scoped reasoning effort)
   reserved 23; // was dictation_settings (dictation removed)
   reserved 38; // was skills_enabled (removed - now always enabled)
-  reserved 43; // was lazy_teammate_mode_enabled (removed)
 
   Metadata metadata = 1;
   optional ModelsApiConfiguration api_configuration = 2;
@@ -404,7 +405,7 @@ message UpdateSettingsRequest {
   optional int32 terminal_output_line_limit = 12;
   optional PlanActMode mode = 13;
   optional string preferred_language = 14;
-  reserved 16; // was strict_plan_mode_enabled (removed)
+  optional bool strict_plan_mode_enabled = 16;
   optional FocusChainSettings focus_chain_settings = 17;
   optional bool use_auto_condense = 18;
   optional string custom_prompt = 19;
@@ -428,6 +429,7 @@ message UpdateSettingsRequest {
   optional bool worktrees_enabled = 40;
   optional bool double_check_completion_enabled = 41;
   optional bool show_feature_tips = 42;
+  optional bool lazy_teammate_mode_enabled = 43;
 }
 
 message UpdateTerminalConnectionTimeoutRequest {

apps/vscode/proto/cline/task.proto

@@ -82,14 +82,12 @@ message GetTaskHistoryRequest {
   string search_query = 3;
   string sort_by = 4;
   bool current_workspace_only = 5;
-  int32 limit = 6;
-  int32 offset = 7;
 }
 
 // Response for task history
 message TaskHistoryArray {
   repeated TaskItem tasks = 1;
-  bool has_more = 2;
+  int32 total_count = 2;
 }
 
 // Task item details for history list

apps/vscode/proto/cline/ui.proto

@@ -226,12 +226,6 @@ message ClineMessage {
   ClineAskNewTask ask_new_task = 21;
   ClineApiReqInfo api_req_info = 22;
   ClineModelInfo model_info = 23;
-
-  // Convergent-replica fields (see webview-message-state-design.md):
-  // seq = monotonic freshness (higher seq wins for the same ts/identity)
-  // epoch = conversation/replica fence (older epoch is dropped by the webview)
-  int64 seq = 24;
-  int64 epoch = 25;
 }
 
 message ShowWebviewEvent {

apps/vscode/scripts/find-dead-src.mjs

@@ -1,129 +0,0 @@
-// Dead-source finder: uses esbuild's own bundle reachability (the same analysis
-// that drives tree-shaking + minification mangling) to compute which src/ files
-// are reachable from BOTH shipped entry points:
-//   - src/extension.ts            (VS Code extension host)
-//   - src/standalone/cline-core.ts (standalone host used by JetBrains + CLI)
-//
-// A src/*.ts file that is NOT in the union of metafile inputs for those two
-// builds is unreachable from any shipped entry => dead (modulo dynamic import()
-// of computed specifiers, which esbuild surfaces separately).
-//
-// Run: node scripts/find-dead-src.mjs
-
-import fs from "node:fs"
-import path from "node:path"
-import { fileURLToPath } from "node:url"
-import * as esbuild from "esbuild"
-import { glob } from "glob"
-
-const __dirname = path.dirname(fileURLToPath(import.meta.url))
-const root = path.resolve(__dirname, "..")
-
-const aliases = {
-	"@": path.join(root, "src"),
-	"@core": path.join(root, "src/core"),
-	"@integrations": path.join(root, "src/integrations"),
-	"@services": path.join(root, "src/services"),
-	"@shared": path.join(root, "src/shared"),
-	"@utils": path.join(root, "src/utils"),
-	"@packages": path.join(root, "src/packages"),
-}
-
-const aliasResolverPlugin = {
-	name: "alias-resolver",
-	setup(build) {
-		for (const [alias, aliasPath] of Object.entries(aliases)) {
-			const aliasRegex = new RegExp(`^${alias}($|/.*)`)
-			build.onResolve({ filter: aliasRegex }, (args) => {
-				const importPath = args.path.replace(alias, aliasPath)
-				const exts = [".ts", ".tsx", ".js", ".jsx"]
-				if (fs.existsSync(importPath)) {
-					const stats = fs.statSync(importPath)
-					if (stats.isDirectory()) {
-						for (const ext of exts) {
-							const idx = path.join(importPath, `index${ext}`)
-							if (fs.existsSync(idx)) return { path: idx }
-						}
-					} else {
-						return { path: importPath }
-					}
-				}
-				for (const ext of exts) {
-					if (fs.existsSync(`${importPath}${ext}`)) return { path: `${importPath}${ext}` }
-				}
-				return undefined
-			})
-		}
-	},
-}
-
-const common = {
-	bundle: true,
-	minify: false,
-	sourcemap: false,
-	logLevel: "silent",
-	format: "cjs",
-	platform: "node",
-	metafile: true,
-	write: false,
-	absWorkingDir: root,
-	tsconfig: path.join(root, "tsconfig.json"),
-	packages: "external",
-	plugins: [aliasResolverPlugin],
-	define: { "process.env.IS_DEV": "false", "process.env.IS_TEST": "false" },
-	banner: { js: "const _importMetaUrl=require('url').pathToFileURL(__filename)" },
-}
-
-async function inputsFor(entry, external) {
-	const r = await esbuild.build({ ...common, entryPoints: [entry], external })
-	return new Set(Object.keys(r.metafile.inputs).filter((f) => f.startsWith("src/") && /\.tsx?$/.test(f)))
-}
-
-const ext = await inputsFor("src/extension.ts", ["vscode"])
-const standalone = await inputsFor("src/standalone/cline-core.ts", [
-	"vscode",
-	"@grpc/reflection",
-	"grpc-health-check",
-	"better-sqlite3",
-])
-const live = new Set([...ext, ...standalone])
-
-// Third consumer: the webview (webview-ui/) is a separate Vite/React build that
-// imports extension code ONLY from src/shared (via "@shared/*" alias or relative
-// "../src/shared/*" paths). Any src/shared file referenced from webview-ui/src is
-// therefore live even if the extension-host/standalone bundles don't reach it.
-// Conservatively mark every src/shared file mentioned by the webview as live.
-const webviewFiles = await glob("webview-ui/src/**/*.{ts,tsx}", { cwd: root })
-const sharedMentionedByWebview = new Set()
-for (const wf of webviewFiles) {
-	const text = fs.readFileSync(path.join(root, wf), "utf8")
-	// Match @shared/X or .../src/shared/X import specifiers and map to src/shared/X
-	const re = /(?:@shared\/|src\/shared\/)([A-Za-z0-9_./-]+)/g
-	let m
-	while ((m = re.exec(text))) {
-		const rel = m[1].replace(/\.(ts|tsx|js|jsx)$/, "")
-		for (const cand of [`src/shared/${rel}.ts`, `src/shared/${rel}.tsx`, `src/shared/${rel}/index.ts`]) {
-			if (fs.existsSync(path.join(root, cand))) sharedMentionedByWebview.add(cand)
-		}
-	}
-}
-for (const f of sharedMentionedByWebview) live.add(f)
-console.log(`src/shared files referenced by webview: ${sharedMentionedByWebview.size}`)
-
-// All non-test, non-.d.ts source files on disk.
-const allSrc = (await glob("src/**/*.{ts,tsx}", { cwd: root }))
-	.filter((f) => !/\.test\.tsx?$/.test(f))
-	.filter((f) => !f.endsWith(".d.ts"))
-	.filter((f) => !f.includes("/__tests__/"))
-	.filter((f) => !f.startsWith("src/test/"))
-	.filter((f) => !f.startsWith("src/generated/")) // generated host glue
-	.filter((f) => !f.startsWith("src/dev/")) // dev-only tooling
-
-const dead = allSrc.filter((f) => !live.has(f)).sort()
-
-console.log(`extension inputs: ${ext.size}`)
-console.log(`standalone inputs: ${standalone.size}`)
-console.log(`union live src files: ${live.size}`)
-console.log(`candidate dead files: ${dead.length}`)
-fs.writeFileSync("/tmp/dead-src.json", JSON.stringify(dead, null, "\t"))
-console.log("--- dead candidates written to /tmp/dead-src.json ---")

apps/vscode/scripts/test-hostbridge-server.ts

@@ -87,12 +87,6 @@ function createMockService<T extends grpc.UntypedServiceImplementation>(serviceN
 						})
 						return
 
-					case "openExternal":
-						simulateOAuthBrowserCallback(call.request?.value || "")
-							.then(() => callback(null, {}))
-							.catch((error) => callback(error))
-						return
-
 					case "getWebviewHtml":
 						callback(null, {
 							html: "<html><body>Fake Webview</body></html>",
@@ -149,41 +143,6 @@ function createMockService<T extends grpc.UntypedServiceImplementation>(serviceN
 	return new Proxy({} as T, handler)
 }
 
-async function simulateOAuthBrowserCallback(urlString: string): Promise<void> {
-	let url: URL
-	try {
-		url = new URL(urlString)
-	} catch {
-		return
-	}
-
-	if (!isLoopbackHost(url.hostname) || url.pathname !== "/api/v1/auth/authorize") {
-		return
-	}
-
-	const callbackUrl = url.searchParams.get("callback_url") ?? url.searchParams.get("redirect_uri")
-	if (!callbackUrl) {
-		return
-	}
-
-	const callback = new URL(callbackUrl)
-	if (!isLoopbackHost(callback.hostname) || callback.pathname !== "/auth") {
-		return
-	}
-
-	callback.searchParams.set("code", "test-personal-token")
-	callback.searchParams.set("provider", "cline")
-
-	const response = await fetch(callback.toString())
-	if (!response.ok) {
-		throw new Error(`Mock OAuth callback failed: ${response.status} ${response.statusText}`)
-	}
-}
-
-function isLoopbackHost(hostname: string): boolean {
-	return hostname === "localhost" || hostname === "127.0.0.1" || hostname === "::1"
-}
-
 if (require.main === module) {
 	startTestHostBridgeServer().catch((err) => {
 		console.error("Failed to start test host bridge server:", err)

apps/vscode/scripts/test-standalone-core-api-server.ts

@@ -115,8 +115,7 @@ async function main(): Promise<void> {
 
 	const covDir = path.join(projectRoot, `coverage/coverage-core-${PROTOBUS_PORT}`)
 
-	const workosFetchMockPath = path.join(projectRoot, "scripts", "testing-platform-workos-fetch-mock.cjs")
-	const baseArgs = ["--enable-source-maps", "--require", workosFetchMockPath, path.join(distDir, "cline-core.js")]
+	const baseArgs = ["--enable-source-maps", path.join(distDir, "cline-core.js")]
 
 	const spawnArgs = USE_C8 ? ["c8", "--report-dir", covDir, "node", ...baseArgs] : ["node", ...baseArgs]
 

apps/vscode/scripts/testing-platform-workos-fetch-mock.cjs

@@ -1,56 +0,0 @@
-// Preload used by the standalone testing platform.
-// It makes the SDK WorkOS device-auth flow deterministic and fully local while
-// leaving production auth code on the same device-auth path used by users.
-
-const originalFetch = globalThis.fetch?.bind(globalThis)
-
-const WORKOS_ORIGIN = "https://api.workos.com"
-const DEVICE_CODE = "test-device-code"
-const USER_CODE = "PTBC-TXTP"
-const ACCESS_TOKEN = "test-personal-token"
-const REFRESH_TOKEN = "test-personal-token_refresh"
-
-function jsonResponse(body, init = {}) {
-	return new Response(JSON.stringify(body), {
-		status: init.status ?? 200,
-		headers: { "Content-Type": "application/json", ...(init.headers ?? {}) },
-	})
-}
-
-function inputUrl(input) {
-	if (typeof input === "string") return input
-	if (input instanceof URL) return input.toString()
-	if (input && typeof input === "object" && "url" in input) return input.url
-	return String(input)
-}
-
-globalThis.fetch = async (input, init) => {
-	const urlString = inputUrl(input)
-	let url
-	try {
-		url = new URL(urlString)
-	} catch {
-		return originalFetch(input, init)
-	}
-
-	if (url.origin === WORKOS_ORIGIN && url.pathname === "/user_management/authorize/device") {
-		return jsonResponse({
-			device_code: DEVICE_CODE,
-			user_code: USER_CODE,
-			verification_uri: "https://login.workos.test/device",
-			verification_uri_complete: `https://login.workos.test/device?user_code=${USER_CODE}`,
-			expires_in: 300,
-			interval: 1,
-		})
-	}
-
-	if (url.origin === WORKOS_ORIGIN && url.pathname === "/user_management/authenticate") {
-		return jsonResponse({
-			access_token: ACCESS_TOKEN,
-			refresh_token: REFRESH_TOKEN,
-			token_type: "Bearer",
-		})
-	}
-
-	return originalFetch(input, init)
-}

apps/vscode/src/__tests__/config.test.ts

@@ -21,7 +21,9 @@ describe("ClineEndpoint configuration", () => {
 
 		// Stub os.homedir to return our temp directory
 		originalHomedir = os.homedir
-		sandbox.stub(os, "homedir").returns(tempDir)
+		sandbox
+			.stub(os, "homedir")
+			.returns(tempDir)
 
 		// Reset the singleton state using internal method
 		;(ClineEndpoint as any)._instance = null

apps/vscode/src/common.ts

@@ -4,6 +4,7 @@ import "./utils/path" // necessary to have access to String.prototype.toPosix
 import { HostProvider } from "@/hosts/host-provider"
 import { Logger } from "@/shared/services/Logger"
 import type { StorageContext } from "@/shared/storage/storage-context"
+import { FileContextTracker } from "./core/context/context-tracking/FileContextTracker"
 import { clearOnboardingModelsCache } from "./core/controller/models/getClineOnboardingModels"
 import { HookDiscoveryCache } from "./core/hooks/HookDiscoveryCache"
 import { HookProcessRegistry } from "./core/hooks/HookProcessRegistry"
@@ -73,6 +74,8 @@ export async function initialize(storageContext: StorageContext): Promise<Webvie
 	syncWorker().init({ ...blobStoreSettings, userDistinctId: getDistinctId() })
 	// Clean up old temp files in background (non-blocking) and start periodic cleanup every 24 hours
 	ClineTempManager.startPeriodicCleanup()
+	// Clean up orphaned file context warnings (startup cleanup)
+	FileContextTracker.cleanupOrphanedWarnings(stateManager)
 
 	telemetryService.captureExtensionActivated()
 
@@ -103,7 +106,7 @@ async function showVersionUpdateAnnouncement(stateManager: StateManager) {
 				})
 			}
 			// Always update the main version tracker for the next launch.
-			stateManager.setGlobalState("clineVersion", currentVersion)
+			await stateManager.setGlobalState("clineVersion", currentVersion)
 		}
 	} catch (error) {
 		const errorMessage = error instanceof Error ? error.message : String(error)

apps/vscode/src/core/api/index.ts

@@ -1,24 +1,52 @@
-import { type ApiHandler as SdkApiHandler, type ApiStreamChunk as SdkApiStreamChunk } from "@cline/llms"
-import { ApiConfiguration, ModelInfo } from "@shared/api"
+import { ApiConfiguration, ModelInfo, QwenApiRegions } from "@shared/api"
 import { Mode } from "@shared/storage/types"
 import { ClineStorageMessage } from "@/shared/messages/content"
+import { Logger } from "@/shared/services/Logger"
 import { ClineTool } from "@/shared/tools"
+import { AIhubmixHandler } from "./providers/aihubmix"
+import { AnthropicHandler } from "./providers/anthropic"
+import { AskSageHandler } from "./providers/asksage"
+import { BasetenHandler } from "./providers/baseten"
+import { AwsBedrockHandler } from "./providers/bedrock"
+import { CerebrasHandler } from "./providers/cerebras"
+import { ClaudeCodeHandler } from "./providers/claude-code"
+import { ClineHandler } from "./providers/cline"
+import { DeepSeekHandler } from "./providers/deepseek"
+import { DifyHandler } from "./providers/dify"
+import { DoubaoHandler } from "./providers/doubao"
+import { FireworksHandler } from "./providers/fireworks"
+import { GeminiHandler } from "./providers/gemini"
+import { GroqHandler } from "./providers/groq"
+import { HicapHandler } from "./providers/hicap"
+import { HuaweiCloudMaaSHandler } from "./providers/huawei-cloud-maas"
+import { HuggingFaceHandler } from "./providers/huggingface"
+import { LiteLlmHandler } from "./providers/litellm"
+import { LmStudioHandler } from "./providers/lmstudio"
+import { MinimaxHandler } from "./providers/minimax"
+import { MistralHandler } from "./providers/mistral"
+import { MoonshotHandler } from "./providers/moonshot"
+import { NebiusHandler } from "./providers/nebius"
+import { NousResearchHandler } from "./providers/nousresearch"
+import { OcaHandler } from "./providers/oca"
+import { OllamaHandler } from "./providers/ollama"
+import { OpenAiHandler } from "./providers/openai"
+import { OpenAiCodexHandler } from "./providers/openai-codex"
+import { OpenAiNativeHandler } from "./providers/openai-native"
+import { OpenRouterHandler } from "./providers/openrouter"
+import { QwenHandler } from "./providers/qwen"
+import { QwenCodeHandler } from "./providers/qwen-code"
+import { RequestyHandler } from "./providers/requesty"
+import { SambanovaHandler } from "./providers/sambanova"
+import { SapAiCoreHandler } from "./providers/sapaicore"
+import { TogetherHandler } from "./providers/together"
+import { VercelAIGatewayHandler } from "./providers/vercel-ai-gateway"
+import { VertexHandler } from "./providers/vertex"
+import { VsCodeLmHandler } from "./providers/vscode-lm"
+import { WandbHandler } from "./providers/wandb"
+import { XAIHandler } from "./providers/xai"
+import { ZAiHandler } from "./providers/zai"
 import { ApiStream, ApiStreamUsageChunk } from "./transform/stream"
 
-// buildApiHandler now routes inference through the Cline SDK. It lives in
-// apps/vscode/src/sdk/sdk-api-handler.ts and callers import it directly from
-// there. It is deliberately NOT re-exported here: this barrel is imported
-// widely for *types* only, and re-exporting a value from the SDK module would
-// pull the entire SDK/session-factory runtime graph into every type importer
-// at module-eval time (which can break extension activation). Keep this file
-// types-only.
-
-// Re-export the SDK inference contracts so callers can depend on the SDK types
-// through the existing @core/api entry point. These are the canonical handler
-// and stream types going forward; the local interfaces below remain for the
-// classic provider classes until they are removed.
-export type { SdkApiHandler, SdkApiStreamChunk }
-
 export type CommonApiHandlerOptions = {
 	onRetryAttempt?: ApiConfiguration["onRetryAttempt"]
 }
@@ -32,7 +60,6 @@ export interface ApiHandler {
 export interface ApiHandlerModel {
 	id: string
 	info: ModelInfo
-	providerId?: string
 }
 
 export interface ApiProviderInfo {
@@ -45,3 +72,436 @@ export interface ApiProviderInfo {
 export interface SingleCompletionHandler {
 	completePrompt(prompt: string): Promise<string>
 }
+
+function createHandlerForProvider(
+	apiProvider: string | undefined,
+	options: Omit<ApiConfiguration, "apiProvider">,
+	mode: Mode,
+): ApiHandler {
+	switch (apiProvider) {
+		case "anthropic":
+			return new AnthropicHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				apiKey: options.apiKey,
+				anthropicBaseUrl: options.anthropicBaseUrl,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+			})
+		case "openrouter":
+			return new OpenRouterHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				openRouterApiKey: options.openRouterApiKey,
+				openRouterModelId: mode === "plan" ? options.planModeOpenRouterModelId : options.actModeOpenRouterModelId,
+				openRouterModelInfo: mode === "plan" ? options.planModeOpenRouterModelInfo : options.actModeOpenRouterModelInfo,
+				openRouterProviderSorting: options.openRouterProviderSorting,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+				enableParallelToolCalling: options.enableParallelToolCalling,
+			})
+		case "bedrock":
+			return new AwsBedrockHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+				awsAccessKey: options.awsAccessKey,
+				awsSecretKey: options.awsSecretKey,
+				awsSessionToken: options.awsSessionToken,
+				awsRegion: options.awsRegion,
+				awsAuthentication: options.awsAuthentication,
+				awsBedrockApiKey: options.awsBedrockApiKey,
+				awsUseCrossRegionInference: options.awsUseCrossRegionInference,
+				awsUseGlobalInference: options.awsUseGlobalInference,
+				awsBedrockUsePromptCache: options.awsBedrockUsePromptCache,
+				awsUseProfile: options.awsUseProfile,
+				awsProfile: options.awsProfile,
+				awsBedrockEndpoint: options.awsBedrockEndpoint,
+				awsBedrockCustomSelected:
+					mode === "plan" ? options.planModeAwsBedrockCustomSelected : options.actModeAwsBedrockCustomSelected,
+				awsBedrockCustomModelBaseId:
+					mode === "plan" ? options.planModeAwsBedrockCustomModelBaseId : options.actModeAwsBedrockCustomModelBaseId,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+			})
+		case "vertex":
+			return new VertexHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				vertexProjectId: options.vertexProjectId,
+				vertexRegion: options.vertexRegion,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+				geminiApiKey: options.geminiApiKey,
+				geminiBaseUrl: options.geminiBaseUrl,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+				ulid: options.ulid,
+			})
+		case "openai":
+			return new OpenAiHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				openAiApiKey: options.openAiApiKey,
+				openAiBaseUrl: options.openAiBaseUrl,
+				azureApiVersion: options.azureApiVersion,
+				azureIdentity: options.azureIdentity,
+				openAiHeaders: options.openAiHeaders,
+				openAiModelId: mode === "plan" ? options.planModeOpenAiModelId : options.actModeOpenAiModelId,
+				openAiModelInfo: mode === "plan" ? options.planModeOpenAiModelInfo : options.actModeOpenAiModelInfo,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+			})
+		case "ollama":
+			return new OllamaHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				ollamaBaseUrl: options.ollamaBaseUrl,
+				ollamaApiKey: options.ollamaApiKey,
+				ollamaModelId: mode === "plan" ? options.planModeOllamaModelId : options.actModeOllamaModelId,
+				ollamaApiOptionsCtxNum: options.ollamaApiOptionsCtxNum,
+				requestTimeoutMs: options.requestTimeoutMs,
+			})
+		case "lmstudio":
+			return new LmStudioHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				lmStudioBaseUrl: options.lmStudioBaseUrl,
+				lmStudioModelId: mode === "plan" ? options.planModeLmStudioModelId : options.actModeLmStudioModelId,
+				lmStudioMaxTokens: options.lmStudioMaxTokens,
+			})
+		case "gemini":
+			return new GeminiHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				vertexProjectId: options.vertexProjectId,
+				vertexRegion: options.vertexRegion,
+				geminiApiKey: options.geminiApiKey,
+				geminiBaseUrl: options.geminiBaseUrl,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+				ulid: options.ulid,
+			})
+		case "openai-native":
+			return new OpenAiNativeHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				openAiNativeApiKey: options.openAiNativeApiKey,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+			})
+		case "openai-codex":
+			return new OpenAiCodexHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		case "deepseek":
+			return new DeepSeekHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				deepSeekApiKey: options.deepSeekApiKey,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		case "requesty":
+			return new RequestyHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				requestyBaseUrl: options.requestyBaseUrl,
+				requestyApiKey: options.requestyApiKey,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+				requestyModelId: mode === "plan" ? options.planModeRequestyModelId : options.actModeRequestyModelId,
+				requestyModelInfo: mode === "plan" ? options.planModeRequestyModelInfo : options.actModeRequestyModelInfo,
+			})
+		case "fireworks":
+			return new FireworksHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				fireworksApiKey: options.fireworksApiKey,
+				fireworksModelId: mode === "plan" ? options.planModeFireworksModelId : options.actModeFireworksModelId,
+			})
+		case "together":
+			return new TogetherHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				togetherApiKey: options.togetherApiKey,
+				togetherModelId: mode === "plan" ? options.planModeTogetherModelId : options.actModeTogetherModelId,
+			})
+		case "qwen":
+			return new QwenHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				qwenApiKey: options.qwenApiKey,
+				qwenApiLine:
+					options.qwenApiLine === QwenApiRegions.INTERNATIONAL ? QwenApiRegions.INTERNATIONAL : QwenApiRegions.CHINA,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+			})
+		case "qwen-code":
+			return new QwenCodeHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				qwenCodeOauthPath: options.qwenCodeOauthPath,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		case "doubao":
+			return new DoubaoHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				doubaoApiKey: options.doubaoApiKey,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		case "mistral":
+			return new MistralHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				mistralApiKey: options.mistralApiKey,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		case "vscode-lm":
+			return new VsCodeLmHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				vsCodeLmModelSelector:
+					mode === "plan" ? options.planModeVsCodeLmModelSelector : options.actModeVsCodeLmModelSelector,
+			})
+		case "cline": {
+			const clineModelId =
+				(mode === "plan" ? options.planModeClineModelId : options.actModeClineModelId) ||
+				(mode === "plan" ? options.planModeOpenRouterModelId : options.actModeOpenRouterModelId)
+			const clineModelInfo =
+				(mode === "plan" ? options.planModeClineModelInfo : options.actModeClineModelInfo) ||
+				(mode === "plan" ? options.planModeOpenRouterModelInfo : options.actModeOpenRouterModelInfo)
+			return new ClineHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				clineAccountId: options.clineAccountId,
+				clineApiKey: options.clineApiKey,
+				ulid: options.ulid,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+				openRouterProviderSorting: options.openRouterProviderSorting,
+				openRouterModelId: clineModelId,
+				openRouterModelInfo: clineModelInfo,
+				enableParallelToolCalling: options.enableParallelToolCalling,
+			})
+		}
+		case "litellm":
+			return new LiteLlmHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				liteLlmApiKey: options.liteLlmApiKey,
+				liteLlmBaseUrl: options.liteLlmBaseUrl,
+				liteLlmModelId: mode === "plan" ? options.planModeLiteLlmModelId : options.actModeLiteLlmModelId,
+				liteLlmModelInfo: mode === "plan" ? options.planModeLiteLlmModelInfo : options.actModeLiteLlmModelInfo,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+				liteLlmUsePromptCache: options.liteLlmUsePromptCache,
+				ulid: options.ulid,
+			})
+		case "moonshot":
+			return new MoonshotHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				moonshotApiKey: options.moonshotApiKey,
+				moonshotApiLine: options.moonshotApiLine,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		case "huggingface":
+			return new HuggingFaceHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				huggingFaceApiKey: options.huggingFaceApiKey,
+				huggingFaceModelId: mode === "plan" ? options.planModeHuggingFaceModelId : options.actModeHuggingFaceModelId,
+				huggingFaceModelInfo:
+					mode === "plan" ? options.planModeHuggingFaceModelInfo : options.actModeHuggingFaceModelInfo,
+			})
+		case "nebius":
+			return new NebiusHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				nebiusApiKey: options.nebiusApiKey,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		case "asksage":
+			return new AskSageHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				asksageApiKey: options.asksageApiKey,
+				asksageApiUrl: options.asksageApiUrl,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		case "xai":
+			return new XAIHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				xaiApiKey: options.xaiApiKey,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		case "sambanova":
+			return new SambanovaHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				sambanovaApiKey: options.sambanovaApiKey,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		case "cerebras":
+			return new CerebrasHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				cerebrasApiKey: options.cerebrasApiKey,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		case "groq":
+			return new GroqHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				groqApiKey: options.groqApiKey,
+				groqModelId: mode === "plan" ? options.planModeGroqModelId : options.actModeGroqModelId,
+				groqModelInfo: mode === "plan" ? options.planModeGroqModelInfo : options.actModeGroqModelInfo,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		case "baseten":
+			return new BasetenHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				basetenApiKey: options.basetenApiKey,
+				basetenModelId: mode === "plan" ? options.planModeBasetenModelId : options.actModeBasetenModelId,
+				basetenModelInfo: mode === "plan" ? options.planModeBasetenModelInfo : options.actModeBasetenModelInfo,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		case "sapaicore":
+			return new SapAiCoreHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				sapAiCoreClientId: options.sapAiCoreClientId,
+				sapAiCoreClientSecret: options.sapAiCoreClientSecret,
+				sapAiCoreTokenUrl: options.sapAiCoreTokenUrl,
+				sapAiResourceGroup: options.sapAiResourceGroup,
+				sapAiCoreBaseUrl: options.sapAiCoreBaseUrl,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+				deploymentId: mode === "plan" ? options.planModeSapAiCoreDeploymentId : options.actModeSapAiCoreDeploymentId,
+				sapAiCoreUseOrchestrationMode: options.sapAiCoreUseOrchestrationMode,
+			})
+		case "claude-code":
+			return new ClaudeCodeHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				claudeCodePath: options.claudeCodePath,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+			})
+		case "huawei-cloud-maas":
+			return new HuaweiCloudMaaSHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				huaweiCloudMaasApiKey: options.huaweiCloudMaasApiKey,
+				huaweiCloudMaasModelId:
+					mode === "plan" ? options.planModeHuaweiCloudMaasModelId : options.actModeHuaweiCloudMaasModelId,
+				huaweiCloudMaasModelInfo:
+					mode === "plan" ? options.planModeHuaweiCloudMaasModelInfo : options.actModeHuaweiCloudMaasModelInfo,
+			})
+		case "dify": // Add Dify.ai handler
+			return new DifyHandler({
+				difyApiKey: options.difyApiKey,
+				difyBaseUrl: options.difyBaseUrl,
+			})
+		case "vercel-ai-gateway":
+			return new VercelAIGatewayHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				vercelAiGatewayApiKey: options.vercelAiGatewayApiKey,
+				openRouterModelId:
+					mode === "plan" ? options.planModeVercelAiGatewayModelId : options.actModeVercelAiGatewayModelId,
+				openRouterModelInfo:
+					mode === "plan" ? options.planModeVercelAiGatewayModelInfo : options.actModeVercelAiGatewayModelInfo,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+			})
+		case "zai":
+			return new ZAiHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				zaiApiLine: options.zaiApiLine,
+				zaiApiKey: options.zaiApiKey,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		case "oca":
+			return new OcaHandler({
+				ocaMode: options.ocaMode || "internal",
+				ocaBaseUrl: options.ocaBaseUrl,
+				ocaModelId: mode === "plan" ? options.planModeOcaModelId : options.actModeOcaModelId,
+				ocaModelInfo: mode === "plan" ? options.planModeOcaModelInfo : options.actModeOcaModelInfo,
+				ocaReasoningEffort: mode === "plan" ? options.planModeOcaReasoningEffort : options.actModeOcaReasoningEffort,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+				ocaUsePromptCache:
+					mode === "plan"
+						? options.planModeOcaModelInfo?.supportsPromptCache
+						: options.actModeOcaModelInfo?.supportsPromptCache,
+				taskId: options.ulid,
+			})
+		case "aihubmix":
+			return new AIhubmixHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				apiKey: options.aihubmixApiKey,
+				baseURL: options.aihubmixBaseUrl,
+				appCode: options.aihubmixAppCode,
+				modelId: mode === "plan" ? (options as any).planModeAihubmixModelId : (options as any).actModeAihubmixModelId,
+				modelInfo:
+					mode === "plan" ? (options as any).planModeAihubmixModelInfo : (options as any).actModeAihubmixModelInfo,
+			})
+		case "minimax":
+			return new MinimaxHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				minimaxApiKey: options.minimaxApiKey,
+				minimaxApiLine: options.minimaxApiLine,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+			})
+		case "hicap":
+			return new HicapHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				hicapApiKey: options.hicapApiKey,
+				hicapModelId: mode === "plan" ? options.planModeHicapModelId : options.actModeHicapModelId,
+			})
+		case "nousResearch":
+			return new NousResearchHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				nousResearchApiKey: options.nousResearchApiKey,
+				apiModelId: mode === "plan" ? options.planModeNousResearchModelId : options.actModeNousResearchModelId,
+			})
+		case "wandb":
+			return new WandbHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				wandbApiKey: options.wandbApiKey,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+			})
+		default:
+			return new AnthropicHandler({
+				onRetryAttempt: options.onRetryAttempt,
+				apiKey: options.apiKey,
+				anthropicBaseUrl: options.anthropicBaseUrl,
+				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
+				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
+				thinkingBudgetTokens:
+					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
+			})
+	}
+}
+
+export function buildApiHandler(configuration: ApiConfiguration, mode: Mode): ApiHandler {
+	const { planModeApiProvider, actModeApiProvider, ...options } = configuration
+
+	const apiProvider = mode === "plan" ? planModeApiProvider : actModeApiProvider
+
+	// Validate thinking budget tokens against model's maxTokens to prevent API errors
+	// wrapped in a try-catch for safety, but this should never throw
+	try {
+		const thinkingBudgetTokens = mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens
+		if (thinkingBudgetTokens && thinkingBudgetTokens > 0) {
+			const handler = createHandlerForProvider(apiProvider, options, mode)
+
+			const modelInfo = handler.getModel().info
+			if (modelInfo?.maxTokens && modelInfo.maxTokens > 0 && thinkingBudgetTokens > modelInfo.maxTokens) {
+				const clippedValue = modelInfo.maxTokens - 1
+				if (mode === "plan") {
+					options.planModeThinkingBudgetTokens = clippedValue
+				} else {
+					options.actModeThinkingBudgetTokens = clippedValue
+				}
+			} else {
+				return handler // don't rebuild unless its necessary
+			}
+		}
+	} catch (error) {
+		Logger.error("buildApiHandler error:", error)
+	}
+
+	return createHandlerForProvider(apiProvider, options, mode)
+}

apps/vscode/src/core/api/providers/__tests__/anthropic.test.ts

@@ -0,0 +1,236 @@
+import { afterEach, describe, it } from "mocha"
+import sinon from "sinon"
+import "should"
+import { anthropicModels } from "@shared/api"
+import { ANTHROPIC_FAST_MODE_BETA, AnthropicHandler } from "../anthropic"
+
+describe("AnthropicHandler", () => {
+	afterEach(() => {
+		sinon.restore()
+	})
+
+	const createAsyncIterable = (data: readonly unknown[] = []) => ({
+		[Symbol.asyncIterator]: async function* () {
+			yield* data
+		},
+	})
+
+	describe("getModel", () => {
+		it("should return the fast mode model when configured", () => {
+			const handler = new AnthropicHandler({
+				apiKey: "test-api-key",
+				apiModelId: "claude-opus-4-6:fast",
+			})
+
+			const result = handler.getModel()
+
+			result.id.should.equal("claude-opus-4-6:fast")
+			result.info.should.deepEqual(anthropicModels["claude-opus-4-6:fast"])
+		})
+
+		it("should return the 1m fast mode model when configured", () => {
+			const handler = new AnthropicHandler({
+				apiKey: "test-api-key",
+				apiModelId: "claude-opus-4-6:1m:fast",
+			})
+
+			const result = handler.getModel()
+
+			result.id.should.equal("claude-opus-4-6:1m:fast")
+			result.info.should.deepEqual(anthropicModels["claude-opus-4-6:1m:fast"])
+		})
+
+		it("should return the 4.7 model when configured", () => {
+			const handler = new AnthropicHandler({
+				apiKey: "test-api-key",
+				apiModelId: "claude-opus-4-7",
+			})
+
+			const result = handler.getModel()
+
+			result.id.should.equal("claude-opus-4-7")
+			result.info.should.deepEqual(anthropicModels["claude-opus-4-7"])
+		})
+
+		it("should return the 4.7 1m model when configured", () => {
+			const handler = new AnthropicHandler({
+				apiKey: "test-api-key",
+				apiModelId: "claude-opus-4-7:1m",
+			})
+
+			const result = handler.getModel()
+
+			result.id.should.equal("claude-opus-4-7:1m")
+			result.info.should.deepEqual(anthropicModels["claude-opus-4-7:1m"])
+		})
+
+		it("should return the 4.8 model when configured", () => {
+			const handler = new AnthropicHandler({
+				apiKey: "test-api-key",
+				apiModelId: "claude-opus-4-8",
+			})
+
+			const result = handler.getModel()
+
+			result.id.should.equal("claude-opus-4-8")
+			result.info.should.deepEqual(anthropicModels["claude-opus-4-8"])
+		})
+
+		it("should return the 4.8 1m model when configured", () => {
+			const handler = new AnthropicHandler({
+				apiKey: "test-api-key",
+				apiModelId: "claude-opus-4-8:1m",
+			})
+
+			const result = handler.getModel()
+
+			result.id.should.equal("claude-opus-4-8:1m")
+			result.info.should.deepEqual(anthropicModels["claude-opus-4-8:1m"])
+		})
+	})
+
+	describe("createMessage", () => {
+		it("should route fast mode requests through the beta messages API", async () => {
+			const handler = new AnthropicHandler({
+				apiKey: "test-api-key",
+				apiModelId: "claude-opus-4-6:fast",
+			})
+
+			const standardCreate = sinon.stub().resolves(createAsyncIterable())
+			const betaCreate = sinon.stub().callsFake(function (this: { _client?: object }, _params: unknown) {
+				should.exist(this._client)
+				return Promise.resolve(createAsyncIterable())
+			})
+
+			sinon.stub(handler as unknown as { ensureClient: () => unknown }, "ensureClient").returns({
+				messages: {
+					create: standardCreate,
+				},
+				beta: {
+					messages: {
+						_client: {},
+						create: betaCreate,
+					},
+				},
+			})
+
+			for await (const _chunk of handler.createMessage("system prompt", [{ role: "user", content: "Hello" }])) {
+			}
+
+			sinon.assert.notCalled(standardCreate)
+			sinon.assert.calledOnce(betaCreate)
+			sinon.assert.calledWithMatch(betaCreate, {
+				model: "claude-opus-4-6",
+				betas: [ANTHROPIC_FAST_MODE_BETA],
+				speed: "fast",
+				stream: true,
+			})
+		})
+
+		it("should include the 1m beta when routing 1m fast mode requests through the beta messages API", async () => {
+			const handler = new AnthropicHandler({
+				apiKey: "test-api-key",
+				apiModelId: "claude-opus-4-6:1m:fast",
+			})
+
+			const standardCreate = sinon.stub().resolves(createAsyncIterable())
+			const betaCreate = sinon.stub().callsFake(function (this: { _client?: object }, _params: unknown) {
+				should.exist(this._client)
+				return Promise.resolve(createAsyncIterable())
+			})
+
+			sinon.stub(handler as unknown as { ensureClient: () => unknown }, "ensureClient").returns({
+				messages: {
+					create: standardCreate,
+				},
+				beta: {
+					messages: {
+						_client: {},
+						create: betaCreate,
+					},
+				},
+			})
+
+			for await (const _chunk of handler.createMessage("system prompt", [{ role: "user", content: "Hello" }])) {
+			}
+
+			sinon.assert.notCalled(standardCreate)
+			sinon.assert.calledOnce(betaCreate)
+			sinon.assert.calledWithMatch(betaCreate, {
+				model: "claude-opus-4-6",
+				betas: [ANTHROPIC_FAST_MODE_BETA, "context-1m-2025-08-07"],
+				speed: "fast",
+				stream: true,
+			})
+		})
+
+		it("should include the 1m beta header for Claude Opus 4.7 1m requests", async () => {
+			const handler = new AnthropicHandler({
+				apiKey: "test-api-key",
+				apiModelId: "claude-opus-4-7:1m",
+				reasoningEffort: "high",
+			})
+
+			const standardCreate = sinon.stub().resolves(createAsyncIterable())
+
+			sinon.stub(handler as unknown as { ensureClient: () => unknown }, "ensureClient").returns({
+				messages: {
+					create: standardCreate,
+				},
+				beta: {
+					messages: {
+						_client: {},
+						create: sinon.stub().resolves(createAsyncIterable()),
+					},
+				},
+			})
+
+			for await (const _chunk of handler.createMessage("system prompt", [{ role: "user", content: "Hello" }])) {
+			}
+
+			sinon.assert.calledOnce(standardCreate)
+			const requestBody = standardCreate.firstCall.args[0] as Record<string, any>
+			const requestOptions = standardCreate.firstCall.args[1] as Record<string, any>
+			requestBody.model.should.equal("claude-opus-4-7")
+			requestBody.thinking.should.deepEqual({ type: "adaptive" })
+			requestOptions.should.deepEqual({
+				headers: {
+					"anthropic-beta": "context-1m-2025-08-07",
+				},
+			})
+		})
+
+		it("should use adaptive thinking and output_config for Claude Opus adaptive models", async () => {
+			const handler = new AnthropicHandler({
+				apiKey: "test-api-key",
+				apiModelId: "claude-opus-4-7",
+				reasoningEffort: "xhigh",
+			})
+
+			const standardCreate = sinon.stub().resolves(createAsyncIterable())
+
+			sinon.stub(handler as unknown as { ensureClient: () => unknown }, "ensureClient").returns({
+				messages: {
+					create: standardCreate,
+				},
+				beta: {
+					messages: {
+						_client: {},
+						create: sinon.stub().resolves(createAsyncIterable()),
+					},
+				},
+			})
+
+			for await (const _chunk of handler.createMessage("system prompt", [{ role: "user", content: "Hello" }])) {
+			}
+
+			sinon.assert.calledOnce(standardCreate)
+			const requestBody = standardCreate.firstCall.args[0] as Record<string, any>
+			requestBody.should.have.property("thinking")
+			requestBody.thinking.should.deepEqual({ type: "adaptive" })
+			requestBody.should.have.property("output_config")
+			requestBody.output_config.should.deepEqual({ effort: "xhigh" })
+			should(requestBody.temperature).equal(undefined)
+		})
+	})
+})

apps/vscode/src/core/api/providers/__tests__/claude-code.test.ts

@@ -0,0 +1,468 @@
+import { afterEach, beforeEach, describe, it } from "mocha"
+import sinon from "sinon"
+import "should"
+import { ClaudeCodeHandler } from "@core/api/providers/claude-code"
+import { ClineStorageMessage } from "@/shared/messages/content"
+
+describe("ClaudeCodeHandler", () => {
+	let handler: ClaudeCodeHandler
+	let sandbox: sinon.SinonSandbox
+
+	beforeEach(() => {
+		sandbox = sinon.createSandbox()
+		handler = new ClaudeCodeHandler({
+			claudeCodePath: "/mock/path",
+			apiModelId: "claude-opus-4-1-20250805",
+		})
+	})
+
+	afterEach(() => {
+		sandbox.restore()
+	})
+
+	describe("token counting", () => {
+		it("should correctly handle token usage from assistant messages", async () => {
+			// The 'input_tokens' field represents the TOTAL number of input tokens used.
+			// See https://docs.anthropic.com/en/api/messages#usage-object
+
+			// Mock the runClaudeCode function
+			const runClaudeCodeModule = await import("@/integrations/claude-code/run")
+			const runClaudeCodeStub = sandbox.stub(runClaudeCodeModule, "runClaudeCode")
+
+			// Create a proper async generator mock for the Claude Code response
+			async function* mockGenerator() {
+				// First yield the system init
+				yield {
+					type: "system",
+					subtype: "init",
+					apiKeySource: "api",
+				}
+
+				// Yield assistant message with usage data
+				// Example: If base input is 70 tokens, cache read is 20, and cache creation is 10,
+				// then input_tokens from Anthropic API will be 100 (70 + 20 + 10)
+				yield {
+					type: "assistant",
+					message: {
+						content: [
+							{
+								type: "text",
+								text: "Test response",
+							},
+						],
+						usage: {
+							input_tokens: 100, // Total including cache (per Anthropic docs)
+							output_tokens: 50,
+							cache_read_input_tokens: 20, // Already included in input_tokens
+							cache_creation_input_tokens: 10, // Already included in input_tokens
+						},
+						stop_reason: "end_turn",
+					},
+				}
+
+				// Yield result with cost
+				yield {
+					type: "result",
+					result: {},
+					total_cost_usd: 0.005,
+				}
+			}
+
+			runClaudeCodeStub.returns(mockGenerator() as any)
+
+			const systemPrompt = "You are a helpful assistant."
+			const messages: ClineStorageMessage[] = [{ role: "user", content: "Hello" }]
+
+			const usageData: any[] = []
+
+			// Collect the results
+			for await (const chunk of handler.createMessage(systemPrompt, messages)) {
+				if (chunk.type === "usage") {
+					usageData.push({
+						inputTokens: chunk.inputTokens,
+						outputTokens: chunk.outputTokens,
+						cacheReadTokens: chunk.cacheReadTokens,
+						cacheWriteTokens: chunk.cacheWriteTokens,
+						totalCost: chunk.totalCost,
+					})
+				}
+			}
+
+			// Verify token counting follows Anthropic API specification
+			usageData.should.have.length(1)
+			usageData[0].should.deepEqual({
+				inputTokens: 100, // Total including cache tokens (per Anthropic API docs)
+				outputTokens: 50,
+				cacheReadTokens: 20, // Tracked separately for reporting
+				cacheWriteTokens: 10, // Tracked separately for reporting
+				totalCost: 0.005,
+			})
+
+			// CRITICAL ASSERTION: Verify that input_tokens is NOT inflated by re-adding cache tokens
+			// The bug would have caused inputTokens to be incorrectly calculated as 130 (100 + 20 + 10)
+			// The fix ensures it remains 100, as per Anthropic's specification
+			usageData[0].inputTokens.should.equal(100) // Correct: matches API response
+			usageData[0].inputTokens.should.not.equal(130) // Would be wrong: double-counting cache tokens
+		})
+
+		it("should handle missing usage fields with nullish coalescing", async () => {
+			// Mock the runClaudeCode function
+			const runClaudeCodeModule = await import("@/integrations/claude-code/run")
+			const runClaudeCodeStub = sandbox.stub(runClaudeCodeModule, "runClaudeCode")
+
+			// Create a proper async generator mock with missing/undefined usage fields
+			async function* mockGenerator() {
+				yield {
+					type: "assistant",
+					message: {
+						content: [
+							{
+								type: "text",
+								text: "Test response",
+							},
+						],
+						usage: {
+							input_tokens: 100,
+							output_tokens: 50,
+							// cache fields are undefined/missing
+						},
+						stop_reason: "end_turn",
+					},
+				}
+
+				yield {
+					type: "result",
+					result: {},
+					total_cost_usd: 0.005,
+				}
+			}
+
+			runClaudeCodeStub.returns(mockGenerator() as any)
+
+			const systemPrompt = "You are a helpful assistant."
+			const messages: ClineStorageMessage[] = [{ role: "user", content: "Hello" }]
+
+			const usageData: any[] = []
+
+			// Collect the results
+			for await (const chunk of handler.createMessage(systemPrompt, messages)) {
+				if (chunk.type === "usage") {
+					usageData.push({
+						inputTokens: chunk.inputTokens,
+						outputTokens: chunk.outputTokens,
+						cacheReadTokens: chunk.cacheReadTokens,
+						cacheWriteTokens: chunk.cacheWriteTokens,
+					})
+				}
+			}
+
+			// Verify that undefined cache tokens default to 0
+			usageData.should.have.length(1)
+			usageData[0].should.deepEqual({
+				inputTokens: 100,
+				outputTokens: 50,
+				cacheReadTokens: 0, // Should default to 0
+				cacheWriteTokens: 0, // Should default to 0
+			})
+		})
+
+		it("should handle completely missing usage object", async () => {
+			// Mock the runClaudeCode function
+			const runClaudeCodeModule = await import("@/integrations/claude-code/run")
+			const runClaudeCodeStub = sandbox.stub(runClaudeCodeModule, "runClaudeCode")
+
+			// Create a proper async generator mock with missing usage object
+			async function* mockGenerator() {
+				yield {
+					type: "assistant",
+					message: {
+						content: [
+							{
+								type: "text",
+								text: "Test response",
+							},
+						],
+						// usage is undefined
+						usage: undefined,
+						stop_reason: "end_turn",
+					},
+				}
+
+				// Need to yield a result chunk to trigger usage data emission
+				yield {
+					type: "result",
+					result: {},
+					total_cost_usd: 0,
+				}
+			}
+
+			runClaudeCodeStub.returns(mockGenerator() as any)
+
+			const systemPrompt = "You are a helpful assistant."
+			const messages: ClineStorageMessage[] = [{ role: "user", content: "Hello" }]
+
+			const usageData: any[] = []
+
+			// Collect the results
+			for await (const chunk of handler.createMessage(systemPrompt, messages)) {
+				if (chunk.type === "usage") {
+					usageData.push({
+						inputTokens: chunk.inputTokens,
+						outputTokens: chunk.outputTokens,
+						cacheReadTokens: chunk.cacheReadTokens,
+						cacheWriteTokens: chunk.cacheWriteTokens,
+					})
+				}
+			}
+
+			// All token counts should default to 0 when usage is undefined
+			usageData.should.have.length(1)
+			usageData[0].should.deepEqual({
+				inputTokens: 0,
+				outputTokens: 0,
+				cacheReadTokens: 0,
+				cacheWriteTokens: 0,
+			})
+		})
+	})
+
+	describe("error handling", () => {
+		it("should not crash when assistant message has empty content array", async () => {
+			const runClaudeCodeModule = await import("@/integrations/claude-code/run")
+			const runClaudeCodeStub = sandbox.stub(runClaudeCodeModule, "runClaudeCode")
+
+			async function* mockGenerator() {
+				yield {
+					type: "assistant",
+					message: {
+						content: [], // empty content — triggered TypeError in older code
+						usage: {
+							input_tokens: 10,
+							output_tokens: 0,
+						},
+						stop_reason: "end_turn",
+					},
+				}
+
+				yield {
+					type: "result",
+					result: {},
+					total_cost_usd: 0,
+				}
+			}
+
+			runClaudeCodeStub.returns(mockGenerator() as any)
+
+			const chunks: any[] = []
+			// Should not throw
+			for await (const chunk of handler.createMessage("system", [{ role: "user", content: "hi" }])) {
+				chunks.push(chunk)
+			}
+
+			const usageChunk = chunks.find((c) => c.type === "usage")
+			usageChunk.should.be.ok()
+			usageChunk.inputTokens.should.equal(10)
+		})
+
+		it("should throw when result has is_error=true (e.g. rate limit with no assistant message)", async () => {
+			const runClaudeCodeModule = await import("@/integrations/claude-code/run")
+			const runClaudeCodeStub = sandbox.stub(runClaudeCodeModule, "runClaudeCode")
+
+			async function* mockGenerator() {
+				yield {
+					type: "system",
+					subtype: "init",
+					apiKeySource: "none",
+				}
+
+				yield {
+					type: "system",
+					subtype: "rate_limit_event",
+					message: "Rate limit hit",
+					retryAfterSeconds: 30,
+				}
+
+				// No assistant message — CLI hit rate limit and gave up
+				yield {
+					type: "result",
+					subtype: "error",
+					is_error: true,
+					result: "Rate limit exceeded",
+					total_cost_usd: 0,
+					duration_ms: 1000,
+					duration_api_ms: 500,
+					num_turns: 0,
+					session_id: "test",
+				}
+			}
+
+			runClaudeCodeStub.returns(mockGenerator() as any)
+
+			let thrownError: Error | undefined
+			try {
+				for await (const _ of handler.createMessage("system", [{ role: "user", content: "hi" }])) {
+					// consume
+				}
+			} catch (err) {
+				thrownError = err as Error
+			}
+
+			thrownError!.message.should.containEql("Rate limit exceeded")
+		})
+
+		it("should ignore rate_limit_event system messages without throwing", async () => {
+			const runClaudeCodeModule = await import("@/integrations/claude-code/run")
+			const runClaudeCodeStub = sandbox.stub(runClaudeCodeModule, "runClaudeCode")
+
+			async function* mockGenerator() {
+				yield {
+					type: "system",
+					subtype: "init",
+					apiKeySource: "none",
+				}
+
+				// Newer Claude Code CLI emits this during rate limiting
+				yield {
+					type: "system",
+					subtype: "rate_limit_event",
+					message: "Rate limit hit, retrying...",
+					retryAfterSeconds: 30,
+				}
+
+				yield {
+					type: "assistant",
+					message: {
+						content: [{ type: "text", text: "Response after retry" }],
+						usage: { input_tokens: 20, output_tokens: 10 },
+						stop_reason: "end_turn",
+					},
+				}
+
+				yield {
+					type: "result",
+					result: {},
+					total_cost_usd: 0,
+				}
+			}
+
+			runClaudeCodeStub.returns(mockGenerator() as any)
+
+			const textChunks: string[] = []
+			for await (const chunk of handler.createMessage("system", [{ role: "user", content: "hi" }])) {
+				if (chunk.type === "text") textChunks.push(chunk.text)
+			}
+
+			textChunks.should.deepEqual(["Response after retry"])
+		})
+	})
+
+	describe("getModel", () => {
+		it("should return the correct model when specified", () => {
+			const handler = new ClaudeCodeHandler({
+				apiModelId: "claude-sonnet-4-5-20250929",
+			})
+
+			const model = handler.getModel()
+			model.id.should.equal("claude-sonnet-4-5-20250929")
+		})
+
+		it("should support Opus 4.6 1m model id", () => {
+			const handler = new ClaudeCodeHandler({
+				apiModelId: "claude-opus-4-6[1m]",
+			})
+
+			const model = handler.getModel()
+			model.id.should.equal("claude-opus-4-6[1m]")
+			model.info.contextWindow.should.equal(1_000_000)
+		})
+
+		it("should support Opus 4.7 model id", () => {
+			const handler = new ClaudeCodeHandler({
+				apiModelId: "claude-opus-4-7",
+			})
+
+			const model = handler.getModel()
+			model.id.should.equal("claude-opus-4-7")
+			model.info.contextWindow.should.equal(200_000)
+		})
+
+		it("should support Opus 4.7 1m model id", () => {
+			const handler = new ClaudeCodeHandler({
+				apiModelId: "claude-opus-4-7[1m]",
+			})
+
+			const model = handler.getModel()
+			model.id.should.equal("claude-opus-4-7[1m]")
+			model.info.contextWindow.should.equal(1_000_000)
+		})
+
+		it("should support Opus 4.8 model id", () => {
+			const handler = new ClaudeCodeHandler({
+				apiModelId: "claude-opus-4-8",
+			})
+
+			const model = handler.getModel()
+			model.id.should.equal("claude-opus-4-8")
+			model.info.contextWindow.should.equal(200_000)
+		})
+
+		it("should support Opus 4.8 1m model id", () => {
+			const handler = new ClaudeCodeHandler({
+				apiModelId: "claude-opus-4-8[1m]",
+			})
+
+			const model = handler.getModel()
+			model.id.should.equal("claude-opus-4-8[1m]")
+			model.info.contextWindow.should.equal(1_000_000)
+		})
+
+		it("should support Opus 1m alias model id", () => {
+			const handler = new ClaudeCodeHandler({
+				apiModelId: "opus[1m]",
+			})
+
+			const model = handler.getModel()
+			model.id.should.equal("opus[1m]")
+			model.info.contextWindow.should.equal(1_000_000)
+		})
+
+		it("should support Sonnet 1m alias model id", () => {
+			const handler = new ClaudeCodeHandler({
+				apiModelId: "sonnet[1m]",
+			})
+
+			const model = handler.getModel()
+			model.id.should.equal("sonnet[1m]")
+			model.info.contextWindow.should.equal(1_000_000)
+		})
+
+		it("should support Sonnet 4.5 1m model id", () => {
+			const handler = new ClaudeCodeHandler({
+				apiModelId: "claude-sonnet-4-5-20250929[1m]",
+			})
+
+			const model = handler.getModel()
+			model.id.should.equal("claude-sonnet-4-5-20250929[1m]")
+			model.info.contextWindow.should.equal(1_000_000)
+		})
+
+		it("should support Sonnet 4.6 1m model id", () => {
+			const handler = new ClaudeCodeHandler({
+				apiModelId: "claude-sonnet-4-6[1m]",
+			})
+
+			const model = handler.getModel()
+			model.id.should.equal("claude-sonnet-4-6[1m]")
+			model.info.contextWindow.should.equal(1_000_000)
+		})
+
+		it("should return default model when not specified", () => {
+			const handler = new ClaudeCodeHandler({})
+
+			const model = handler.getModel()
+			// The default model should be set
+			model.id.should.be.type("string")
+			model.info.should.be.type("object")
+		})
+	})
+})

apps/vscode/src/core/api/providers/__tests__/cline.test.ts

@@ -0,0 +1,166 @@
+import "should"
+import { openRouterDefaultModelInfo } from "@shared/api"
+import sinon from "sinon"
+import { ClineAccountService } from "@/services/account/ClineAccountService"
+import { AuthService } from "@/services/auth/AuthService"
+import { ClineHandler } from "../cline"
+
+describe("ClineHandler", () => {
+	afterEach(() => {
+		sinon.restore()
+	})
+
+	const createAsyncIterable = (data: any[] = []) => ({
+		[Symbol.asyncIterator]: async function* () {
+			yield* data
+		},
+	})
+
+	const createHandler = (options: ConstructorParameters<typeof ClineHandler>[0]) => {
+		sinon.stub(ClineAccountService, "getInstance").returns({} as any)
+		sinon.stub(AuthService, "getInstance").returns({} as any)
+		return new ClineHandler(options)
+	}
+
+	it("should handle usage-only chunks when delta is missing", async () => {
+		const handler = createHandler({})
+		const fakeClient = {
+			chat: {
+				completions: {
+					create: sinon.stub().resolves(
+						createAsyncIterable([
+							{
+								choices: [{}],
+								usage: {
+									prompt_tokens: 17,
+									completion_tokens: 9,
+								},
+							},
+						]),
+					),
+				},
+			},
+		}
+		sinon.stub(handler as any, "ensureClient").resolves(fakeClient as any)
+		sinon.stub(handler, "getModel").returns({
+			id: "openai/gpt-4o-mini",
+			info: openRouterDefaultModelInfo,
+		})
+
+		const chunks: any[] = []
+		for await (const chunk of handler.createMessage("system", [{ role: "user", content: "hi" }])) {
+			chunks.push(chunk)
+		}
+
+		chunks.should.deepEqual([
+			{
+				type: "usage",
+				cacheWriteTokens: 0,
+				cacheReadTokens: 0,
+				inputTokens: 17,
+				outputTokens: 9,
+				totalCost: 0,
+			},
+		])
+	})
+
+	it("should read Anthropic-style cache creation and read tokens from usage chunks", async () => {
+		const handler = createHandler({})
+		const fakeClient = {
+			chat: {
+				completions: {
+					create: sinon.stub().resolves(
+						createAsyncIterable([
+							{
+								choices: [{}],
+								usage: {
+									prompt_tokens: 1000,
+									completion_tokens: 200,
+									prompt_tokens_details: {
+										cached_tokens: 500,
+									},
+									cache_creation_input_tokens: 300,
+								},
+							},
+						]),
+					),
+				},
+			},
+		}
+		sinon.stub(handler as any, "ensureClient").resolves(fakeClient as any)
+		sinon.stub(handler, "getModel").returns({
+			id: "anthropic/claude-sonnet-4.6",
+			info: openRouterDefaultModelInfo,
+		})
+
+		const chunks: any[] = []
+		for await (const chunk of handler.createMessage("system", [{ role: "user", content: "hi" }])) {
+			chunks.push(chunk)
+		}
+
+		chunks.should.deepEqual([
+			{
+				type: "usage",
+				cacheWriteTokens: 300,
+				cacheReadTokens: 500,
+				inputTokens: 200,
+				outputTokens: 200,
+				totalCost: 0,
+			},
+		])
+	})
+
+	it("should forward enableParallelToolCalling to OpenRouter payload", async () => {
+		const handler = createHandler({ enableParallelToolCalling: true })
+		const createStub = sinon.stub().resolves(createAsyncIterable([]))
+		const fakeClient = {
+			chat: {
+				completions: {
+					create: createStub,
+				},
+			},
+		}
+		sinon.stub(handler as any, "ensureClient").resolves(fakeClient as any)
+		sinon.stub(handler, "getModel").returns({
+			id: "openai/gpt-4o-mini",
+			info: openRouterDefaultModelInfo,
+		})
+
+		const tools = [
+			{ type: "function", function: { name: "read_file", description: "", parameters: { type: "object" } } },
+		] as any
+		for await (const _chunk of handler.createMessage("system", [{ role: "user", content: "hi" }], tools)) {
+			// drain stream
+		}
+
+		const payload = createStub.firstCall.args[0]
+		payload.parallel_tool_calls.should.equal(true)
+	})
+
+	it("should send cache_control for qwen3.7-max without changing the selected Cline model id", async () => {
+		const handler = createHandler({
+			openRouterModelId: "qwen/qwen3.7-max",
+			openRouterModelInfo: openRouterDefaultModelInfo,
+		})
+		const createStub = sinon.stub().resolves(createAsyncIterable([]))
+		const fakeClient = {
+			chat: {
+				completions: {
+					create: createStub,
+				},
+			},
+		}
+		sinon.stub(handler as any, "ensureClient").resolves(fakeClient as any)
+		sinon.stub(handler as any, "getFreeModelIdSet").resolves(new Set())
+
+		for await (const _chunk of handler.createMessage("system", [{ role: "user", content: "hi" }])) {
+			// drain stream
+		}
+
+		handler.getModel().id.should.equal("qwen/qwen3.7-max")
+		const payload = createStub.firstCall.args[0]
+		payload.model.should.equal("qwen/qwen3.7-max")
+		payload.messages[0].content[0].cache_control.should.deepEqual({ type: "ephemeral" })
+		payload.messages[1].content[0].cache_control.should.deepEqual({ type: "ephemeral" })
+	})
+})

apps/vscode/src/core/api/providers/__tests__/fireworks.test.ts

@@ -0,0 +1,97 @@
+import "should"
+import sinon from "sinon"
+import { FireworksHandler } from "../fireworks"
+
+describe("FireworksHandler", () => {
+	afterEach(() => {
+		sinon.restore()
+	})
+
+	const createAsyncIterable = (data: any[] = []) => ({
+		[Symbol.asyncIterator]: async function* () {
+			yield* data
+		},
+	})
+
+	it("should handle usage-only chunks when delta is missing", async () => {
+		const handler = new FireworksHandler({
+			fireworksApiKey: "test-api-key",
+			fireworksModelId: "accounts/fireworks/models/llama-v3p1-8b-instruct",
+		})
+		const fakeClient = {
+			chat: {
+				completions: {
+					create: sinon.stub().resolves(
+						createAsyncIterable([
+							{
+								choices: [{}],
+								usage: {
+									prompt_tokens: 19,
+									completion_tokens: 4,
+								},
+							},
+						]),
+					),
+				},
+			},
+		}
+		sinon.stub(handler as any, "ensureClient").returns(fakeClient as any)
+
+		const chunks: any[] = []
+		for await (const chunk of handler.createMessage("system", [{ role: "user", content: "hi" }])) {
+			chunks.push(chunk)
+		}
+
+		chunks.should.deepEqual([
+			{
+				type: "usage",
+				inputTokens: 19,
+				outputTokens: 4,
+				cacheReadTokens: 0,
+				cacheWriteTokens: 0,
+			},
+		])
+	})
+
+	it("should read cache hits from prompt_tokens_details when hit tokens are not present", async () => {
+		const handler = new FireworksHandler({
+			fireworksApiKey: "test-api-key",
+			fireworksModelId: "accounts/fireworks/models/llama-v3p1-8b-instruct",
+		})
+		const fakeClient = {
+			chat: {
+				completions: {
+					create: sinon.stub().resolves(
+						createAsyncIterable([
+							{
+								choices: [{}],
+								usage: {
+									prompt_tokens: 60,
+									completion_tokens: 12,
+									prompt_tokens_details: { cached_tokens: 20 },
+									prompt_cache_miss_tokens: 40,
+								},
+							},
+						]),
+					),
+				},
+			},
+		}
+		sinon.stub(handler as any, "ensureClient").returns(fakeClient as any)
+
+		const chunks: any[] = []
+		for await (const chunk of handler.createMessage("system", [{ role: "user", content: "hi" }])) {
+			chunks.push(chunk)
+		}
+
+		chunks.should.deepEqual([
+			{
+				type: "usage",
+				inputTokens: 60,
+				outputTokens: 12,
+				cacheReadTokens: 20,
+				cacheWriteTokens: 40,
+			},
+		])
+	})
+})

apps/vscode/src/core/api/providers/__tests__/gemini.test.ts

@@ -0,0 +1,235 @@
+import "should"
+import sinon from "sinon"
+import { GeminiHandler } from "../gemini"
+
+describe("GeminiHandler", () => {
+	afterEach(() => {
+		sinon.restore()
+	})
+
+	const createAsyncIterable = (data: any[] = []) => ({
+		[Symbol.asyncIterator]: async function* () {
+			yield* data
+		},
+	})
+
+	it("caps maxOutputTokens to 8192 for Flash models", async () => {
+		const handler = new GeminiHandler({
+			geminiApiKey: "test-api-key",
+			apiModelId: "gemini-2.5-flash",
+		})
+
+		const generateContentStream = sinon.stub().resolves(
+			createAsyncIterable([
+				{
+					responseId: "resp-1",
+					usageMetadata: {
+						promptTokenCount: 10,
+						candidatesTokenCount: 20,
+						cachedContentTokenCount: 0,
+						thoughtsTokenCount: 0,
+					},
+				},
+			]),
+		)
+		sinon.stub(handler as any, "ensureClient").returns({
+			models: { generateContentStream },
+		} as any)
+
+		for await (const _chunk of handler.createMessage("system", [{ role: "user", content: "hi" }] as any)) {
+			// Consume stream to trigger request execution.
+		}
+
+		const requestArgs = generateContentStream.firstCall.args[0] as Record<string, any>
+		requestArgs.config.should.have.property("maxOutputTokens", 8_192)
+	})
+
+	it("supports Gemini 3.5 Flash model metadata", async () => {
+		const handler = new GeminiHandler({
+			geminiApiKey: "test-api-key",
+			apiModelId: "gemini-3.5-flash",
+		})
+
+		const model = handler.getModel()
+		model.id.should.equal("gemini-3.5-flash")
+		model.info.contextWindow!.should.equal(1_048_576)
+		model.info.inputPrice!.should.equal(1.5)
+		model.info.outputPrice!.should.equal(9)
+		model.info.cacheReadsPrice!.should.equal(0.15)
+		model.info.supportsReasoning!.should.equal(true)
+
+		const generateContentStream = sinon.stub().resolves(
+			createAsyncIterable([
+				{
+					responseId: "resp-35",
+					usageMetadata: {
+						promptTokenCount: 10,
+						candidatesTokenCount: 20,
+						cachedContentTokenCount: 0,
+						thoughtsTokenCount: 0,
+					},
+				},
+			]),
+		)
+		sinon.stub(handler as any, "ensureClient").returns({
+			models: { generateContentStream },
+		} as any)
+
+		for await (const _chunk of handler.createMessage("system", [{ role: "user", content: "hi" }] as any)) {
+			// Consume stream to trigger request execution.
+		}
+
+		const requestArgs = generateContentStream.firstCall.args[0] as Record<string, any>
+		requestArgs.model.should.equal("gemini-3.5-flash")
+		requestArgs.config.should.have.property("maxOutputTokens", 8_192)
+		requestArgs.config.thinkingConfig.should.deepEqual({
+			thinkingBudget: undefined,
+			thinkingLevel: "LOW",
+			includeThoughts: true,
+		})
+	})
+
+	it("does not set maxOutputTokens for non-Flash models", async () => {
+		const handler = new GeminiHandler({
+			geminiApiKey: "test-api-key",
+			apiModelId: "gemini-2.5-pro",
+		})
+
+		const generateContentStream = sinon.stub().resolves(
+			createAsyncIterable([
+				{
+					responseId: "resp-2",
+					usageMetadata: {
+						promptTokenCount: 10,
+						candidatesTokenCount: 20,
+						cachedContentTokenCount: 0,
+						thoughtsTokenCount: 0,
+					},
+				},
+			]),
+		)
+		sinon.stub(handler as any, "ensureClient").returns({
+			models: { generateContentStream },
+		} as any)
+
+		for await (const _chunk of handler.createMessage("system", [{ role: "user", content: "hi" }] as any)) {
+			// Consume stream to trigger request execution.
+		}
+
+		const requestArgs = generateContentStream.firstCall.args[0] as Record<string, any>
+		requestArgs.config.should.not.have.property("maxOutputTokens")
+	})
+
+	it("should emit unique tool call IDs when multiple function calls share one responseId", async () => {
+		const handler = new GeminiHandler({
+			geminiApiKey: "test-api-key",
+		})
+
+		const fakeClient = {
+			models: {
+				generateContentStream: sinon.stub().resolves(
+					createAsyncIterable([
+						{
+							responseId: "resp_1",
+							candidates: [
+								{
+									content: {
+										parts: [
+											{
+												functionCall: {
+													name: "read_file",
+													args: { path: ".nvmrc" },
+												},
+											},
+										],
+									},
+								},
+							],
+						},
+						{
+							responseId: "resp_1",
+							candidates: [
+								{
+									content: {
+										parts: [
+											{
+												functionCall: {
+													name: "read_file",
+													args: { path: ".gitattributes" },
+												},
+											},
+										],
+									},
+								},
+							],
+						},
+					]),
+				),
+			},
+		}
+		sinon.stub(handler as any, "ensureClient").returns(fakeClient as any)
+
+		const tools = [{ name: "read_file", description: "read file", parameters: { type: "OBJECT" } }] as any
+		const chunks: any[] = []
+		for await (const chunk of handler.createMessage("system", [{ role: "user", content: "hi" }], tools)) {
+			if (chunk.type === "tool_calls") {
+				chunks.push(chunk)
+			}
+		}
+
+		chunks.should.have.length(2)
+		chunks[0].tool_call.function.id.should.equal("resp_1-tool-0")
+		chunks[1].tool_call.function.id.should.equal("resp_1-tool-1")
+		chunks[0].tool_call.call_id.should.equal(chunks[0].tool_call.function.id)
+		chunks[1].tool_call.call_id.should.equal(chunks[1].tool_call.function.id)
+		JSON.parse(chunks[0].tool_call.function.arguments).path.should.equal(".nvmrc")
+		JSON.parse(chunks[1].tool_call.function.arguments).path.should.equal(".gitattributes")
+	})
+
+	it("should preserve Gemini-provided functionCall.id when present", async () => {
+		const handler = new GeminiHandler({
+			geminiApiKey: "test-api-key",
+		})
+
+		const fakeClient = {
+			models: {
+				generateContentStream: sinon.stub().resolves(
+					createAsyncIterable([
+						{
+							responseId: "resp_2",
+							candidates: [
+								{
+									content: {
+										parts: [
+											{
+												functionCall: {
+													id: "call_alpha",
+													name: "read_file",
+													args: { path: ".nvmrc" },
+												},
+											},
+										],
+									},
+								},
+							],
+						},
+					]),
+				),
+			},
+		}
+		sinon.stub(handler as any, "ensureClient").returns(fakeClient as any)
+
+		const tools = [{ name: "read_file", description: "read file", parameters: { type: "OBJECT" } }] as any
+		const chunks: any[] = []
+		for await (const chunk of handler.createMessage("system", [{ role: "user", content: "hi" }], tools)) {
+			if (chunk.type === "tool_calls") {
+				chunks.push(chunk)
+			}
+		}
+
+		chunks.should.have.length(1)
+		chunks[0].tool_call.function.id.should.equal("call_alpha")
+		chunks[0].tool_call.call_id.should.equal("call_alpha")
+		JSON.parse(chunks[0].tool_call.function.arguments).path.should.equal(".nvmrc")
+	})
+})

apps/vscode/src/core/api/providers/__tests__/litellm.test.ts

@@ -0,0 +1,326 @@
+import { LiteLlmHandler, type LiteLlmModelInfoResponse } from "@core/api/providers/litellm"
+import { convertToOpenAiMessages } from "@core/api/transform/openai-format"
+import { liteLlmModelInfoSaneDefaults } from "@shared/api" // used in getModel tests
+import { expect } from "chai"
+import sinon from "sinon"
+import { StateManager } from "@/core/storage/StateManager" // used in getModel tests
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { mockFetchForTesting } from "@/shared/net"
+
+const fakeClient = {
+	chat: {
+		completions: {
+			create: sinon.stub(),
+		},
+	},
+	baseURL: "https://fake.example",
+}
+
+describe("LiteLlmHandler", () => {
+	const mockFetch = sinon.stub()
+	let doneMockingFetch: (value: any) => void = () => {}
+
+	const mockModelFetch = (modelInfo: LiteLlmModelInfoResponse["data"][number]) => {
+		mockFetch.resolves({
+			ok: true,
+			json: () =>
+				Promise.resolve({
+					data: [modelInfo],
+				}),
+		})
+	}
+
+	let handler: LiteLlmHandler
+
+	const mockHandlerChat = () => {
+		sinon.stub(handler, "ensureClient" as any).returns(fakeClient)
+	}
+
+	const initializeHandler = (model: string) => {
+		handler = new LiteLlmHandler({
+			liteLlmApiKey: "test-api-key",
+			liteLlmBaseUrl: "http://localhost:4000",
+			liteLlmUsePromptCache: true,
+			liteLlmModelId: model,
+		})
+
+		mockHandlerChat()
+	}
+
+	beforeEach(() => {
+		fakeClient.chat.completions.create.resetHistory()
+
+		mockFetchForTesting(mockFetch, () => {
+			return new Promise((resolve) => {
+				doneMockingFetch = resolve
+			})
+		})
+
+		// Configure the stub to return a stream that closes immediately with usage data
+		fakeClient.chat.completions.create.resolves(
+			createAsyncIterable([
+				{
+					choices: [{ delta: { content: "test response" } }],
+				},
+				{
+					choices: [{}],
+					usage: {
+						prompt_tokens: 100,
+						completion_tokens: 50,
+						cache_creation_input_tokens: 20,
+						cache_read_input_tokens: 10,
+					},
+				},
+			]),
+		)
+	})
+
+	afterEach(() => {
+		sinon.reset()
+		doneMockingFetch(void 0)
+	})
+
+	const createAsyncIterable = (data: any[] = []) => {
+		return {
+			[Symbol.asyncIterator]: async function* () {
+				yield* data
+			},
+		}
+	}
+
+	describe("prompt cache", () => {
+		const setModelData = (model: string, supportsPromptCaching: boolean) => {
+			mockModelFetch({
+				model_name: model,
+				litellm_params: {
+					model,
+				},
+				model_info: {
+					supports_prompt_caching: supportsPromptCaching,
+					input_cost_per_token: 0.01,
+					output_cost_per_token: 0.02,
+				},
+			})
+		}
+
+		describe("when the model doesn't support prompt caching", () => {
+			const model = "openai/gpt-5"
+
+			beforeEach(() => {
+				initializeHandler(model)
+				setModelData(model, false)
+			})
+
+			it("sends the system prompt and messages with the openai format", async () => {
+				const systemPrompt = "Test System Prompt"
+				const messages: ClineStorageMessage[] = [
+					{
+						role: "user",
+						content: "first message",
+					},
+					{
+						role: "assistant",
+						content: "first response",
+					},
+					{
+						role: "user",
+						content: [
+							{
+								type: "text",
+								text: "test",
+							},
+							{
+								type: "text",
+								text: "second message",
+							},
+						],
+					},
+				]
+
+				for await (const _ of handler.createMessage(systemPrompt, messages)) {
+				}
+
+				sinon.assert.calledOnce(fakeClient.chat.completions.create)
+
+				const callArgs = fakeClient.chat.completions.create.getCall(0).args[0]
+
+				const systemPromptMessage = callArgs.messages.shift()
+				expect(systemPromptMessage).to.deep.equal({
+					role: "system",
+					content: systemPrompt,
+				})
+
+				expect(callArgs.messages).to.deep.equal(convertToOpenAiMessages(messages))
+			})
+		})
+
+		describe("when the model supports prompt caching", () => {
+			const model = "anthropic/claude-sonnet-4-20250514"
+
+			beforeEach(() => {
+				initializeHandler(model)
+
+				setModelData(model, true)
+			})
+
+			it("inserts the cache control in the system prompt and the last two user messages", async () => {
+				const systemPrompt = "Test System Prompt"
+				const messages: ClineStorageMessage[] = [
+					{
+						role: "user",
+						content: "first message",
+					},
+					{
+						role: "assistant",
+						content: "first response",
+					},
+					{
+						role: "user",
+						content: [
+							{
+								type: "text",
+								text: "test",
+							},
+							{
+								type: "text",
+								text: "second message",
+							},
+						],
+					},
+				]
+
+				for await (const _ of handler.createMessage(systemPrompt, messages)) {
+				}
+
+				sinon.assert.calledOnce(fakeClient.chat.completions.create)
+
+				const callArgs = fakeClient.chat.completions.create.getCall(0).args[0]
+
+				expect(callArgs.messages[0]).to.deep.equal({
+					role: "system",
+					content: [
+						{
+							text: systemPrompt,
+							type: "text",
+							cache_control: {
+								type: "ephemeral",
+							},
+						},
+					],
+				})
+
+				const sentMessages = callArgs.messages
+				expect(sentMessages.length).to.equal(4)
+
+				const firstUserMessage = sentMessages[1]
+
+				expect(firstUserMessage).to.deep.equal({
+					role: "user",
+					content: [
+						{
+							type: "text",
+							text: "first message",
+							cache_control: {
+								type: "ephemeral",
+							},
+						},
+					],
+				})
+
+				const lastUserMessage = sentMessages[3]
+				expect(lastUserMessage.content[0]).to.deep.equal({
+					type: "text",
+					text: "test",
+				})
+
+				const lastContentBlock = lastUserMessage.content[lastUserMessage.content.length - 1]
+				expect(lastContentBlock).to.deep.equal({
+					type: "text",
+					text: "second message",
+					cache_control: {
+						type: "ephemeral",
+					},
+				})
+
+				expect(callArgs.model).to.be.a("string")
+				expect(callArgs.stream).to.equal(true)
+				expect(callArgs.stream_options).to.deep.equal({ include_usage: true })
+			})
+		})
+	})
+
+	describe("getModel", () => {
+		let stateManagerStub: sinon.SinonStub
+
+		beforeEach(() => {
+			stateManagerStub = sinon.stub(StateManager, "get").returns({
+				getModelInfo: () => null,
+			} as any)
+		})
+
+		afterEach(() => {
+			stateManagerStub.restore()
+		})
+
+		it("returns sane defaults when no liteLlmModelInfo option is provided", () => {
+			const h = new LiteLlmHandler({
+				liteLlmApiKey: "test",
+				liteLlmModelId: "some-model",
+			})
+			const model = h.getModel()
+			expect(model.id).to.equal("some-model")
+			expect(model.info.contextWindow).to.equal(liteLlmModelInfoSaneDefaults.contextWindow)
+		})
+
+		it("returns user-configured model info when liteLlmModelInfo is provided and no cache exists", () => {
+			const h = new LiteLlmHandler({
+				liteLlmApiKey: "test",
+				liteLlmModelId: "claude-sonnet-4-6",
+				liteLlmModelInfo: {
+					contextWindow: 1_000_000,
+					maxTokens: 8192,
+					supportsImages: true,
+					supportsPromptCache: true,
+					inputPrice: 3,
+					outputPrice: 15,
+					cacheWritesPrice: 3.75,
+					cacheReadsPrice: 0.3,
+				},
+			})
+			const model = h.getModel()
+			expect(model.id).to.equal("claude-sonnet-4-6")
+			expect(model.info.contextWindow).to.equal(1_000_000)
+		})
+
+		it("prefers StateManager cached model info over user-configured liteLlmModelInfo", () => {
+			const cachedInfo = {
+				contextWindow: 200_000,
+				maxTokens: 4096,
+				supportsImages: true,
+				supportsPromptCache: true,
+				inputPrice: 0,
+				outputPrice: 0,
+			}
+			stateManagerStub.returns({
+				getModelInfo: () => cachedInfo,
+			} as any)
+
+			const h = new LiteLlmHandler({
+				liteLlmApiKey: "test",
+				liteLlmModelId: "claude-sonnet-4-6",
+				liteLlmModelInfo: {
+					contextWindow: 1_000_000,
+					maxTokens: 8192,
+					supportsImages: true,
+					supportsPromptCache: true,
+					inputPrice: 3,
+					outputPrice: 15,
+					cacheWritesPrice: 3.75,
+					cacheReadsPrice: 0.3,
+				},
+			})
+			const model = h.getModel()
+			expect(model.info.contextWindow).to.equal(200_000)
+		})
+	})
+})

apps/vscode/src/core/api/providers/__tests__/moonshot.test.ts

@@ -0,0 +1,53 @@
+import "should"
+import { moonshotModels } from "@shared/api"
+import type { ClineStorageMessage } from "@shared/messages/content"
+import sinon from "sinon"
+import { MoonshotHandler } from "../moonshot"
+
+interface MoonshotRequestPayload {
+	model: string
+	temperature: number
+	max_tokens: number
+}
+
+describe("MoonshotHandler", () => {
+	afterEach(() => {
+		sinon.restore()
+	})
+
+	const createAsyncIterable = (data: unknown[] = []): AsyncIterable<unknown> => ({
+		[Symbol.asyncIterator]: async function* () {
+			yield* data
+		},
+	})
+
+	it("supports kimi-k2.6 model metadata", async () => {
+		const handler = new MoonshotHandler({
+			moonshotApiKey: "test-api-key",
+			apiModelId: "kimi-k2.6",
+		})
+
+		const model = handler.getModel()
+		model.id.should.equal("kimi-k2.6")
+		model.info.should.deepEqual(moonshotModels["kimi-k2.6"])
+
+		const createStub = sinon.stub().resolves(createAsyncIterable([]))
+		sinon.stub(handler as unknown as { ensureClient: () => unknown }, "ensureClient").returns({
+			chat: {
+				completions: {
+					create: createStub,
+				},
+			},
+		})
+
+		const messages: ClineStorageMessage[] = [{ role: "user", content: "hi" }]
+		for await (const _chunk of handler.createMessage("system", messages)) {
+			// Consume stream to trigger request execution.
+		}
+
+		const payload = createStub.firstCall.args[0] as MoonshotRequestPayload
+		payload.model.should.equal("kimi-k2.6")
+		payload.temperature.should.equal(moonshotModels["kimi-k2.6"].temperature)
+		payload.max_tokens.should.equal(moonshotModels["kimi-k2.6"].maxTokens)
+	})
+})

apps/vscode/src/core/api/providers/__tests__/oca.test.ts

@@ -0,0 +1,93 @@
+import { expect } from "chai"
+import { afterEach, describe, it } from "mocha"
+import sinon from "sinon"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { ApiFormat } from "@/shared/proto/index.cline"
+import { OcaHandler } from "../oca"
+
+const messages: ClineStorageMessage[] = [{ role: "user", content: "Hello" }]
+
+async function collectChunks(stream: AsyncGenerator<any>) {
+	const chunks: any[] = []
+	for await (const chunk of stream) {
+		chunks.push(chunk)
+	}
+	return chunks
+}
+
+describe("OcaHandler.createMessage", () => {
+	afterEach(() => {
+		sinon.restore()
+	})
+
+	it("routes OPENAI_RESPONSES models to createMessageResponsesApi", async () => {
+		const handler = new OcaHandler({
+			ocaModelInfo: { apiFormat: ApiFormat.OPENAI_RESPONSES } as any,
+		})
+
+		const chatStub = sinon.stub(handler as any, "createMessageChatApi").callsFake(async function* () {
+			yield { type: "text", text: "chat" }
+		})
+		const responsesStub = sinon.stub(handler as any, "createMessageResponsesApi").callsFake(async function* () {
+			yield { type: "text", text: "responses" }
+		})
+		const messagesStub = sinon.stub(handler as any, "createMessageMessagesApi").callsFake(async function* () {
+			yield { type: "text", text: "messages" }
+		})
+
+		const chunks = await collectChunks(handler.createMessage("system", messages))
+
+		expect(chunks).to.deep.equal([{ type: "text", text: "responses" }])
+		sinon.assert.notCalled(chatStub)
+		sinon.assert.calledOnce(responsesStub)
+		sinon.assert.notCalled(messagesStub)
+	})
+
+	it("routes ANTHROPIC_CHAT models to createMessageMessagesApi", async () => {
+		const handler = new OcaHandler({
+			ocaModelInfo: { apiFormat: ApiFormat.ANTHROPIC_CHAT } as any,
+		})
+
+		const chatStub = sinon.stub(handler as any, "createMessageChatApi").callsFake(async function* () {
+			yield { type: "text", text: "chat" }
+		})
+		const responsesStub = sinon.stub(handler as any, "createMessageResponsesApi").callsFake(async function* () {
+			yield { type: "text", text: "responses" }
+		})
+		const messagesStub = sinon.stub(handler as any, "createMessageMessagesApi").callsFake(async function* () {
+			yield { type: "text", text: "messages" }
+		})
+
+		const chunks = await collectChunks(handler.createMessage("system", messages))
+
+		expect(chunks).to.deep.equal([{ type: "text", text: "messages" }])
+		sinon.assert.notCalled(chatStub)
+		sinon.assert.notCalled(responsesStub)
+		sinon.assert.calledOnce(messagesStub)
+	})
+
+	it("defaults to createMessageChatApi for OPENAI_CHAT and undefined apiFormat", async () => {
+		for (const apiFormat of [ApiFormat.OPENAI_CHAT, undefined]) {
+			const handler = new OcaHandler({
+				ocaModelInfo: { apiFormat } as any,
+			})
+
+			const chatStub = sinon.stub(handler as any, "createMessageChatApi").callsFake(async function* () {
+				yield { type: "text", text: "chat" }
+			})
+			const responsesStub = sinon.stub(handler as any, "createMessageResponsesApi").callsFake(async function* () {
+				yield { type: "text", text: "responses" }
+			})
+			const messagesStub = sinon.stub(handler as any, "createMessageMessagesApi").callsFake(async function* () {
+				yield { type: "text", text: "messages" }
+			})
+
+			const chunks = await collectChunks(handler.createMessage("system", messages))
+
+			expect(chunks).to.deep.equal([{ type: "text", text: "chat" }])
+			sinon.assert.calledOnce(chatStub)
+			sinon.assert.notCalled(responsesStub)
+			sinon.assert.notCalled(messagesStub)
+		}
+	})
+})

apps/vscode/src/core/api/providers/__tests__/ollama.test.ts

@@ -0,0 +1,231 @@
+import { afterEach, before, beforeEach, describe, it } from "mocha"
+import "should"
+import { ApiHandlerOptions } from "@shared/api"
+import axios from "axios"
+import sinon from "sinon"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { OllamaHandler } from "../ollama"
+
+describe("OllamaHandler", () => {
+	let ollamaAvailable = false
+
+	// Check if Ollama is running before running tests
+	before(async function () {
+		this.timeout(5000)
+		try {
+			await axios.get("http://localhost:11434/api/version", { timeout: 2000 })
+			ollamaAvailable = true
+		} catch (_error) {
+			console.log("Ollama server not available, skipping tests")
+			ollamaAvailable = false
+		}
+	})
+	let handler: OllamaHandler
+	let options: ApiHandlerOptions
+	let clock: sinon.SinonFakeTimers
+
+	beforeEach(() => {
+		options = {
+			actModeOllamaModelId: "llama2",
+			ollamaBaseUrl: "http://localhost:11434",
+		}
+		handler = new OllamaHandler(options)
+		// Use fake timers for testing timeouts
+		clock = sinon.useFakeTimers()
+	})
+
+	afterEach(() => {
+		clock.restore()
+		sinon.restore()
+	})
+
+	describe("createMessage", () => {
+		it("should handle successful responses", async function () {
+			if (!ollamaAvailable) {
+				this.skip()
+			}
+			this.timeout(5000)
+			// Ensure client is initialized
+			const client = (handler as any).ensureClient()
+			// Mock the Ollama client's chat method
+			const chatStub = sinon.stub(client, "chat").resolves({
+				[Symbol.asyncIterator]: async function* () {
+					yield {
+						message: { content: "Hello, world!" },
+						eval_count: 10,
+						prompt_eval_count: 20,
+					}
+				},
+			} as any)
+
+			const systemPrompt = "You are a helpful assistant."
+			const messages: ClineStorageMessage[] = [{ role: "user", content: "Hello" }]
+
+			const result = []
+			const usageInfo = []
+
+			// Collect the results
+			for await (const chunk of handler.createMessage(systemPrompt, messages)) {
+				if (chunk.type === "text") {
+					result.push(chunk.text)
+				} else if (chunk.type === "usage") {
+					usageInfo.push({
+						inputTokens: chunk.inputTokens,
+						outputTokens: chunk.outputTokens,
+					})
+				}
+			}
+
+			// Verify the results
+			result.should.deepEqual(["Hello, world!"])
+			usageInfo.should.deepEqual([{ inputTokens: 20, outputTokens: 10 }])
+			chatStub.calledOnce.should.be.true()
+		})
+
+		it("should handle timeout errors", async function () {
+			if (!ollamaAvailable) {
+				this.skip()
+			}
+			this.timeout(10000)
+			// Restore real timers for this test
+			clock.restore()
+
+			// Create a handler with a very short timeout for testing
+			const testHandler = new OllamaHandler(options)
+
+			// Replace the createMessage method with one that has a shorter timeout
+			testHandler.createMessage = async function* (_systemPrompt, _messages) {
+				try {
+					// Create a promise that rejects after a short timeout
+					const timeoutPromise = new Promise<never>((_, reject) => {
+						setTimeout(() => reject(new Error("Ollama request timed out after 120 seconds")), 100)
+					})
+
+					// Create a promise that never resolves
+					const neverPromise = new Promise(() => {})
+
+					// Race them
+					await Promise.race([timeoutPromise, neverPromise])
+				} catch (error: any) {
+					// Enhance error reporting
+					console.error(`Ollama API error: ${error.message}`)
+					throw error
+				}
+			}
+
+			const systemPrompt = "You are a helpful assistant."
+			const messages: ClineStorageMessage[] = [{ role: "user", content: "Hello" }]
+
+			// Start the request and catch the error
+			let errorMessage = ""
+			try {
+				for await (const _ of testHandler.createMessage(systemPrompt, messages)) {
+					// This should not be reached
+				}
+			} catch (error: any) {
+				errorMessage = error.message
+			}
+
+			// Check the result
+			errorMessage.should.equal("Ollama request timed out after 120 seconds")
+
+			// Restore the fake timers for other tests
+			clock = sinon.useFakeTimers()
+		})
+
+		it("should retry on errors when using the withRetry decorator", async function () {
+			if (!ollamaAvailable) {
+				this.skip()
+			}
+			this.timeout(10000)
+			// Restore real timers for this test
+			clock.restore()
+
+			// Ensure client is initialized and mock the Ollama client's chat method to fail on first call and succeed on second
+			const client = (handler as any).ensureClient()
+			const chatStub = sinon.stub(client, "chat")
+
+			// First call throws an error
+			chatStub.onFirstCall().rejects(new Error("API Error"))
+
+			// Second call succeeds
+			chatStub.onSecondCall().resolves({
+				[Symbol.asyncIterator]: async function* () {
+					yield {
+						message: { content: "Success after retry" },
+					}
+				},
+			} as any)
+
+			const systemPrompt = "You are a helpful assistant."
+			const messages: ClineStorageMessage[] = [{ role: "user", content: "Hello" }]
+
+			const result = []
+
+			// Add a small delay to ensure the retry mechanism has time to work
+			await new Promise((resolve) => setTimeout(resolve, 100))
+
+			// Collect the results
+			for await (const chunk of handler.createMessage(systemPrompt, messages)) {
+				if (chunk.type === "text") {
+					result.push(chunk.text)
+				}
+			}
+
+			// Verify the results
+			result.should.deepEqual(["Success after retry"])
+			chatStub.calledTwice.should.be.true()
+
+			// Restore the fake timers for other tests
+			clock = sinon.useFakeTimers()
+		})
+
+		it("should handle stream processing errors", async function () {
+			if (!ollamaAvailable) {
+				this.skip()
+			}
+			this.timeout(10000)
+			// Restore real timers for this test
+			clock.restore()
+
+			// Create a handler with a custom implementation for testing
+			const testHandler = new OllamaHandler(options)
+
+			// Replace the createMessage method with one that simulates a stream error
+			testHandler.createMessage = async function* (_systemPrompt, _messages) {
+				// First yield a successful chunk
+				yield {
+					type: "text",
+					text: "Partial response",
+				}
+
+				// Then throw an error in the stream
+				throw new Error("Ollama stream processing error: Stream error")
+			}
+
+			const systemPrompt = "You are a helpful assistant."
+			const messages: ClineStorageMessage[] = [{ role: "user", content: "Hello" }]
+
+			const result = []
+
+			// Collect the results and catch the error
+			let errorMessage = ""
+			try {
+				for await (const chunk of testHandler.createMessage(systemPrompt, messages)) {
+					if (chunk.type === "text") {
+						result.push(chunk.text)
+					}
+				}
+			} catch (error: any) {
+				errorMessage = error.message
+			}
+
+			// Verify the results
+			errorMessage.should.equal("Ollama stream processing error: Stream error")
+			result.should.deepEqual(["Partial response"])
+
+			// Restore the fake timers for other tests
+			clock = sinon.useFakeTimers()
+		})
+	})
+})

apps/vscode/src/core/api/providers/__tests__/openrouter.test.ts

@@ -0,0 +1,164 @@
+import "should"
+import { openRouterDefaultModelInfo } from "@shared/api"
+import sinon from "sinon"
+import { OpenRouterHandler } from "../openrouter"
+
+describe("OpenRouterHandler", () => {
+	afterEach(() => {
+		sinon.restore()
+	})
+
+	const createAsyncIterable = (data: any[] = []) => ({
+		[Symbol.asyncIterator]: async function* () {
+			yield* data
+		},
+	})
+
+	const tools = [{ type: "function", function: { name: "read_file", description: "", parameters: { type: "object" } } }] as any
+
+	it("should handle usage-only chunks when delta is missing", async () => {
+		const handler = new OpenRouterHandler({
+			openRouterApiKey: "test-api-key",
+		})
+		const fakeClient = {
+			chat: {
+				completions: {
+					create: sinon.stub().resolves(
+						createAsyncIterable([
+							{
+								choices: [{}],
+								usage: {
+									prompt_tokens: 13,
+									completion_tokens: 5,
+								},
+							},
+						]),
+					),
+				},
+			},
+		}
+		sinon.stub(handler as any, "ensureClient").returns(fakeClient as any)
+		sinon.stub(handler, "getModel").returns({
+			id: "openai/gpt-4o-mini",
+			info: openRouterDefaultModelInfo,
+		})
+
+		const chunks: any[] = []
+		for await (const chunk of handler.createMessage("system", [{ role: "user", content: "hi" }])) {
+			chunks.push(chunk)
+		}
+
+		chunks.should.deepEqual([
+			{
+				type: "usage",
+				cacheWriteTokens: 0,
+				cacheReadTokens: 0,
+				inputTokens: 13,
+				outputTokens: 5,
+				totalCost: 0,
+			},
+		])
+	})
+
+	it("should read cache_write_tokens from prompt_tokens_details", async () => {
+		const handler = new OpenRouterHandler({
+			openRouterApiKey: "test-api-key",
+		})
+		const fakeClient = {
+			chat: {
+				completions: {
+					create: sinon.stub().resolves(
+						createAsyncIterable([
+							{
+								choices: [{}],
+								usage: {
+									prompt_tokens: 1000,
+									completion_tokens: 200,
+									prompt_tokens_details: {
+										cached_tokens: 500,
+										cache_write_tokens: 300,
+									},
+								},
+							},
+						]),
+					),
+				},
+			},
+		}
+		sinon.stub(handler as any, "ensureClient").returns(fakeClient as any)
+		sinon.stub(handler, "getModel").returns({
+			id: "anthropic/claude-sonnet-4.6",
+			info: openRouterDefaultModelInfo,
+		})
+
+		const chunks: any[] = []
+		for await (const chunk of handler.createMessage("system", [{ role: "user", content: "hi" }])) {
+			chunks.push(chunk)
+		}
+
+		chunks.should.deepEqual([
+			{
+				type: "usage",
+				cacheWriteTokens: 300,
+				cacheReadTokens: 500,
+				inputTokens: 200,
+				outputTokens: 200,
+				totalCost: 0,
+			},
+		])
+	})
+
+	type ParallelToolCallsTestCase = {
+		modelId: string
+		enableParallelToolCalling: boolean
+		expectedParallelToolCalls: boolean
+	}
+
+	const parallelToolCallsTestCases: ParallelToolCallsTestCase[] = [
+		{
+			modelId: "openai/gpt-4o-mini",
+			enableParallelToolCalling: true,
+			expectedParallelToolCalls: true,
+		},
+		{
+			modelId: "openai/gpt-4o-mini",
+			enableParallelToolCalling: false,
+			expectedParallelToolCalls: false,
+		},
+		{
+			modelId: "google/gemini-3-flash-preview",
+			enableParallelToolCalling: true,
+			expectedParallelToolCalls: true,
+		},
+	]
+
+	for (const testCase of parallelToolCallsTestCases) {
+		const settingLabel = testCase.enableParallelToolCalling ? "enabled" : "disabled"
+		it(`should set parallel_tool_calls=${testCase.expectedParallelToolCalls} for ${testCase.modelId} when setting is ${settingLabel}`, async () => {
+			const handler = new OpenRouterHandler({
+				openRouterApiKey: "test-api-key",
+				enableParallelToolCalling: testCase.enableParallelToolCalling,
+			})
+			const createStub = sinon.stub().resolves(createAsyncIterable([]))
+			const fakeClient = {
+				chat: {
+					completions: {
+						create: createStub,
+					},
+				},
+			}
+			sinon.stub(handler as any, "ensureClient").returns(fakeClient as any)
+			sinon.stub(handler, "getModel").returns({
+				id: testCase.modelId,
+				info: openRouterDefaultModelInfo,
+			})
+
+			for await (const _chunk of handler.createMessage("system", [{ role: "user", content: "hi" }], tools)) {
+				// drain stream
+			}
+
+			const payload = createStub.firstCall.args[0]
+			payload.parallel_tool_calls.should.equal(testCase.expectedParallelToolCalls)
+		})
+	}
+})

apps/vscode/src/core/api/providers/__tests__/sapaicore.test.ts

@@ -0,0 +1,132 @@
+import "should"
+import { Anthropic } from "@anthropic-ai/sdk"
+import { SapAiCoreHandler } from "../sapaicore"
+
+describe("SapAiCoreHandler", () => {
+	let handler: SapAiCoreHandler
+
+	beforeEach(() => {
+		const mockOptions = {
+			sapAiCoreClientId: "test-client-id",
+			sapAiCoreClientSecret: "test-client-secret",
+			sapAiCoreTokenUrl: "https://test.auth.sap.com",
+			sapAiResourceGroup: "default",
+			sapAiCoreBaseUrl: "https://test.api.sap.com",
+			apiModelId: "anthropic--claude-3.5-sonnet",
+		}
+		handler = new SapAiCoreHandler(mockOptions)
+	})
+
+	describe("image processing", () => {
+		// Test image processing through the public interface
+		// This tests the complete flow including processImageContent internally
+
+		it("should handle image processing for Claude 4 models", () => {
+			// Create handler with Claude 4 model
+			const claude4Handler = new SapAiCoreHandler({
+				sapAiCoreClientId: "test-client-id",
+				sapAiCoreClientSecret: "test-client-secret",
+				sapAiCoreTokenUrl: "https://test.auth.sap.com",
+				sapAiResourceGroup: "default",
+				sapAiCoreBaseUrl: "https://test.api.sap.com",
+				apiModelId: "anthropic--claude-4-sonnet",
+			})
+
+			const model = claude4Handler.getModel()
+			model.id.should.equal("anthropic--claude-4-sonnet")
+			model.info.should.have.property("supportsImages", true)
+		})
+
+		it("should create proper user readable request with images", () => {
+			const testImageData =
+				"iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mP8/5+hHgAHggJ/PchI7wAAAABJRU5ErkJggg=="
+
+			const userContent: (Anthropic.TextBlockParam | Anthropic.ImageBlockParam)[] = [
+				{
+					type: "text",
+					text: "Here's an image:",
+				},
+				{
+					type: "image",
+					source: {
+						type: "base64",
+						media_type: "image/png",
+						data: testImageData,
+					},
+				},
+			]
+
+			const result = handler.createUserReadableRequest(userContent)
+
+			result.should.have.property("model")
+			result.should.have.property("max_tokens")
+			result.should.have.property("system")
+			result.should.have.property("messages")
+			result.messages.should.be.Array()
+			result.messages[1].should.have.property("role", "user")
+			result.messages[1].should.have.property("content", userContent)
+		})
+
+		it("should support different Claude model variants", () => {
+			const modelVariants = [
+				"anthropic--claude-4.6-sonnet",
+				"anthropic--claude-4-sonnet",
+				"anthropic--claude-4-opus",
+				"anthropic--claude-3.7-sonnet",
+				"anthropic--claude-3.5-sonnet",
+				"anthropic--claude-3-sonnet",
+				"anthropic--claude-3-haiku",
+				"anthropic--claude-3-opus",
+			]
+
+			modelVariants.forEach((modelId) => {
+				const testHandler = new SapAiCoreHandler({
+					apiModelId: modelId,
+				})
+
+				const model = testHandler.getModel()
+				model.id.should.equal(modelId)
+				model.info.should.have.property("maxTokens")
+				model.info.should.have.property("contextWindow")
+			})
+		})
+	})
+
+	describe("getModel", () => {
+		it("should return default model when no apiModelId is provided", () => {
+			const result = handler.getModel()
+			result.should.have.property("id")
+			result.should.have.property("info")
+			result.info.should.have.property("maxTokens")
+		})
+
+		it("should return specified model when apiModelId is provided", () => {
+			const customHandler = new SapAiCoreHandler({
+				apiModelId: "anthropic--claude-4-sonnet",
+			})
+
+			const result = customHandler.getModel()
+			result.id.should.equal("anthropic--claude-4-sonnet")
+		})
+	})
+
+	describe("createUserReadableRequest", () => {
+		it("should create a readable request format", () => {
+			const userContent: Anthropic.TextBlockParam[] = [
+				{
+					type: "text",
+					text: "Hello, world!",
+				},
+			]
+
+			const result = handler.createUserReadableRequest(userContent)
+
+			result.should.have.property("model")
+			result.should.have.property("max_tokens")
+			result.should.have.property("system")
+			result.should.have.property("messages")
+			result.should.have.property("tools")
+			result.should.have.property("tool_choice")
+		})
+	})
+})

apps/vscode/src/core/api/providers/__tests__/vercel-ai-gateway.test.ts

@@ -0,0 +1,138 @@
+import "should"
+import { openRouterDefaultModelId, openRouterDefaultModelInfo } from "@shared/api"
+import sinon from "sinon"
+import { VercelAIGatewayHandler } from "../vercel-ai-gateway"
+
+describe("VercelAIGatewayHandler", () => {
+	afterEach(() => {
+		sinon.restore()
+	})
+
+	const createAsyncIterable = (data: any[] = []) => ({
+		[Symbol.asyncIterator]: async function* () {
+			yield* data
+		},
+	})
+
+	describe("getModel", () => {
+		it("should return configured model and info when both are provided", () => {
+			const customModelInfo = {
+				...openRouterDefaultModelInfo,
+				maxTokens: 123456,
+			}
+
+			const handler = new VercelAIGatewayHandler({
+				openRouterModelId: "google/gemini-3.1-pro-preview",
+				openRouterModelInfo: customModelInfo,
+			})
+
+			const result = handler.getModel()
+			result.id.should.equal("google/gemini-3.1-pro-preview")
+			result.info.should.deepEqual(customModelInfo)
+		})
+
+		it("should preserve configured model ID when model info is missing", () => {
+			const handler = new VercelAIGatewayHandler({
+				openRouterModelId: "google/gemini-3.1-pro-preview",
+			})
+
+			const result = handler.getModel()
+			result.id.should.equal("google/gemini-3.1-pro-preview")
+			result.info.should.deepEqual(openRouterDefaultModelInfo)
+		})
+
+		it("should fall back to default model when model ID is missing", () => {
+			const handler = new VercelAIGatewayHandler({})
+			const result = handler.getModel()
+
+			result.id.should.equal(openRouterDefaultModelId)
+			result.info.should.deepEqual(openRouterDefaultModelInfo)
+		})
+	})
+
+	describe("createMessage", () => {
+		it("should handle usage-only chunks when delta is missing", async () => {
+			const handler = new VercelAIGatewayHandler({
+				vercelAiGatewayApiKey: "test-api-key",
+			})
+			const fakeClient = {
+				chat: {
+					completions: {
+						create: sinon.stub().resolves(
+							createAsyncIterable([
+								{
+									choices: [{}],
+									usage: {
+										prompt_tokens: 11,
+										completion_tokens: 7,
+									},
+								},
+							]),
+						),
+					},
+				},
+			}
+			sinon.stub(handler as any, "ensureClient").returns(fakeClient as any)
+
+			const chunks: any[] = []
+			for await (const chunk of handler.createMessage("system", [{ role: "user", content: "hi" }])) {
+				chunks.push(chunk)
+			}
+
+			chunks.should.deepEqual([
+				{
+					type: "usage",
+					cacheWriteTokens: 0,
+					cacheReadTokens: 0,
+					inputTokens: 11,
+					outputTokens: 7,
+					totalCost: 0,
+				},
+			])
+		})
+
+		it("should read Anthropic-style cache creation and read tokens from usage chunks", async () => {
+			const handler = new VercelAIGatewayHandler({
+				vercelAiGatewayApiKey: "test-api-key",
+			})
+			const fakeClient = {
+				chat: {
+					completions: {
+						create: sinon.stub().resolves(
+							createAsyncIterable([
+								{
+									choices: [{}],
+									usage: {
+										prompt_tokens: 1000,
+										completion_tokens: 200,
+										prompt_tokens_details: {
+											cached_tokens: 500,
+										},
+										cache_creation_input_tokens: 300,
+									},
+								},
+							]),
+						),
+					},
+				},
+			}
+			sinon.stub(handler as any, "ensureClient").returns(fakeClient as any)
+
+			const chunks: any[] = []
+			for await (const chunk of handler.createMessage("system", [{ role: "user", content: "hi" }])) {
+				chunks.push(chunk)
+			}
+
+			chunks.should.deepEqual([
+				{
+					type: "usage",
+					cacheWriteTokens: 300,
+					cacheReadTokens: 500,
+					inputTokens: 200,
+					outputTokens: 200,
+					totalCost: 0,
+				},
+			])
+		})
+	})
+})

apps/vscode/src/core/api/providers/__tests__/vertex.test.ts

@@ -0,0 +1,23 @@
+import "should"
+import { vertexGlobalModels } from "@shared/api"
+import { VertexHandler } from "../vertex"
+
+describe("VertexHandler", () => {
+	it("supports Gemini 3.5 Flash model metadata", () => {
+		const handler = new VertexHandler({
+			vertexProjectId: "test-project",
+			vertexRegion: "global",
+			apiModelId: "gemini-3.5-flash",
+		})
+
+		const model = handler.getModel()
+		model.id.should.equal("gemini-3.5-flash")
+		model.info.contextWindow!.should.equal(1_048_576)
+		model.info.inputPrice!.should.equal(1.5)
+		model.info.outputPrice!.should.equal(9)
+		model.info.cacheReadsPrice!.should.equal(0.15)
+		model.info.supportsGlobalEndpoint!.should.equal(true)
+		model.info.supportsReasoning!.should.equal(true)
+		vertexGlobalModels.should.have.property("gemini-3.5-flash")
+	})
+})

apps/vscode/src/core/api/providers/__tests__/wandb.test.ts

@@ -0,0 +1,42 @@
+import "should"
+import { openAiModelInfoSaneDefaults, wandbDefaultModelId, wandbModels } from "@shared/api"
+import { WandbHandler } from "../wandb"
+
+describe("WandbHandler", () => {
+	it("returns known catalog model metadata when model id is recognized", () => {
+		const modelId = "meta-llama/Llama-3.3-70B-Instruct"
+		const handler = new WandbHandler({
+			wandbApiKey: "test-api-key",
+			apiModelId: modelId,
+		})
+
+		const model = handler.getModel()
+
+		model.id.should.equal(modelId)
+		model.info.should.deepEqual(wandbModels[modelId])
+	})
+
+	it("passes through an explicit unknown model id instead of silently falling back", () => {
+		const unknownModelId = "moonshotai/Kimi-K2.5"
+		const handler = new WandbHandler({
+			wandbApiKey: "test-api-key",
+			apiModelId: unknownModelId,
+		})
+
+		const model = handler.getModel()
+
+		model.id.should.equal(unknownModelId)
+		model.info.should.deepEqual(openAiModelInfoSaneDefaults)
+	})
+
+	it("uses the default W&B model when no model id is configured", () => {
+		const handler = new WandbHandler({
+			wandbApiKey: "test-api-key",
+		})
+
+		const model = handler.getModel()
+
+		model.id.should.equal(wandbDefaultModelId)
+		model.info.should.deepEqual(wandbModels[wandbDefaultModelId])
+	})
+})

apps/vscode/src/core/api/providers/aihubmix.ts

@@ -0,0 +1,327 @@
+import { Anthropic } from "@anthropic-ai/sdk"
+import { GenerateContentConfig, GoogleGenAI } from "@google/genai"
+import { ModelInfo } from "@shared/api"
+import OpenAI from "openai"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { ApiHandler, CommonApiHandlerOptions } from "../index"
+import { withRetry } from "../retry"
+import { sanitizeAnthropicMessages } from "../transform/anthropic-format"
+import { convertAnthropicMessageToGemini } from "../transform/gemini-format"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+
+interface AIhubmixHandlerOptions extends CommonApiHandlerOptions {
+	apiKey?: string
+	baseURL?: string
+	appCode?: string
+	modelId?: string
+	modelInfo?: ModelInfo
+	thinkingBudgetTokens?: number
+}
+
+export class AIhubmixHandler implements ApiHandler {
+	private options: AIhubmixHandlerOptions
+	private anthropicClient: Anthropic | undefined
+	private openaiClient: OpenAI | undefined
+	private geminiClient: GoogleGenAI | undefined
+
+	constructor(options: AIhubmixHandlerOptions) {
+		const { baseURL, appCode, ...rest } = options
+		this.options = {
+			baseURL: baseURL ?? "https://aihubmix.com",
+			appCode: appCode ?? "KUWF9311",
+			...rest,
+		}
+	}
+
+	private ensureAnthropicClient(): Anthropic {
+		if (!this.anthropicClient) {
+			if (!this.options.apiKey) {
+				throw new Error("AIhubmix API key is required")
+			}
+			try {
+				this.anthropicClient = new Anthropic({
+					apiKey: this.options.apiKey,
+					baseURL: this.options.baseURL,
+					defaultHeaders: {
+						"APP-Code": this.options.appCode,
+						...buildExternalBasicHeaders(),
+					},
+				})
+			} catch (error) {
+				throw new Error(`Error creating Anthropic client: ${error.message}`)
+			}
+		}
+		return this.anthropicClient
+	}
+
+	private ensureOpenaiClient(): OpenAI {
+		if (!this.openaiClient) {
+			if (!this.options.apiKey) {
+				throw new Error("AIhubmix API key is required")
+			}
+			try {
+				this.openaiClient = new OpenAI({
+					apiKey: this.options.apiKey,
+					baseURL: `${this.options.baseURL}/v1`,
+					defaultHeaders: {
+						"APP-Code": this.options.appCode,
+						...buildExternalBasicHeaders(),
+					},
+				})
+			} catch (error) {
+				throw new Error(`Error creating OpenAI client: ${error.message}`)
+			}
+		}
+		return this.openaiClient
+	}
+
+	private ensureGeminiClient(): GoogleGenAI {
+		if (!this.geminiClient) {
+			if (!this.options.apiKey) {
+				throw new Error("AIhubmix API key is required")
+			}
+			try {
+				this.geminiClient = new GoogleGenAI({
+					apiKey: this.options.apiKey,
+					httpOptions: {
+						baseUrl: `${this.options.baseURL}/gemini`,
+						headers: {
+							// @ts-expect-error
+							"APP-Code": this.options.appCode,
+							Authorization: `Bearer ${this.options.apiKey ?? ""}`,
+							...buildExternalBasicHeaders(),
+						},
+					},
+				})
+			} catch (error) {
+				throw new Error(`Error creating Gemini client: ${error.message}`)
+			}
+		}
+		return this.geminiClient
+	}
+
+	private routeModel(modelName: string): "anthropic" | "openai" | "gemini" | "openai-response" {
+		const id = modelName || ""
+		if (id.startsWith("claude")) {
+			return "anthropic"
+		}
+		if (id.startsWith("gemini") && !id.endsWith("-nothink") && !id.endsWith("-search")) {
+			return "gemini"
+		}
+		if (id === "gpt-5-pro" || id === "gpt-5-codex") {
+			return "openai-response"
+		}
+		return "openai"
+	}
+
+	private fixToolChoice(requestBody: any): any {
+		if (requestBody.tools?.length === 0 && requestBody.tool_choice) {
+			delete requestBody.tool_choice
+		}
+		return requestBody
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: any[]): ApiStream {
+		const modelId = this.options.modelId || ""
+		const route = this.routeModel(modelId)
+
+		switch (route) {
+			case "anthropic":
+				yield* this.createAnthropicMessage(systemPrompt, messages)
+				break
+			case "gemini":
+				yield* this.createGeminiMessage(systemPrompt, messages)
+				break
+			case "openai-response":
+				yield* this.createOpenaiResponseMessage(systemPrompt, messages)
+				break
+			case "openai":
+				yield* this.createOpenaiMessage(systemPrompt, messages)
+				break
+			default:
+				throw new Error(`Unsupported model route: ${route}`)
+		}
+	}
+
+	private async *createAnthropicMessage(systemPrompt: string, messages: any[]): ApiStream {
+		const client = this.ensureAnthropicClient()
+		const modelId = this.options.modelId || "claude-3-5-sonnet-20241022"
+
+		// Sanitize messages to remove Cline-specific fields like call_id that are not allowed by Anthropic API
+		const sanitizedMessages = sanitizeAnthropicMessages(messages, false)
+
+		const stream = await client.messages.create({
+			model: modelId,
+			temperature: 0,
+			max_tokens: this.options.modelInfo?.maxTokens || 8192,
+			system: [{ text: systemPrompt, type: "text" }],
+			messages: sanitizedMessages,
+			stream: true,
+		})
+
+		for await (const chunk of stream) {
+			switch (chunk?.type) {
+				case "message_start":
+					const usage = chunk.message.usage
+					yield {
+						type: "usage",
+						inputTokens: usage.input_tokens || 0,
+						outputTokens: usage.output_tokens || 0,
+						cacheWriteTokens: usage.cache_creation_input_tokens || undefined,
+						cacheReadTokens: usage.cache_read_input_tokens || undefined,
+					}
+					break
+				case "message_delta":
+					yield {
+						type: "usage",
+						inputTokens: 0,
+						outputTokens: chunk.usage.output_tokens || 0,
+					}
+					break
+				case "content_block_start":
+					if (chunk.content_block.type === "text") {
+						yield {
+							type: "text",
+							text: chunk.content_block.text,
+						}
+					}
+					break
+				case "content_block_delta":
+					if (chunk.delta.type === "text_delta") {
+						yield {
+							type: "text",
+							text: chunk.delta.text,
+						}
+					}
+					break
+			}
+		}
+	}
+
+	private async *createOpenaiResponseMessage(systemPrompt: string, messages: any[]): ApiStream {
+		const client = this.ensureOpenaiClient()
+		const modelId = this.options.modelId || "gpt-4o-mini"
+
+		const input = (messages || []).map((m: any) => {
+			const role = m.role || "user"
+			const contentArray = Array.isArray(m.content) ? m.content : [{ type: "text", text: m.content }]
+			const content = contentArray
+				.filter((c: any) => c != null)
+				.map((c: any) => {
+					if (c.type === "image" || c.type === "input_image" || c.type === "image_url") {
+						return { type: "input_image", image_url: c.image_url || c.url || c.source?.url }
+					}
+					const text = c.text ?? (typeof c === "string" ? c : "")
+					return { type: role === "assistant" ? "output_text" : "input_text", text }
+				})
+			return { role, content }
+		})
+
+		const stream = await (client as any).responses.stream({
+			model: modelId,
+			instructions: systemPrompt,
+			input,
+		})
+
+		for await (const event of stream as any) {
+			if (event?.type === "response.output_text.delta") {
+				yield { type: "text", text: event.delta || "" }
+				continue
+			}
+			if (event?.type === "response.completed") {
+				const usage = event.response?.usage || {}
+				yield {
+					type: "usage",
+					inputTokens: usage.input_tokens || 0,
+					outputTokens: usage.output_tokens || 0,
+				}
+				continue
+			}
+			if (event?.type === "response.error") {
+				throw new Error(event.error?.message || "responses error")
+			}
+		}
+	}
+
+	private async *createOpenaiMessage(systemPrompt: string, messages: any[]): ApiStream {
+		const client = this.ensureOpenaiClient()
+		const modelId = this.options.modelId || "gpt-4o-mini"
+
+		const openaiMessages = [{ role: "system", content: systemPrompt }, ...convertToOpenAiMessages(messages)]
+
+		const requestBody = {
+			model: modelId,
+			messages: openaiMessages,
+			temperature: 0,
+			stream: true,
+		}
+
+		const fixedRequestBody = this.fixToolChoice(requestBody)
+
+		const stream = await client.chat.completions.create(fixedRequestBody)
+
+		for await (const chunk of stream as any) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (chunk.usage) {
+				yield {
+					type: "usage",
+					inputTokens: chunk.usage.prompt_tokens || 0,
+					outputTokens: chunk.usage.completion_tokens || 0,
+				}
+			}
+		}
+	}
+
+	private async *createGeminiMessage(systemPrompt: string, messages: any[]): ApiStream {
+		const client = this.ensureGeminiClient()
+		const modelId = this.options.modelId || "gemini-2.0-flash-exp"
+
+		const contents = messages.map(convertAnthropicMessageToGemini)
+
+		const requestConfig: GenerateContentConfig = {
+			systemInstruction: systemPrompt,
+			temperature: 0,
+		}
+
+		if (this.options.thinkingBudgetTokens) {
+			requestConfig.thinkingConfig = {
+				thinkingBudget: this.options.thinkingBudgetTokens,
+				includeThoughts: true,
+			}
+		}
+
+		const stream = await client.models.generateContentStream({
+			model: modelId,
+			contents,
+			config: requestConfig,
+		})
+
+		for await (const chunk of stream as any) {
+			if (chunk?.text) {
+				yield { type: "text", text: chunk.text }
+			}
+		}
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		return {
+			id: this.options.modelId || "gpt-4o-mini",
+			info: this.options.modelInfo || {
+				maxTokens: 8192,
+				contextWindow: 128000,
+				supportsImages: true,
+				supportsPromptCache: false,
+				description: "AIhubmix unified model provider",
+			},
+		}
+	}
+}

apps/vscode/src/core/api/providers/anthropic.ts

@@ -0,0 +1,316 @@
+import { Anthropic } from "@anthropic-ai/sdk"
+import type {
+	MessageCreateParamsStreaming as BetaMessageCreateParamsStreaming,
+	BetaRawMessageStreamEvent,
+} from "@anthropic-ai/sdk/resources/beta/messages/messages"
+import { Tool as AnthropicTool } from "@anthropic-ai/sdk/resources/index"
+import type { MessageCreateParamsStreaming as AnthropicMessageCreateParamsStreaming } from "@anthropic-ai/sdk/resources/messages/messages"
+import { Stream as AnthropicStream } from "@anthropic-ai/sdk/streaming"
+import {
+	ANTHROPIC_FAST_MODE_SUFFIX,
+	AnthropicModelId,
+	anthropicDefaultModelId,
+	anthropicModels,
+	CLAUDE_SONNET_1M_SUFFIX,
+	ModelInfo,
+} from "@shared/api"
+import { isClaudeOpusAdaptiveThinkingModel, resolveClaudeOpusAdaptiveThinking } from "@shared/utils/reasoning-support"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { fetch } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../index"
+import { withRetry } from "../retry"
+import { sanitizeAnthropicMessages } from "../transform/anthropic-format"
+import { ApiStream } from "../transform/stream"
+
+export const ANTHROPIC_FAST_MODE_BETA = "fast-mode-2026-02-01"
+
+interface AnthropicHandlerOptions extends CommonApiHandlerOptions {
+	apiKey?: string
+	anthropicBaseUrl?: string
+	apiModelId?: string
+	reasoningEffort?: string
+	thinkingBudgetTokens?: number
+}
+
+export class AnthropicHandler implements ApiHandler {
+	private options: AnthropicHandlerOptions
+	private client: Anthropic | undefined
+
+	constructor(options: AnthropicHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): Anthropic {
+		if (!this.client) {
+			if (!this.options.apiKey) {
+				throw new Error("Anthropic API key is required")
+			}
+			try {
+				this.client = new Anthropic({
+					apiKey: this.options.apiKey,
+					baseURL: this.options.anthropicBaseUrl || undefined,
+					defaultHeaders: buildExternalBasicHeaders(),
+					fetch, // Use configured fetch with proxy support
+				})
+			} catch (error) {
+				throw new Error(`Error creating Anthropic client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: AnthropicTool[]): ApiStream {
+		const client = this.ensureClient()
+
+		const model = this.getModel()
+		let stream: AnthropicStream<Anthropic.RawMessageStreamEvent> | AsyncIterable<BetaRawMessageStreamEvent>
+
+		const useFastMode = model.id.endsWith(ANTHROPIC_FAST_MODE_SUFFIX)
+		const baseModelId = useFastMode ? model.id.slice(0, -ANTHROPIC_FAST_MODE_SUFFIX.length) : model.id
+		const modelId = baseModelId.endsWith(CLAUDE_SONNET_1M_SUFFIX)
+			? baseModelId.slice(0, -CLAUDE_SONNET_1M_SUFFIX.length)
+			: baseModelId
+		const enable1mContextWindow = baseModelId.endsWith(CLAUDE_SONNET_1M_SUFFIX)
+		const fastModeBetas = enable1mContextWindow
+			? [ANTHROPIC_FAST_MODE_BETA, "context-1m-2025-08-07"]
+			: [ANTHROPIC_FAST_MODE_BETA]
+		const createFastModeMessage = (
+			body: AnthropicMessageCreateParamsStreaming,
+		): Promise<AsyncIterable<BetaRawMessageStreamEvent>> => {
+			return (
+				client.beta.messages.create as unknown as (
+					params: BetaMessageCreateParamsStreaming & { speed: "fast" },
+				) => Promise<AsyncIterable<BetaRawMessageStreamEvent>>
+			)({
+				...body,
+				betas: fastModeBetas,
+				speed: "fast",
+			})
+		}
+
+		const budget_tokens = this.options.thinkingBudgetTokens || 0
+
+		// Tools are available only when native tools are enabled.
+		const nativeToolsOn = tools?.length && tools?.length > 0
+		const reasoningOn = (model.info.supportsReasoning ?? false) && budget_tokens !== 0
+
+		// Claude Opus 4.5+ uses adaptive thinking instead of budgeted extended thinking.
+		const isAdaptiveThinkingModel = isClaudeOpusAdaptiveThinkingModel(modelId)
+		const adaptiveThinking = isAdaptiveThinkingModel
+			? resolveClaudeOpusAdaptiveThinking(this.options.reasoningEffort, budget_tokens)
+			: undefined
+		const adaptiveThinkingEnabled = adaptiveThinking?.enabled === true
+		const adaptiveThinkingEffort = adaptiveThinking?.effort
+		const thinkingEnabled = isAdaptiveThinkingModel ? adaptiveThinkingEnabled : reasoningOn
+		const thinkingConfig = thinkingEnabled
+			? isAdaptiveThinkingModel
+				? ({ type: "adaptive" } as any)
+				: { type: "enabled", budget_tokens: budget_tokens }
+			: undefined
+		const outputConfig = isAdaptiveThinkingModel && adaptiveThinkingEffort ? { effort: adaptiveThinkingEffort } : undefined
+
+		if (model.info.supportsPromptCache) {
+			const anthropicMessages = sanitizeAnthropicMessages(messages, true)
+			const requestBody: AnthropicMessageCreateParamsStreaming & Record<string, unknown> = {
+				model: modelId,
+				thinking: thinkingConfig,
+				max_tokens: model.info.maxTokens || 8192,
+				// "Thinking isn't compatible with temperature, top_p, or top_k modifications as well as forced tool use."
+				// (https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking#important-considerations-when-using-extended-thinking)
+				// Adaptive Claude Opus models do not support temperature.
+				temperature: isAdaptiveThinkingModel ? undefined : reasoningOn ? undefined : 0,
+				system: [
+					{
+						text: systemPrompt,
+						type: "text",
+						cache_control: { type: "ephemeral" },
+					},
+				], // setting cache breakpoint for system prompt so new tasks can reuse it
+				messages: anthropicMessages,
+				// tools, // cache breakpoints go from tools > system > messages, and since tools dont change, we can just set the breakpoint at the end of system (this avoids having to set a breakpoint at the end of tools which by itself does not meet min requirements for haiku caching)
+				stream: true,
+				tools: nativeToolsOn ? tools : undefined,
+				// tool_choice options:
+				// - none: disables tool use, even if tools are provided. Claude will not call any tools.
+				// - auto: allows Claude to decide whether to call any provided tools or not. This is the default value when tools are provided.
+				// - any: tells Claude that it must use one of the provided tools, but doesn’t force a particular tool.
+				// NOTE: Forcing tool use when tools are provided will result in error when thinking is also enabled.
+				tool_choice: nativeToolsOn && !thinkingEnabled ? { type: "any" } : undefined,
+			}
+			if (outputConfig) {
+				requestBody.output_config = outputConfig
+			}
+
+			stream = useFastMode
+				? await createFastModeMessage(requestBody)
+				: await client.messages.create(
+						requestBody,
+						(() => {
+							// 1m context window beta header
+							if (enable1mContextWindow) {
+								return {
+									headers: {
+										"anthropic-beta": "context-1m-2025-08-07",
+									},
+								}
+							}
+							return undefined
+						})(),
+					)
+		} else {
+			const requestBody: AnthropicMessageCreateParamsStreaming & Record<string, unknown> = {
+				model: modelId,
+				max_tokens: model.info.maxTokens || 8192,
+				temperature: isAdaptiveThinkingModel ? undefined : reasoningOn ? undefined : 0,
+				system: [{ text: systemPrompt, type: "text" }],
+				messages: sanitizeAnthropicMessages(messages, false),
+				tools: nativeToolsOn ? tools : undefined,
+				tool_choice: thinkingEnabled ? undefined : { type: "auto" },
+				stream: true,
+				thinking: thinkingConfig,
+			}
+			if (outputConfig) {
+				requestBody.output_config = outputConfig
+			}
+
+			stream = useFastMode ? await createFastModeMessage(requestBody) : await client.messages.create(requestBody)
+		}
+
+		const lastStartedToolCall = { id: "", name: "", arguments: "" }
+
+		for await (const chunk of stream) {
+			switch (chunk?.type) {
+				case "message_start":
+					{
+						// tells us cache reads/writes/input/output
+						const usage = chunk.message.usage
+						yield {
+							type: "usage",
+							inputTokens: usage.input_tokens || 0,
+							outputTokens: usage.output_tokens || 0,
+							cacheWriteTokens: usage.cache_creation_input_tokens || undefined,
+							cacheReadTokens: usage.cache_read_input_tokens || undefined,
+						}
+					}
+					break
+				case "message_delta":
+					// tells us stop_reason, stop_sequence, and output tokens along the way and at the end of the message
+
+					yield {
+						type: "usage",
+						inputTokens: 0,
+						outputTokens: chunk.usage.output_tokens || 0,
+					}
+					break
+				case "message_stop":
+					// no usage data, just an indicator that the message is done
+					break
+				case "content_block_start":
+					switch (chunk.content_block.type) {
+						case "thinking":
+							yield {
+								type: "reasoning",
+								reasoning: chunk.content_block.thinking || "",
+								signature: chunk.content_block.signature,
+							}
+							break
+						case "redacted_thinking":
+							// Content is encrypted, and we don't to pass placeholder text back to the API
+							yield {
+								type: "reasoning",
+								reasoning: "[Redacted thinking block]",
+								redacted_data: chunk.content_block.data,
+							}
+							break
+						case "tool_use":
+							if (chunk.content_block.id && chunk.content_block.name) {
+								// Convert Anthropic tool_use to OpenAI-compatible format
+								lastStartedToolCall.id = chunk.content_block.id
+								lastStartedToolCall.name = chunk.content_block.name
+								lastStartedToolCall.arguments = ""
+							}
+							break
+						case "text":
+							// we may receive multiple text blocks, in which case just insert a line break between them
+							if (chunk.index > 0) {
+								yield {
+									type: "text",
+									text: "\n",
+								}
+							}
+							yield {
+								type: "text",
+								text: chunk.content_block.text,
+							}
+							break
+					}
+					break
+				case "content_block_delta":
+					switch (chunk.delta.type) {
+						case "thinking_delta":
+							// 'reasoning' type just displays in the UI, but ant_thinking will be used to send the thinking traces back to the API
+							yield {
+								type: "reasoning",
+								reasoning: chunk.delta.thinking,
+							}
+							break
+						case "signature_delta":
+							// It's used when sending the thinking block back to the API
+							// API expects this in completed form, not as array of deltas
+							if (chunk.delta.signature) {
+								yield {
+									type: "reasoning",
+									reasoning: "", // reasoning text is already sent via thinking_delta
+									signature: chunk.delta.signature,
+								}
+							}
+							break
+						case "text_delta":
+							yield {
+								type: "text",
+								text: chunk.delta.text,
+							}
+							break
+						case "input_json_delta":
+							if (lastStartedToolCall.id && lastStartedToolCall.name && chunk.delta.partial_json) {
+								// 	// Convert Anthropic tool_use to OpenAI-compatible format
+								yield {
+									type: "tool_calls",
+									tool_call: {
+										...lastStartedToolCall,
+										function: {
+											...lastStartedToolCall,
+											id: lastStartedToolCall.id,
+											name: lastStartedToolCall.name,
+											arguments: chunk.delta.partial_json,
+										},
+									},
+								}
+							}
+							break
+					}
+					break
+
+				case "content_block_stop":
+					lastStartedToolCall.id = ""
+					lastStartedToolCall.name = ""
+					lastStartedToolCall.arguments = ""
+					break
+			}
+		}
+	}
+
+	getModel(): { id: AnthropicModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+		if (modelId && modelId in anthropicModels) {
+			const id = modelId as AnthropicModelId
+			return { id, info: anthropicModels[id] }
+		}
+		return {
+			id: anthropicDefaultModelId,
+			info: anthropicModels[anthropicDefaultModelId],
+		}
+	}
+}

apps/vscode/src/core/api/providers/asksage.ts

@@ -0,0 +1,200 @@
+import { AskSageModelId, askSageDefaultModelId, askSageDefaultURL, askSageModels, ModelInfo } from "@shared/api"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { fetch } from "@/shared/net"
+import { Logger } from "@/shared/services/Logger"
+import { ApiHandler, CommonApiHandlerOptions } from ".."
+import { withRetry } from "../retry"
+import { ApiStream } from "../transform/stream"
+
+interface AskSageHandlerOptions extends CommonApiHandlerOptions {
+	asksageApiKey?: string
+	asksageApiUrl?: string
+	apiModelId?: string
+}
+
+type AskSageRequest = {
+	system_prompt: string
+	message: {
+		user: "gpt" | "me"
+		message: string
+	}[]
+	model: string
+	dataset: "none"
+	usage: boolean
+}
+
+type AskSageUsage = {
+	model_tokens: {
+		completion_tokens: number
+		prompt_tokens: number
+		total_tokens: number
+	}
+	asksage_tokens: number
+}
+
+type AskSageResponse = {
+	uuid: string
+	status: number
+	// Response status
+	response: string
+	// Generated response message
+	message: string
+	// whether embedding & vector systems are down
+	embedding_down: boolean
+	vectors_down: boolean
+	// references if dataset is not none
+	references: string
+	type: string
+	added_obj: any
+	tool_calls: any
+	// usage metrics
+	usage: AskSageUsage | null
+	tool_responses: any[]
+	tool_calls_unified: any[]
+}
+
+export class AskSageHandler implements ApiHandler {
+	private options: AskSageHandlerOptions
+	private apiUrl: string
+	private apiKey: string
+
+	constructor(options: AskSageHandlerOptions) {
+		Logger.log("init api url", options.asksageApiUrl, askSageDefaultURL)
+		this.options = options
+		this.apiKey = options.asksageApiKey || ""
+		this.apiUrl = options.asksageApiUrl || askSageDefaultURL
+
+		if (!this.apiKey) {
+			throw new Error("AskSage API key is required")
+		}
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[]): ApiStream {
+		try {
+			const model = this.getModel()
+			// Transform messages into AskSageRequest format
+			const formattedMessages = messages.map((msg) => {
+				const content = Array.isArray(msg.content)
+					? msg.content.map((block) => ("text" in block ? block.text : "")).join("")
+					: msg.content
+
+				return {
+					user: msg.role === "assistant" ? ("gpt" as const) : ("me" as const),
+					message: content,
+				}
+			})
+
+			const request: AskSageRequest = {
+				system_prompt: systemPrompt,
+				message: formattedMessages,
+				model: model.id,
+				dataset: "none",
+				usage: true,
+			}
+
+			// Make request to AskSage API
+			const response = await fetch(`${this.apiUrl}/query`, {
+				method: "POST",
+				headers: this.headers(),
+				body: JSON.stringify(request),
+			})
+
+			if (!response.ok) {
+				const error = await response.text()
+				throw new Error(`AskSage API error: ${error}`)
+			}
+
+			const result = (await response.json()) as AskSageResponse
+
+			if (!result.message) {
+				throw new Error("No content in AskSage response")
+			}
+
+			// Yield tool responses if they exist
+			if (result.tool_responses && result.tool_responses.length > 0) {
+				for (const toolResponse of result.tool_responses) {
+					yield {
+						type: "text",
+						text: `[Tool Response: ${JSON.stringify(toolResponse)}]\n`,
+					}
+				}
+			}
+
+			// Yield the main response text
+			yield {
+				type: "text",
+				text: result.message,
+			}
+
+			// Yield usage information if available
+			if (result.usage) {
+				yield {
+					type: "usage",
+					inputTokens: result.usage.model_tokens.prompt_tokens,
+					outputTokens: result.usage.model_tokens.completion_tokens,
+					cacheReadTokens: 0,
+					cacheWriteTokens: 0,
+					totalCost: result.usage.asksage_tokens, // Cost = Consumed AskSage tokens
+				}
+			}
+		} catch (error) {
+			if (error instanceof Error) {
+				throw new Error(`AskSage request failed: ${error.message}`)
+			}
+			throw error
+		}
+	}
+
+	async getApiStreamUsage() {
+		if (!this.apiKey) {
+			return undefined
+		}
+
+		try {
+			const response = await fetch(`${this.apiUrl}/count-monthly-tokens`, {
+				method: "POST",
+				headers: this.headers(),
+				body: JSON.stringify({ app_name: "asksage" }),
+			})
+
+			if (!response.ok) {
+				Logger.error("Failed to fetch AskSage usage", await response.text())
+				return undefined
+			}
+
+			const data = await response.json()
+			const usedTokens = data.response as number
+
+			return {
+				type: "usage" as const,
+				inputTokens: usedTokens,
+				outputTokens: 0,
+			}
+		} catch (error) {
+			Logger.error("Error fetching AskSage usage:", error)
+			return undefined
+		}
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+		if (modelId && modelId in askSageModels) {
+			const id = modelId as AskSageModelId
+			return { id, info: askSageModels[id] }
+		}
+		return {
+			id: askSageDefaultModelId,
+			info: askSageModels[askSageDefaultModelId],
+		}
+	}
+
+	private headers() {
+		return {
+			"Content-Type": "application/json",
+			"x-access-tokens": this.apiKey,
+			...buildExternalBasicHeaders(),
+		}
+	}
+}

apps/vscode/src/core/api/providers/baseten.ts

@@ -0,0 +1,172 @@
+import { BasetenModelId, basetenDefaultModelId, basetenModels, ModelInfo } from "@shared/api"
+import { calculateApiCostOpenAI } from "@utils/cost"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { fetch } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+import { ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface BasetenHandlerOptions extends CommonApiHandlerOptions {
+	basetenApiKey?: string
+	basetenModelId?: string
+	basetenModelInfo?: ModelInfo
+	apiModelId?: string // For backward compatibility
+}
+
+export class BasetenHandler implements ApiHandler {
+	private options: BasetenHandlerOptions
+	private client: OpenAI | undefined
+
+	constructor(options: BasetenHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.basetenApiKey) {
+				throw new Error("Baseten API key is required")
+			}
+			try {
+				this.client = new OpenAI({
+					baseURL: "https://inference.baseten.co/v1",
+					apiKey: this.options.basetenApiKey,
+					defaultHeaders: buildExternalBasicHeaders(),
+					fetch, // Use configured fetch with proxy support
+				})
+			} catch (error) {
+				throw new Error(`Error creating Baseten client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	/**
+	 * Gets the optimal max_tokens based on model capabilities
+	 */
+	private getOptimalMaxTokens(model: { id: BasetenModelId; info: ModelInfo }): number {
+		// Use model-specific max tokens if available
+		if (model.info.maxTokens && model.info.maxTokens > 0) {
+			return model.info.maxTokens
+		}
+
+		// Default fallback
+		return 8192
+	}
+
+	getModel(): { id: BasetenModelId; info: ModelInfo } {
+		// First priority: basetenModelId and basetenModelInfo
+		const basetenModelId = this.options.basetenModelId
+		const basetenModelInfo = this.options.basetenModelInfo
+		if (basetenModelId && basetenModelInfo) {
+			return { id: basetenModelId as BasetenModelId, info: basetenModelInfo }
+		}
+
+		// Second priority: basetenModelId with static model info
+		if (basetenModelId && basetenModelId in basetenModels) {
+			const id = basetenModelId as BasetenModelId
+			return { id, info: basetenModels[id] }
+		}
+
+		// Third priority: apiModelId (for backward compatibility)
+		const apiModelId = this.options.apiModelId
+		if (apiModelId && apiModelId in basetenModels) {
+			const id = apiModelId as BasetenModelId
+			return { id, info: basetenModels[id] }
+		}
+
+		// Default fallback
+		return {
+			id: basetenDefaultModelId,
+			info: basetenModels[basetenDefaultModelId],
+		}
+	}
+
+	private async *yieldUsage(modelInfo: ModelInfo, usage: any): ApiStream {
+		if (usage.prompt_tokens || usage.completion_tokens) {
+			const cost = calculateApiCostOpenAI(modelInfo, usage.prompt_tokens || 0, usage.completion_tokens || 0)
+
+			yield {
+				type: "usage",
+				inputTokens: usage.prompt_tokens || 0,
+				outputTokens: usage.completion_tokens || 0,
+				cacheWriteTokens: 0,
+				cacheReadTokens: 0,
+				totalCost: cost,
+			}
+		}
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+		const maxTokens = this.getOptimalMaxTokens(model)
+		const toolCallProcessor = new ToolCallProcessor()
+
+		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+
+		const stream = await client.chat.completions.create({
+			model: model.id,
+			max_tokens: maxTokens,
+			messages: openAiMessages,
+			stream: true,
+			stream_options: { include_usage: true },
+			temperature: 0,
+			tools,
+			tool_choice: tools && tools.length > 0 ? "auto" : undefined,
+		})
+
+		let didOutputUsage = false
+
+		for await (const chunk of stream) {
+			const delta = chunk?.choices?.[0]?.delta
+
+			// Handle reasoning field if present (for reasoning models with parsed output)
+			if (delta && "reasoning" in delta && delta?.reasoning) {
+				const reasoning = typeof delta.reasoning === "string" ? delta.reasoning : JSON.stringify(delta.reasoning)
+				yield {
+					type: "reasoning",
+					reasoning,
+				}
+			}
+
+			// Handle content field
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			// Handle usage information - only output once
+			if (!didOutputUsage && chunk.usage) {
+				yield* this.yieldUsage(model.info, chunk.usage)
+				didOutputUsage = true
+			}
+		}
+	}
+
+	/**
+	 * Checks if the current model supports tools
+	 */
+	supportsTools(): boolean {
+		const model = this.getModel()
+		const modelInfo = model.info as any
+
+		// Use dynamic API data when available, fallback to true since all current Baseten models support tools
+		// (as of 2025-09-16 - could change if Baseten add non-tool models in future, currently no plans to do so)
+		return modelInfo.supportedFeatures ? modelInfo.supportedFeatures.includes("tools") : true
+	}
+}

apps/vscode/src/core/api/providers/cerebras.ts

@@ -0,0 +1,275 @@
+import Cerebras from "@cerebras/cerebras_cloud_sdk"
+import { CerebrasModelId, cerebrasDefaultModelId, cerebrasModels, ModelInfo } from "@shared/api"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { fetch } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../index"
+import { withRetry } from "../retry"
+import { ApiStream } from "../transform/stream"
+
+interface CerebrasHandlerOptions extends CommonApiHandlerOptions {
+	cerebrasApiKey?: string
+	apiModelId?: string
+}
+
+// Conservative max_tokens for Cerebras to avoid premature rate limiting.
+// Cerebras rate limiter estimates token consumption using max_completion_tokens upfront,
+// so requesting the model maximum (e.g., 64K) reserves that quota even if actual usage is low.
+// 16K is sufficient for most agentic tool use while preserving rate limit headroom.
+const CEREBRAS_DEFAULT_MAX_TOKENS = 16_384
+
+export class CerebrasHandler implements ApiHandler {
+	private options: CerebrasHandlerOptions
+	private client: Cerebras | undefined
+
+	constructor(options: CerebrasHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): Cerebras {
+		if (!this.client) {
+			// Clean and validate the API key
+			const cleanApiKey = this.options.cerebrasApiKey?.trim()
+
+			if (!cleanApiKey) {
+				throw new Error("Cerebras API key is required")
+			}
+
+			try {
+				const externalHeaders = buildExternalBasicHeaders()
+				this.client = new Cerebras({
+					apiKey: cleanApiKey,
+					timeout: 30000, // 30 second timeout
+					fetch, // Use configured fetch with proxy support
+					defaultHeaders: {
+						...externalHeaders,
+						"X-Cerebras-3rd-Party-Integration": "cline",
+					},
+				})
+			} catch (error) {
+				throw new Error(`Error creating Cerebras client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry({
+		maxRetries: 6, // More retries to be patient with rate limits
+		baseDelay: 5000, // Start with 5 second delay
+		maxDelay: 60000, // Allow up to 60 second delays to respect rate limits
+	})
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[]): ApiStream {
+		const client = this.ensureClient()
+
+		// Convert Anthropic messages to Cerebras format
+		const cerebrasMessages: Array<{
+			role: "system" | "user" | "assistant"
+			content: string
+		}> = [{ role: "system", content: systemPrompt }]
+
+		// Helper function to strip thinking tags from content
+		const stripThinkingTags = (content: string): string => {
+			return content.replace(/<think>[\s\S]*?<\/think>/g, "").trim()
+		}
+
+		// Check if this is a reasoning model that uses thinking tags
+		const modelId = this.getModel().id
+		const isReasoningModel = modelId.includes("qwen")
+
+		// Convert Anthropic messages to Cerebras format
+		for (const message of messages) {
+			if (message.role === "user") {
+				const content = Array.isArray(message.content)
+					? message.content
+							.map((block) => {
+								if (block.type === "text") {
+									return block.text
+								}
+								if (block.type === "image") {
+									return "[Image content not supported in Cerebras]"
+								}
+								return ""
+							})
+							.join("\n")
+					: message.content
+				cerebrasMessages.push({ role: "user", content })
+			} else if (message.role === "assistant") {
+				let content = Array.isArray(message.content)
+					? message.content
+							.map((block) => {
+								if (block.type === "text") {
+									return block.text
+								}
+								return ""
+							})
+							.join("\n")
+					: message.content || ""
+
+				// Strip thinking tags from assistant messages for reasoning models
+				// so the model doesn't see its own thinking in the conversation history
+				if (isReasoningModel) {
+					content = stripThinkingTags(content)
+				}
+
+				cerebrasMessages.push({ role: "assistant", content })
+			}
+		}
+
+		try {
+			const model = this.getModel()
+			const stream = await client.chat.completions.create({
+				model: model.id,
+				messages: cerebrasMessages,
+				temperature: model.info.temperature ?? 0,
+				stream: true,
+				max_tokens: CEREBRAS_DEFAULT_MAX_TOKENS,
+			})
+
+			// Handle streaming response
+			let reasoning: string | null = null // Track reasoning content for models that support thinking
+
+			for await (const chunk of stream as any) {
+				// Type assertion for the streaming chunk
+				const streamChunk = chunk as any
+
+				if (streamChunk.choices?.[0]?.delta?.content) {
+					const content = streamChunk.choices[0].delta.content
+
+					// Handle reasoning models (Qwen and DeepSeek R1 Distill) that use <think> tags
+					if (isReasoningModel) {
+						// Check if we're entering or continuing reasoning mode
+						if (reasoning || content.includes("<think>")) {
+							reasoning = (reasoning || "") + content
+
+							// Clean the content by removing think tags for display
+							const cleanContent = content.replace(/<think>/g, "").replace(/<\/think>/g, "")
+
+							// Only yield reasoning content if there's actual content after cleaning
+							if (cleanContent.trim()) {
+								yield {
+									type: "reasoning",
+									reasoning: cleanContent,
+								}
+							}
+
+							// Check if reasoning is complete
+							if (reasoning.includes("</think>")) {
+								reasoning = null
+							}
+						} else {
+							// Regular content outside of thinking tags
+							yield {
+								type: "text",
+								text: content,
+							}
+						}
+					} else {
+						// Non-reasoning models - just yield text content
+						yield {
+							type: "text",
+							text: content,
+						}
+					}
+				}
+
+				// Handle usage information from Cerebras API
+				// Usage is typically only available in the final chunk
+				if (streamChunk.usage) {
+					const totalCost = this.calculateCost({
+						inputTokens: streamChunk.usage.prompt_tokens || 0,
+						outputTokens: streamChunk.usage.completion_tokens || 0,
+					})
+
+					yield {
+						type: "usage",
+						inputTokens: streamChunk.usage.prompt_tokens || 0,
+						outputTokens: streamChunk.usage.completion_tokens || 0,
+						cacheReadTokens: 0,
+						cacheWriteTokens: 0,
+						totalCost,
+					}
+				}
+			}
+		} catch (error: any) {
+			// Enhanced error handling for Cerebras API
+			if (error?.status === 429 || error?.code === "rate_limit_exceeded") {
+				// Rate limit error - will be handled by retry decorator with patient backoff
+				const _limits = this.getRateLimits()
+				throw new Error(`Cerebras API rate limit exceeded.`)
+			}
+			if (error?.status === 401) {
+				throw new Error("Cerebras API authentication failed. Please check your API key.")
+			}
+			if (error?.status === 403) {
+				throw new Error("Cerebras API access denied. Please check your API key permissions.")
+			}
+			if (error?.status >= 500) {
+				// Server errors - retryable
+				throw new Error(`Cerebras API server error (${error.status}): ${error.message || "Unknown server error"}`)
+			}
+			if (error?.status === 400) {
+				// Client errors - not retryable
+				throw new Error(`Cerebras API bad request: ${error.message || "Invalid request parameters"}`)
+			}
+
+			// Re-throw original error for other cases
+			throw error
+		}
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		const originalModelId = this.options.apiModelId
+		let apiModelId = originalModelId
+		if (originalModelId === "qwen-3-coder-480b-free") {
+			apiModelId = "qwen-3-coder-480b"
+			return { id: apiModelId, info: cerebrasModels[originalModelId as CerebrasModelId] }
+		}
+
+		if (originalModelId && originalModelId in cerebrasModels) {
+			const id = originalModelId as CerebrasModelId
+			return { id, info: cerebrasModels[id] }
+		}
+		return {
+			id: cerebrasDefaultModelId,
+			info: cerebrasModels[cerebrasDefaultModelId],
+		}
+	}
+
+	/**
+	 * Get rate limit information for the current model
+	 *
+	 * These limits are used for informational purposes and to calculate appropriate
+	 * retry delays. Since Cerebras inference is extremely fast, users hit these limits
+	 * quickly, so we need to be patient with retries to maximize usage efficiency.
+	 *
+	 * @returns Rate limit configuration for the model
+	 */
+	private getRateLimits(): { requestsPerMinute: number; tokensPerMinute: number } {
+		const modelId = this.getModel().id
+
+		switch (modelId) {
+			case "qwen-3-coder-480b":
+			case "qwen-3-coder-480b-free":
+				return { requestsPerMinute: 10, tokensPerMinute: 150_000 }
+			case "qwen-3-235b-a22b-instruct-2507":
+			case "qwen-3-235b-a22b-thinking-2507":
+				return { requestsPerMinute: 30, tokensPerMinute: 60_000 }
+			case "gpt-oss-120b":
+				return { requestsPerMinute: 30, tokensPerMinute: 64_000 }
+			default:
+				// Default rate limits for unknown models
+				return { requestsPerMinute: 30, tokensPerMinute: 60_000 }
+		}
+	}
+
+	private calculateCost({ inputTokens, outputTokens }: { inputTokens: number; outputTokens: number }): number {
+		const model = this.getModel()
+		const inputPrice = model.info.inputPrice || 0
+		const outputPrice = model.info.outputPrice || 0
+
+		const inputCost = (inputPrice / 1_000_000) * inputTokens
+		const outputCost = (outputPrice / 1_000_000) * outputTokens
+
+		return inputCost + outputCost
+	}
+}

apps/vscode/src/core/api/providers/claude-code.ts

@@ -0,0 +1,229 @@
+import { filterMessagesForClaudeCode } from "@/integrations/claude-code/message-filter"
+import { runClaudeCode } from "@/integrations/claude-code/run"
+import { ClaudeCodeModelId, claudeCodeDefaultModelId, claudeCodeModels } from "@/shared/api"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { Logger } from "@/shared/services/Logger"
+import { type ApiHandler, CommonApiHandlerOptions } from ".."
+import { withRetry } from "../retry"
+import { type ApiStream, ApiStreamUsageChunk } from "../transform/stream"
+
+interface ClaudeCodeHandlerOptions extends CommonApiHandlerOptions {
+	claudeCodePath?: string
+	apiModelId?: string
+	thinkingBudgetTokens?: number
+}
+
+export class ClaudeCodeHandler implements ApiHandler {
+	private options: ClaudeCodeHandlerOptions
+
+	constructor(options: ClaudeCodeHandlerOptions) {
+		this.options = options
+	}
+
+	@withRetry({
+		maxRetries: 4,
+		baseDelay: 2000,
+		maxDelay: 15000,
+	})
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[]): ApiStream {
+		// Filter out image blocks since Claude Code doesn't support them
+		const filteredMessages = filterMessagesForClaudeCode(messages)
+
+		const claudeProcess = runClaudeCode({
+			systemPrompt,
+			messages: filteredMessages,
+			path: this.options.claudeCodePath,
+			modelId: this.getModel().id,
+			thinkingBudgetTokens: this.options.thinkingBudgetTokens,
+		})
+
+		// Usage is included with assistant messages,
+		// but cost is included in the result chunk
+		const usage: ApiStreamUsageChunk = {
+			type: "usage",
+			inputTokens: 0,
+			outputTokens: 0,
+			cacheReadTokens: 0,
+			cacheWriteTokens: 0,
+		}
+
+		let isPaidUsage = true
+
+		for await (const chunk of claudeProcess) {
+			if (typeof chunk === "string") {
+				yield {
+					type: "text",
+					text: chunk,
+				}
+
+				continue
+			}
+
+			// Handle system init messages
+			if (chunk.type === "system" && "subtype" in chunk) {
+				if (chunk.subtype === "init") {
+					// Based on my tests, subscription usage sets the `apiKeySource` to "none"
+					isPaidUsage = (chunk as any).apiKeySource !== "none"
+				}
+				// Also handles legacy rate_limit_event format (type: "system", subtype: "rate_limit_event")
+				// by falling through — no special handling needed.
+				continue
+			}
+
+			// Handle rate_limit_event (newer CLI format: top-level type)
+			if (chunk.type === "rate_limit_event") {
+				// Rate limit events are informational. Log them but don't yield anything.
+				// If the rate limit blocks the response, the stream will end without
+				// assistant messages and the task loop will handle the empty response.
+				Logger.log("Claude Code rate limit event:", JSON.stringify(chunk))
+				continue
+			}
+
+			// Skip user messages (tool results from Claude Code's own tool execution)
+			if (chunk.type === "user") {
+				continue
+			}
+
+			if (chunk.type === "assistant" && "message" in chunk) {
+				const message = chunk.message
+
+				// Check for error field on the message (newer CLI format)
+				if (message.error) {
+					const firstContent = message.content?.[0]
+					const errorText = firstContent && "text" in firstContent ? firstContent.text : undefined
+					throw new Error(errorText ?? `Claude Code error: ${message.error}`)
+				}
+
+				if (message.stop_reason !== null) {
+					const firstContent = message.content?.[0]
+					const content = firstContent && "text" in firstContent ? firstContent : undefined
+
+					// Check if content exists before accessing its properties
+					if (content && content.text.startsWith(`API Error`)) {
+						// Error messages are formatted as: `API Error: <<status code>> <<json>>`
+						const errorMessageStart = content.text.indexOf("{")
+						const errorMessage = content.text.slice(errorMessageStart)
+
+						const error = this.attemptParse(errorMessage)
+						if (!error) {
+							throw new Error(content.text)
+						}
+
+						if (error.error.message.includes("Invalid model name")) {
+							throw new Error(
+								content.text +
+									`\n\nAPI keys and subscription plans allow different models. Make sure the selected model is included in your plan.`,
+							)
+						}
+
+						throw new Error(errorMessage)
+					}
+				}
+
+				for (const content of message.content) {
+					switch (content.type) {
+						case "text":
+							yield {
+								type: "text",
+								text: content.text,
+							}
+							break
+						case "thinking":
+							yield {
+								type: "reasoning",
+								reasoning: content.thinking || "",
+							}
+							break
+						case "redacted_thinking":
+							yield {
+								type: "reasoning",
+								reasoning: "[Redacted thinking block]",
+							}
+							break
+						case "tool_use":
+							// Yield tool_use blocks to the streaming pipeline for proper tool execution
+							yield {
+								type: "tool_calls",
+								tool_call: {
+									call_id: content.id,
+									function: {
+										id: content.id,
+										name: content.name,
+										arguments: JSON.stringify(content.input),
+									},
+								},
+							}
+							break
+						default: {
+							// Handle unknown content block types gracefully.
+							// Newer Anthropic models or CLI versions may introduce new content types
+							// (e.g., server_tool_use, mcp_tool_use). Log them instead of silently dropping.
+							const unknownBlock = content as { type: string; text?: string }
+							Logger.warn(`Unhandled content type in Claude Code response: ${unknownBlock.type}`)
+							// If the unknown block has a text-like field, try to yield it as text
+							if (typeof unknownBlock.text === "string") {
+								yield {
+									type: "text",
+									text: unknownBlock.text,
+								}
+							}
+							break
+						}
+					}
+				}
+
+				// According to Anthropic's API documentation:
+				// https://docs.anthropic.com/en/api/messages#usage-object
+				// The `input_tokens` field already includes both `cache_read_input_tokens` and `cache_creation_input_tokens`.
+				// Therefore, we should not add cache tokens to the input_tokens count again, as this would result in double-counting.
+				usage.inputTokens = message.usage?.input_tokens ?? 0
+				usage.outputTokens = message.usage?.output_tokens ?? 0
+				usage.cacheReadTokens = message.usage?.cache_read_input_tokens ?? 0
+				usage.cacheWriteTokens = message.usage?.cache_creation_input_tokens ?? 0
+
+				continue
+			}
+
+			if (chunk.type === "result" && "result" in chunk) {
+				if (chunk.is_error) {
+					throw new Error(`Claude Code returned an error: ${chunk.result}`)
+				}
+
+				usage.totalCost = isPaidUsage ? chunk.total_cost_usd : 0
+
+				yield usage
+				continue
+			}
+
+			// ErrorMessage — log it explicitly and skip
+			if ((chunk as any).type === "error") {
+				Logger.warn("Claude Code emitted an error-type chunk:", JSON.stringify(chunk))
+				continue
+			}
+
+			// Any completely unrecognized chunk type — log and skip
+			Logger.warn(`Unrecognized Claude Code chunk type: ${(chunk as any).type}`)
+		}
+	}
+
+	private attemptParse(str: string) {
+		try {
+			return JSON.parse(str)
+		} catch (_err) {
+			return null
+		}
+	}
+
+	getModel() {
+		const modelId = this.options.apiModelId
+		if (modelId && modelId in claudeCodeModels) {
+			const id = modelId as ClaudeCodeModelId
+			return { id, info: claudeCodeModels[id] }
+		}
+
+		return {
+			id: claudeCodeDefaultModelId,
+			info: claudeCodeModels[claudeCodeDefaultModelId],
+		}
+	}
+}

apps/vscode/src/core/api/providers/cline.ts

@@ -0,0 +1,343 @@
+import { type ModelInfo, openRouterDefaultModelId, openRouterDefaultModelInfo } from "@shared/api"
+import { shouldSkipReasoningForModel } from "@utils/model-utils"
+import axios from "axios"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { ClineEnv } from "@/config"
+import { refreshClineRecommendedModels } from "@/core/controller/models/refreshClineRecommendedModels"
+import { ClineAccountService } from "@/services/account/ClineAccountService"
+import { AuthService } from "@/services/auth/AuthService"
+import { buildClineExtraHeaders } from "@/services/EnvUtils"
+import { CLINE_ACCOUNT_AUTH_ERROR_MESSAGE } from "@/shared/ClineAccount"
+import { CLINE_RECOMMENDED_MODELS_FALLBACK } from "@/shared/cline/recommended-models"
+import type { ClineStorageMessage } from "@/shared/messages/content"
+import { fetch, getAxiosSettings } from "@/shared/net"
+import { Logger } from "@/shared/services/Logger"
+import type { ApiHandler, CommonApiHandlerOptions } from "../"
+import { withRetry } from "../retry"
+import { createOpenRouterStream } from "../transform/openrouter-stream"
+import type { ApiStream, ApiStreamUsageChunk } from "../transform/stream"
+import { ToolCallProcessor } from "../transform/tool-call-processor"
+import type { OpenRouterErrorResponse } from "./types"
+
+interface ClineHandlerOptions extends CommonApiHandlerOptions {
+	ulid?: string
+	taskId?: string
+	reasoningEffort?: string
+	thinkingBudgetTokens?: number
+	openRouterProviderSorting?: string
+	openRouterModelId?: string
+	openRouterModelInfo?: ModelInfo
+	clineAccountId?: string
+	clineApiKey?: string
+	enableParallelToolCalling?: boolean
+}
+
+function normalizeModelId(modelId: string): string {
+	return modelId.trim().toLowerCase()
+}
+
+const CLINE_FREE_MODEL_IDS = new Set(CLINE_RECOMMENDED_MODELS_FALLBACK.free.map((model) => normalizeModelId(model.id)))
+
+function getCacheReadTokens(usage: any): number {
+	return usage?.prompt_tokens_details?.cached_tokens || usage?.cache_read_input_tokens || 0
+}
+
+function getCacheWriteTokens(usage: any): number {
+	return usage?.prompt_tokens_details?.cache_write_tokens || usage?.cache_creation_input_tokens || 0
+}
+
+export class ClineHandler implements ApiHandler {
+	private options: ClineHandlerOptions
+	private clineAccountService = ClineAccountService.getInstance()
+	private _authService: AuthService
+	private client: OpenAI | undefined
+	lastGenerationId?: string
+	private lastRequestId?: string
+
+	private get _baseUrl(): string {
+		return ClineEnv.config().apiBaseUrl
+	}
+
+	constructor(options: ClineHandlerOptions) {
+		this.options = options
+		this._authService = AuthService.getInstance()
+	}
+
+	private async getFreeModelIdSet(): Promise<Set<string>> {
+		try {
+			const models = await refreshClineRecommendedModels()
+			const freeModelIds = models.free.map((model) => normalizeModelId(model.id)).filter((modelId) => modelId.length > 0)
+			if (freeModelIds.length > 0) {
+				return new Set(freeModelIds)
+			}
+		} catch (error) {
+			Logger.error("Error resolving Cline free model IDs from recommended models:", error)
+		}
+
+		return CLINE_FREE_MODEL_IDS
+	}
+
+	private async ensureClient(): Promise<OpenAI> {
+		const clineAccountAuthToken = this.options.clineApiKey || (await this._authService.getAuthToken())
+		if (!clineAccountAuthToken) {
+			throw new Error(CLINE_ACCOUNT_AUTH_ERROR_MESSAGE)
+		}
+		if (!this.client) {
+			try {
+				const defaultHeaders: Record<string, string> = {
+					"HTTP-Referer": "https://cline.bot",
+					"X-Title": "Cline",
+					"X-Task-ID": this.options.ulid || "",
+				}
+				Object.assign(defaultHeaders, await buildClineExtraHeaders())
+
+				this.client = new OpenAI({
+					baseURL: `${this._baseUrl}/api/v1`,
+					apiKey: clineAccountAuthToken,
+					defaultHeaders,
+					// Capture real HTTP request ID from initial streaming response headers
+					fetch: async (...args: Parameters<typeof fetch>): Promise<Awaited<ReturnType<typeof fetch>>> => {
+						const [input, init] = args
+						const resp = await fetch(input, init)
+						try {
+							let urlStr = ""
+							if (typeof input === "string") {
+								urlStr = input
+							} else if (input instanceof URL) {
+								urlStr = input.toString()
+							} else if (typeof (input as { url?: unknown }).url === "string") {
+								urlStr = (input as { url: string }).url
+							}
+							// Only record for chat completions (the primary streaming request)
+							if (urlStr.includes("/chat/completions")) {
+								const rid = resp.headers.get("x-request-id") || resp.headers.get("request-id")
+								if (rid) {
+									this.lastRequestId = rid
+								}
+							}
+						} catch {
+							// ignore header capture errors
+						}
+						return resp
+					},
+				})
+			} catch (error: any) {
+				throw new Error(`Error creating Cline client: ${error.message}`)
+			}
+		}
+		// Ensure the client is always using the latest auth token
+		this.client.apiKey = clineAccountAuthToken
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		try {
+			const client = await this.ensureClient()
+
+			this.lastGenerationId = undefined
+			this.lastRequestId = undefined
+
+			let didOutputUsage = false
+			const freeModelIds = await this.getFreeModelIdSet()
+
+			const stream = await createOpenRouterStream(
+				client,
+				systemPrompt,
+				messages,
+				this.getModel(),
+				this.options.reasoningEffort,
+				this.options.thinkingBudgetTokens,
+				this.options.openRouterProviderSorting,
+				tools,
+				this.options.enableParallelToolCalling,
+			)
+
+			const toolCallProcessor = new ToolCallProcessor()
+
+			for await (const chunk of stream) {
+				Logger.debug("ClineHandler chunk:" + JSON.stringify(chunk))
+				// openrouter returns an error object instead of the openai sdk throwing an error
+				if ("error" in chunk) {
+					const error = chunk.error as OpenRouterErrorResponse["error"]
+					Logger.error(`Cline API Error: ${error?.code} - ${error?.message}`)
+					// Include metadata in the error message if available
+					const metadataStr = error.metadata ? `\nMetadata: ${JSON.stringify(error.metadata, null, 2)}` : ""
+					throw new Error(`Cline API Error ${error.code}: ${error.message}${metadataStr}`)
+				}
+
+				if (!this.lastGenerationId && chunk.id) {
+					this.lastGenerationId = chunk.id
+				}
+
+				// Check for mid-stream error via finish_reason
+				const choice = chunk.choices?.[0]
+				// OpenRouter may return finish_reason = "error" with error details
+				if ((choice?.finish_reason as string) === "error") {
+					const choiceWithError = choice as any
+					if (choiceWithError.error) {
+						const error = choiceWithError.error
+						Logger.error(`Cline Mid-Stream Error: ${error.code || error.type || "Unknown"} - ${error.message}`)
+						throw new Error(`Cline Mid-Stream Error: ${error.code || error.type || "Unknown"} - ${error.message}`)
+					}
+					throw new Error("Cline Mid-Stream Error: Stream terminated with error status but no error details provided")
+				}
+
+				const delta = choice?.delta
+
+				if (delta?.content) {
+					yield {
+						type: "text",
+						text: delta.content,
+					}
+				}
+
+				if (delta?.tool_calls) {
+					yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+				}
+
+				// Reasoning tokens are returned separately from the content
+				// Skip reasoning content for Grok 4 models since it only displays "thinking" without providing useful information
+				if (
+					delta &&
+					"reasoning" in delta &&
+					delta.reasoning &&
+					!shouldSkipReasoningForModel(this.options.openRouterModelId)
+				) {
+					yield {
+						type: "reasoning",
+						reasoning: typeof delta.reasoning === "string" ? delta.reasoning : JSON.stringify(delta.reasoning),
+					}
+				}
+
+				/* 
+				OpenRouter passes reasoning details that we can pass back unmodified in api requests to preserve reasoning traces for model
+				  - The reasoning_details array in each chunk may contain one or more reasoning objects
+				  - For encrypted reasoning, the content may appear as [REDACTED] in streaming responses
+				  - The complete reasoning sequence is built by concatenating all chunks in order
+				See: https://openrouter.ai/docs/use-cases/reasoning-tokens#preserving-reasoning-blocks
+				*/
+				if (
+					delta &&
+					"reasoning_details" in delta &&
+					delta.reasoning_details &&
+					// @ts-expect-error-next-line
+					delta?.reasoning_details?.length && // exists and non-0
+					!shouldSkipReasoningForModel(this.options.openRouterModelId)
+				) {
+					yield {
+						type: "reasoning",
+						reasoning: "",
+						details: delta.reasoning_details,
+					}
+				}
+
+				if (!didOutputUsage && chunk.usage) {
+					// @ts-expect-error-next-line
+					let totalCost = (chunk.usage.cost || 0) + (chunk.usage.cost_details?.upstream_inference_cost || 0)
+					const modelId = this.getModel().id
+					const isFreeModel = freeModelIds.has(normalizeModelId(modelId))
+					const cacheReadTokens = getCacheReadTokens(chunk.usage)
+					const cacheWriteTokens = getCacheWriteTokens(chunk.usage)
+
+					if (isFreeModel) {
+						totalCost = 0
+					}
+
+					yield {
+						type: "usage",
+						cacheWriteTokens,
+						cacheReadTokens,
+						inputTokens: Math.max(0, (chunk.usage.prompt_tokens || 0) - cacheReadTokens - cacheWriteTokens),
+						outputTokens: chunk.usage.completion_tokens || 0,
+						totalCost,
+					}
+					didOutputUsage = true
+				}
+			}
+
+			// Fallback to generation endpoint if usage chunk not returned
+			if (!didOutputUsage) {
+				Logger.warn("Cline API did not return usage chunk, fetching from generation endpoint")
+				const apiStreamUsage = await this.getApiStreamUsage(freeModelIds)
+				if (apiStreamUsage) {
+					yield apiStreamUsage
+				}
+			}
+		} catch (error) {
+			Logger.error("Cline API Error:", error)
+			throw error
+		}
+	}
+
+	async getApiStreamUsage(freeModelIds?: Set<string>): Promise<ApiStreamUsageChunk | undefined> {
+		if (this.lastGenerationId) {
+			try {
+				const resolvedFreeModelIds = freeModelIds || (await this.getFreeModelIdSet())
+				const clineAccountAuthToken = await this._authService.getAuthToken()
+				if (!clineAccountAuthToken) {
+					throw new Error(CLINE_ACCOUNT_AUTH_ERROR_MESSAGE)
+				}
+				const headers: Record<string, string> = {
+					// Align with backend auth expectations
+					Authorization: `Bearer ${clineAccountAuthToken}`,
+				}
+				Object.assign(headers, await buildClineExtraHeaders())
+
+				const response = await axios.get(`${this.clineAccountService.baseUrl}/generation?id=${this.lastGenerationId}`, {
+					headers,
+					timeout: 15_000, // this request hangs sometimes
+					...getAxiosSettings(),
+				})
+
+				const generation = response.data
+				let totalCost = generation?.total_cost || 0
+				const modelId = this.getModel().id
+				const isFreeModel = resolvedFreeModelIds.has(normalizeModelId(modelId))
+
+				if (isFreeModel) {
+					totalCost = 0
+				}
+
+				return {
+					type: "usage",
+					cacheWriteTokens: generation?.native_tokens_cache_write || 0,
+					cacheReadTokens: generation?.native_tokens_cached || 0,
+					// openrouter generation endpoint fails often
+					inputTokens: Math.max(
+						0,
+						(generation?.native_tokens_prompt || 0) -
+							(generation?.native_tokens_cached || 0) -
+							(generation?.native_tokens_cache_write || 0),
+					),
+					outputTokens: generation?.native_tokens_completion || 0,
+					totalCost,
+				}
+			} catch (error) {
+				// ignore if fails
+				Logger.error("Error fetching cline generation details:", error)
+			}
+		}
+		return undefined
+	}
+
+	// Expose the last HTTP request ID captured from response headers (X-Request-ID)
+	getLastRequestId(): string | undefined {
+		return this.lastRequestId
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		const modelId = this.options.openRouterModelId
+		const modelInfo = this.options.openRouterModelInfo
+		if (modelId && modelInfo) {
+			return { id: modelId, info: modelInfo }
+		}
+		// If we have a model ID but no model info (e.g., CLI featured models),
+		// use the ID with default model info rather than falling back to a different model
+		if (modelId) {
+			return { id: modelId, info: openRouterDefaultModelInfo }
+		}
+		return { id: openRouterDefaultModelId, info: openRouterDefaultModelInfo }
+	}
+}

apps/vscode/src/core/api/providers/deepseek.ts

@@ -0,0 +1,142 @@
+import { DeepSeekModelId, deepSeekDefaultModelId, deepSeekModels, ModelInfo } from "@shared/api"
+import { calculateApiCostOpenAI } from "@utils/cost"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { fetch } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { addReasoningContent } from "../transform/r1-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface DeepSeekHandlerOptions extends CommonApiHandlerOptions {
+	deepSeekApiKey?: string
+	apiModelId?: string
+}
+
+export class DeepSeekHandler implements ApiHandler {
+	private options: DeepSeekHandlerOptions
+	private client: OpenAI | undefined
+
+	constructor(options: DeepSeekHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.deepSeekApiKey) {
+				throw new Error("DeepSeek API key is required")
+			}
+			try {
+				this.client = new OpenAI({
+					baseURL: "https://api.deepseek.com/v1",
+					apiKey: this.options.deepSeekApiKey,
+					defaultHeaders: buildExternalBasicHeaders(),
+					fetch, // Use configured fetch with proxy support
+				})
+			} catch (error) {
+				throw new Error(`Error creating DeepSeek client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	private async *yieldUsage(info: ModelInfo, usage: OpenAI.Completions.CompletionUsage | undefined): ApiStream {
+		// Deepseek reports total input AND cache reads/writes,
+		// see context caching: https://api-docs.deepseek.com/guides/kv_cache)
+		// where the input tokens is the sum of the cache hits/misses, just like OpenAI.
+		// This affects:
+		// 1) context management truncation algorithm, and
+		// 2) cost calculation
+
+		// Deepseek usage includes extra fields.
+		// Safely cast the prompt token details section to the appropriate structure.
+		interface DeepSeekUsage extends OpenAI.CompletionUsage {
+			prompt_cache_hit_tokens?: number
+			prompt_cache_miss_tokens?: number
+		}
+		const deepUsage = usage as DeepSeekUsage
+
+		const inputTokens = deepUsage?.prompt_tokens || 0 // sum of cache hits and misses
+		const outputTokens = deepUsage?.completion_tokens || 0
+		const cacheReadTokens = deepUsage?.prompt_cache_hit_tokens || 0
+		const cacheWriteTokens = deepUsage?.prompt_cache_miss_tokens || 0
+		const totalCost = calculateApiCostOpenAI(info, inputTokens, outputTokens, cacheWriteTokens, cacheReadTokens)
+		const nonCachedInputTokens = Math.max(0, inputTokens - cacheReadTokens - cacheWriteTokens) // this will always be 0
+		yield {
+			type: "usage",
+			inputTokens: nonCachedInputTokens,
+			outputTokens: outputTokens,
+			cacheWriteTokens: cacheWriteTokens,
+			cacheReadTokens: cacheReadTokens,
+			totalCost: totalCost,
+		}
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+
+		const isDeepSeekThinkingModel =
+			model.id.includes("deepseek-reasoner") || model.id === "deepseek-v4-flash" || model.id === "deepseek-v4-pro"
+
+		const convertedMessages = convertToOpenAiMessages(messages)
+		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = isDeepSeekThinkingModel
+			? [{ role: "system", content: systemPrompt }, ...addReasoningContent(convertedMessages, messages)]
+			: [{ role: "system", content: systemPrompt }, ...convertedMessages]
+
+		const stream = await client.chat.completions.create({
+			model: model.id,
+			max_completion_tokens: model.info.maxTokens,
+			messages: openAiMessages,
+			stream: true,
+			stream_options: { include_usage: true },
+			// Only set temperature for non-thinking models
+			...(isDeepSeekThinkingModel ? {} : { temperature: 0 }),
+			...getOpenAIToolParams(tools),
+		})
+
+		const toolCallProcessor = new ToolCallProcessor()
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			if (delta && "reasoning_content" in delta && delta.reasoning_content) {
+				yield {
+					type: "reasoning",
+					reasoning: (delta.reasoning_content as string | undefined) || "",
+				}
+			}
+
+			if (chunk.usage) {
+				yield* this.yieldUsage(model.info, chunk.usage)
+			}
+		}
+	}
+
+	getModel(): { id: DeepSeekModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+		if (modelId && modelId in deepSeekModels) {
+			const id = modelId as DeepSeekModelId
+			return { id, info: deepSeekModels[id] }
+		}
+		return {
+			id: deepSeekDefaultModelId,
+			info: deepSeekModels[deepSeekDefaultModelId],
+		}
+	}
+}

apps/vscode/src/core/api/providers/dify.ts

@@ -0,0 +1,657 @@
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { fetch } from "@/shared/net"
+import { Logger } from "@/shared/services/Logger"
+import { ModelInfo } from "../../../shared/api"
+import { ApiHandler } from "../index"
+import { ApiStream } from "../transform/stream"
+
+interface DifyHandlerOptions {
+	difyApiKey?: string
+	difyBaseUrl?: string
+}
+
+// Dify API Response Types
+export interface DifyFileResponse {
+	id: string
+	name: string
+	size: number
+	extension: string
+	mime_type: string
+	created_by: string
+	created_at: number
+}
+
+export interface DifyMessage {
+	id: string
+	conversation_id: string
+	inputs: Record<string, any>
+	query: string
+	message_files: Array<{
+		id: string
+		type: string
+		url: string
+		belongs_to: string
+	}>
+	answer: string
+	created_at: number
+	feedback?: {
+		rating: string
+	}
+	retriever_resources?: any[]
+}
+
+interface DifyHistoryResponse {
+	data: DifyMessage[]
+	has_more: boolean
+	limit: number
+}
+
+interface DifyConversation {
+	id: string
+	name: string
+	inputs: Record<string, any>
+	status: string
+	introduction: string
+	created_at: number
+	updated_at: number
+}
+
+interface DifyConversationsResponse {
+	data: DifyConversation[]
+	has_more: boolean
+	limit: number
+}
+
+interface DifyConversationResponse {
+	id: string
+	name: string
+	inputs: Record<string, any>
+	status: string
+	introduction: string
+	created_at: number
+	updated_at: number
+}
+
+export class DifyHandler implements ApiHandler {
+	private options: DifyHandlerOptions
+	private baseUrl: string
+	private apiKey: string
+	private conversationId: string | null = null
+	private currentTaskId: string | null = null
+	private abortController: AbortController | null = null
+
+	constructor(options: DifyHandlerOptions) {
+		this.options = options
+		this.apiKey = options.difyApiKey || ""
+		this.baseUrl = options.difyBaseUrl || ""
+
+		Logger.log("[DIFY DEBUG] Constructor called with:", {
+			hasApiKey: !!this.apiKey,
+			baseUrl: this.baseUrl,
+		})
+
+		if (!this.apiKey) {
+			throw new Error("Dify API key is required")
+		}
+		if (!this.baseUrl) {
+			throw new Error("Dify base URL is required")
+		}
+	}
+
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[]): ApiStream {
+		Logger.log("[DIFY DEBUG] createMessage called with:", {
+			systemPromptLength: systemPrompt?.length || 0,
+			messagesCount: messages?.length || 0,
+		})
+
+		// Convert messages to Dify format
+		const query = this.convertMessagesToQuery(systemPrompt, messages)
+		const requestBody = {
+			inputs: {},
+			query: query,
+			response_mode: "streaming",
+			conversation_id: this.conversationId || "",
+			user: "cline-user", // A unique user identifier
+			files: [],
+		}
+
+		const fullUrl = `${this.baseUrl}/chat-messages`
+		Logger.log("[DIFY DEBUG] Making request to:", fullUrl)
+		Logger.log("[DIFY DEBUG] Request body:", JSON.stringify(requestBody, null, 2))
+
+		let response: Response
+		try {
+			response = await fetch(fullUrl, {
+				method: "POST",
+				headers: this.jsonHeaders(),
+				body: JSON.stringify(requestBody),
+			})
+		} catch (error: any) {
+			Logger.error("[DIFY DEBUG] Network error during fetch:", error)
+			const cause = error.cause ? ` | Cause: ${error.cause}` : ""
+			throw new Error(`Dify API network error: ${error.message}${cause}`)
+		}
+
+		Logger.log("[DIFY DEBUG] Response status:", response.status)
+		const headersObj: Record<string, string> = {}
+		response.headers.forEach((value, key) => {
+			headersObj[key] = value
+		})
+		Logger.log("[DIFY DEBUG] Response headers:", headersObj)
+
+		if (!response.ok) {
+			const errorText = await response.text()
+			Logger.error("[DIFY DEBUG] Error response:", errorText)
+			throw new Error(`Dify API error: ${response.status} ${response.statusText} - ${errorText}`)
+		}
+
+		if (!response.body) {
+			throw new Error("No response body from Dify API")
+		}
+
+		const reader = response.body.getReader()
+		const decoder = new TextDecoder()
+		let buffer = ""
+		let fullText = ""
+		let hasYieldedContent = false
+		const processedEvents: string[] = []
+		let lastEventTime = Date.now()
+
+		Logger.log("[DIFY DEBUG] Starting to read streaming response...")
+
+		try {
+			while (true) {
+				const { done, value } = await reader.read()
+				if (done) {
+					Logger.log("[DIFY DEBUG] Stream ended naturally")
+					Logger.log(
+						"[DIFY DEBUG] Final state - hasYieldedContent:",
+						hasYieldedContent,
+						"fullText length:",
+						fullText.length,
+						"processedEvents:",
+						processedEvents,
+					)
+					break
+				}
+
+				const chunk = decoder.decode(value, { stream: true })
+				Logger.log("[DIFY DEBUG] Raw chunk received:", JSON.stringify(chunk))
+
+				buffer += chunk
+				const lines = buffer.split("\n")
+
+				// Keep the last incomplete line in the buffer
+				buffer = lines.pop() || ""
+
+				for (const line of lines) {
+					Logger.log("[DIFY DEBUG] Processing line:", JSON.stringify(line))
+
+					if (line.startsWith("data: ")) {
+						const data = line.slice(6).trim()
+						Logger.log("[DIFY DEBUG] Extracted data:", JSON.stringify(data))
+
+						if (data === "[DONE]") {
+							Logger.log("[DIFY DEBUG] Received [DONE] signal")
+							break
+						}
+
+						if (data === "") {
+							Logger.log("[DIFY DEBUG] Empty data line, skipping")
+							continue
+						}
+
+						try {
+							const parsed = JSON.parse(data)
+							Logger.log("[DIFY DEBUG] Parsed JSON:", parsed)
+							processedEvents.push(parsed.event || "unknown")
+							lastEventTime = Date.now()
+
+							// Capture conversation_id as soon as it's available
+							if (parsed.conversation_id && !this.conversationId) {
+								this.conversationId = parsed.conversation_id
+								Logger.log("[DIFY DEBUG] Captured conversation_id:", this.conversationId)
+							}
+
+							// Handle different Dify event types based on actual Dify API
+							if (parsed.event === "message") {
+								Logger.log("[DIFY DEBUG] Message event, answer:", parsed.answer)
+								// Dify sends the full text in each "answer" chunk, so we replace.
+								if (typeof parsed.answer === "string") {
+									fullText = parsed.answer
+									Logger.log("[DIFY DEBUG] Updated fullText length:", fullText.length)
+									yield {
+										type: "text",
+										text: fullText,
+									}
+									hasYieldedContent = true
+								}
+							} else if (parsed.event === "message_replace") {
+								Logger.log("[DIFY DEBUG] Replace message event:", parsed)
+								if (parsed.answer) {
+									fullText = parsed.answer // Replace instead of append
+									Logger.log("[DIFY DEBUG] Replaced fullText length:", fullText.length)
+									yield {
+										type: "text",
+										text: fullText,
+									}
+									hasYieldedContent = true
+								}
+							} else if (parsed.event === "message_end") {
+								Logger.log("[DIFY DEBUG] Message end event", parsed)
+								// Message completed. Yield final text if we have any.
+								if (fullText) {
+									yield {
+										type: "text",
+										text: fullText,
+									}
+									hasYieldedContent = true
+								}
+								// Yield usage data if available
+								if (parsed.usage) {
+									yield {
+										type: "usage",
+										inputTokens: parsed.usage.prompt_tokens || 0,
+										outputTokens: parsed.usage.completion_tokens || parsed.usage.total_tokens || 0,
+										totalCost: parsed.usage.total_price || 0,
+									}
+								}
+								return // End of stream
+							} else if (parsed.event === "error") {
+								Logger.error("[DIFY DEBUG] Error event:", parsed)
+								throw new Error(`Dify API error: ${parsed.message || "Unknown error"}`)
+							} else if (parsed.event === "workflow_started" || parsed.event === "workflow_finished") {
+								Logger.log("[DIFY DEBUG] Workflow event:", parsed.event)
+								// These are informational events, continue processing
+							} else if (parsed.event === "node_started" || parsed.event === "node_finished") {
+								Logger.log("[DIFY DEBUG] Node event:", parsed.event, parsed.data)
+								// These are informational events, continue processing
+							} else if (parsed.event === "ping") {
+								Logger.log("[DIFY DEBUG] Ping event received, keeping connection alive.")
+								// Ping event, do nothing
+							} else {
+								Logger.log("[DIFY DEBUG] Unknown event type:", parsed.event, "Full object:", parsed)
+								// Try to extract text from other possible fields
+								if (parsed.text) {
+									fullText += parsed.text
+									yield {
+										type: "text",
+										text: fullText,
+									}
+									hasYieldedContent = true
+								} else if (parsed.content) {
+									fullText += parsed.content
+									yield {
+										type: "text",
+										text: fullText,
+									}
+									hasYieldedContent = true
+								} else if (parsed.answer) {
+									// Fallback: some events might have answer field even if not "message" type
+									fullText += parsed.answer
+									yield {
+										type: "text",
+										text: fullText,
+									}
+									hasYieldedContent = true
+								}
+							}
+						} catch (e) {
+							Logger.warn("[DIFY DEBUG] Failed to parse JSON:", data, "Error:", e)
+						}
+					} else if (line.trim() !== "") {
+						Logger.log(
+							"[DIFY DEBUG] Non-data line (not starting with 'data:'), trying to parse as direct JSON:",
+							JSON.stringify(line),
+						)
+						// Try to parse as direct JSON (fallback for non-SSE responses, though Dify uses SSE)
+						try {
+							const parsed = JSON.parse(line.trim())
+							Logger.log("[DIFY DEBUG] Parsed direct JSON:", parsed)
+							processedEvents.push(parsed.event || "direct-json")
+
+							// Handle the same event types as above
+							if (parsed.event === "message" && parsed.answer) {
+								fullText += parsed.answer
+								yield {
+									type: "text",
+									text: fullText,
+								}
+								hasYieldedContent = true
+							} else if (parsed.event === "message_end") {
+								if (fullText) {
+									yield {
+										type: "text",
+										text: fullText,
+									}
+									hasYieldedContent = true
+								}
+								return
+							} else if (parsed.event === "error") {
+								Logger.error("[DIFY DEBUG] Direct JSON Error event:", parsed)
+								throw new Error(`Dify API error: ${parsed.message || "Unknown error"}`)
+							} else if (parsed.answer || parsed.text || parsed.content) {
+								// Fallback for any content in direct JSON
+								const content = parsed.answer || parsed.text || parsed.content
+								fullText += content
+								yield {
+									type: "text",
+									text: fullText,
+								}
+								hasYieldedContent = true
+							}
+						} catch (e) {
+							// Not JSON, continue
+							Logger.log("[DIFY DEBUG] Line is not direct JSON, continuing")
+						}
+					}
+				}
+			}
+
+			// Final check - if we haven't yielded any content, provide diagnostic information
+			if (!hasYieldedContent) {
+				const diagnosticInfo = {
+					processedEvents,
+					finalFullTextLength: fullText.length,
+					finalFullText: fullText,
+					streamDuration: Date.now() - lastEventTime,
+					conversationId: this.conversationId,
+				}
+				Logger.error("[DIFY DEBUG] No content was yielded! Diagnostic info:", diagnosticInfo)
+
+				// If we have any accumulated text at all, yield it as a fallback
+				if (fullText.trim()) {
+					Logger.log("[DIFY DEBUG] Yielding accumulated text as fallback:", fullText)
+					yield {
+						type: "text",
+						text: fullText,
+					}
+				} else {
+					// Provide a more informative error
+					throw new Error(
+						`Dify API did not provide any assistant messages. ` +
+							`Events processed: [${processedEvents.join(", ")}]. ` +
+							`Check your Dify application configuration and ensure it's properly set up to return responses. ` +
+							`API URL: ${fullUrl}. Conversation ID: ${this.conversationId || "none"}.`,
+					)
+				}
+			}
+		} finally {
+			reader.releaseLock()
+			Logger.log("[DIFY DEBUG] Stream reader released")
+		}
+	}
+
+	private convertMessagesToQuery(systemPrompt: string, messages: ClineStorageMessage[]): string {
+		// Dify's context is managed by `conversation_id`. The `query` should be the last user message.
+		// The system prompt is typically configured in the Dify App itself.
+		const lastUserMessage = messages.filter((m) => m.role === "user").pop()
+
+		if (!lastUserMessage) {
+			return "" // Should not happen in normal flow
+		}
+
+		const userQuery = Array.isArray(lastUserMessage.content)
+			? lastUserMessage.content.map((c) => ("text" in c ? c.text : "")).join("\n")
+			: (lastUserMessage.content as string)
+
+		// Only prepend the system prompt if it's the very first message of a new conversation.
+		if (!this.conversationId && systemPrompt) {
+			Logger.log("[DIFY DEBUG] Prepending system prompt for new conversation.")
+			return `${systemPrompt}\n\n---\n\n${userQuery}`
+		}
+
+		return userQuery
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		return {
+			id: "dify-workflow",
+			info: {
+				maxTokens: 8192,
+				contextWindow: 128000,
+				supportsImages: true,
+				supportsPromptCache: false,
+				inputPrice: 0,
+				outputPrice: 0,
+				description: "Dify workflow - model selection is configured in your Dify application",
+			},
+		}
+	}
+
+	// Additional Dify API Methods
+
+	/**
+	 * Upload a file for use in conversations
+	 * @param file File buffer to upload
+	 * @param filename Name of the file
+	 * @param user User identifier (defaults to "cline-user")
+	 * @returns Promise with file upload response
+	 */
+	async uploadFile(file: Buffer, filename: string, user: string = "cline-user"): Promise<DifyFileResponse> {
+		const formData = new FormData()
+		formData.append("file", new Blob([new Uint8Array(file)]), filename)
+		formData.append("user", user)
+
+		const response = await fetch(`${this.baseUrl}/files/upload`, {
+			method: "POST",
+			headers: this.headers(),
+			body: formData,
+		})
+
+		if (!response.ok) {
+			const errorText = await response.text()
+			throw new Error(`Dify file upload error: ${response.status} ${response.statusText} - ${errorText}`)
+		}
+
+		return response.json()
+	}
+
+	/**
+	 * Stop generation for a specific task
+	 * @param taskId Task ID from streaming response
+	 * @param user User identifier (defaults to "cline-user")
+	 * @returns Promise that resolves when generation is stopped
+	 */
+	async stopGeneration(taskId: string, user: string = "cline-user"): Promise<void> {
+		const response = await fetch(`${this.baseUrl}/chat-messages/${taskId}/stop`, {
+			method: "POST",
+			headers: this.jsonHeaders(),
+			body: JSON.stringify({ user }),
+		})
+
+		if (!response.ok) {
+			const errorText = await response.text()
+			throw new Error(`Dify stop generation error: ${response.status} ${response.statusText} - ${errorText}`)
+		}
+	}
+
+	/**
+	 * Get conversation history messages with pagination
+	 * @param conversationId Conversation ID
+	 * @param user User identifier (defaults to "cline-user")
+	 * @param firstId First message ID for pagination (optional)
+	 * @param limit Number of messages to return (default: 20)
+	 * @returns Promise with conversation history
+	 */
+	async getConversationHistory(
+		conversationId: string,
+		user: string = "cline-user",
+		firstId?: string,
+		limit: number = 20,
+	): Promise<DifyHistoryResponse> {
+		const params = new URLSearchParams({ user, limit: limit.toString() })
+		if (firstId) {
+			params.append("first_id", firstId)
+		}
+
+		const response = await fetch(`${this.baseUrl}/conversations/${conversationId}/messages?${params}`, {
+			headers: this.headers(),
+		})
+
+		if (!response.ok) {
+			const errorText = await response.text()
+			throw new Error(`Dify get conversation history error: ${response.status} ${response.statusText} - ${errorText}`)
+		}
+
+		return response.json()
+	}
+
+	/**
+	 * Get list of conversations for a user
+	 * @param user User identifier (defaults to "cline-user")
+	 * @param lastId Last conversation ID for pagination (optional)
+	 * @param limit Number of conversations to return (default: 20)
+	 * @param sortBy Sort field (default: "-updated_at")
+	 * @returns Promise with conversations list
+	 */
+	async getConversations(
+		user: string = "cline-user",
+		lastId?: string,
+		limit: number = 20,
+		sortBy: string = "-updated_at",
+	): Promise<DifyConversationsResponse> {
+		const params = new URLSearchParams({
+			user,
+			limit: limit.toString(),
+			sort_by: sortBy,
+		})
+		if (lastId) {
+			params.append("last_id", lastId)
+		}
+
+		const response = await fetch(`${this.baseUrl}/conversations?${params}`, {
+			headers: this.headers(),
+		})
+
+		if (!response.ok) {
+			const errorText = await response.text()
+			throw new Error(`Dify get conversations error: ${response.status} ${response.statusText} - ${errorText}`)
+		}
+
+		return response.json()
+	}
+
+	/**
+	 * Delete a conversation
+	 * @param conversationId Conversation ID to delete
+	 * @param user User identifier (defaults to "cline-user")
+	 * @returns Promise that resolves when conversation is deleted
+	 */
+	async deleteConversation(conversationId: string, user: string = "cline-user"): Promise<void> {
+		const response = await fetch(`${this.baseUrl}/conversations/${conversationId}`, {
+			method: "DELETE",
+			headers: this.jsonHeaders(),
+			body: JSON.stringify({ user }),
+		})
+
+		if (!response.ok) {
+			const errorText = await response.text()
+			throw new Error(`Dify delete conversation error: ${response.status} ${response.statusText} - ${errorText}`)
+		}
+	}
+
+	/**
+	 * Rename a conversation
+	 * @param conversationId Conversation ID to rename
+	 * @param user User identifier (defaults to "cline-user")
+	 * @param name New conversation name (optional if auto_generate is true)
+	 * @param autoGenerate Whether to auto-generate the name (default: false)
+	 * @returns Promise with updated conversation details
+	 */
+	async renameConversation(
+		conversationId: string,
+		user: string = "cline-user",
+		name?: string,
+		autoGenerate: boolean = false,
+	): Promise<DifyConversationResponse> {
+		const body: any = { user, auto_generate: autoGenerate }
+		if (name) {
+			body.name = name
+		}
+
+		const response = await fetch(`${this.baseUrl}/conversations/${conversationId}/name`, {
+			method: "POST",
+			headers: this.jsonHeaders(),
+			body: JSON.stringify(body),
+		})
+
+		if (!response.ok) {
+			const errorText = await response.text()
+			throw new Error(`Dify rename conversation error: ${response.status} ${response.statusText} - ${errorText}`)
+		}
+
+		return response.json()
+	}
+
+	/**
+	 * Submit feedback for a message
+	 * @param messageId Message ID to provide feedback for
+	 * @param rating Rating: "like" or "dislike"
+	 * @param content Optional feedback content
+	 * @param user User identifier (defaults to "cline-user")
+	 * @returns Promise that resolves when feedback is submitted
+	 */
+	async submitMessageFeedback(
+		messageId: string,
+		rating: "like" | "dislike",
+		content?: string,
+		user: string = "cline-user",
+	): Promise<void> {
+		const body: any = { rating, user }
+		if (content) {
+			body.content = content
+		}
+
+		const response = await fetch(`${this.baseUrl}/messages/${messageId}/feedbacks`, {
+			method: "POST",
+			headers: this.jsonHeaders(),
+			body: JSON.stringify(body),
+		})
+
+		if (!response.ok) {
+			const errorText = await response.text()
+			throw new Error(`Dify submit feedback error: ${response.status} ${response.statusText} - ${errorText}`)
+		}
+	}
+
+	/**
+	 * Get current conversation ID
+	 * @returns Current conversation ID or null
+	 */
+	getCurrentConversationId(): string | null {
+		return this.conversationId
+	}
+
+	/**
+	 * Set conversation ID for continuing existing conversations
+	 * @param conversationId Conversation ID to set
+	 */
+	setConversationId(conversationId: string): void {
+		this.conversationId = conversationId
+	}
+
+	/**
+	 * Reset conversation ID to start a new conversation
+	 */
+	resetConversation(): void {
+		this.conversationId = null
+		this.currentTaskId = null
+	}
+
+	private jsonHeaders() {
+		return {
+			...this.headers(),
+			"Content-Type": "application/json",
+		}
+	}
+
+	private headers() {
+		const externalHeaders = buildExternalBasicHeaders()
+		return {
+			...externalHeaders,
+			Authorization: `Bearer ${this.apiKey}`,
+		}
+	}
+}

apps/vscode/src/core/api/providers/doubao.ts

@@ -0,0 +1,90 @@
+import { DoubaoModelId, doubaoDefaultModelId, doubaoModels, ModelInfo } from "@shared/api"
+import OpenAI from "openai"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from ".."
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+
+interface DoubaoHandlerOptions extends CommonApiHandlerOptions {
+	doubaoApiKey?: string
+	apiModelId?: string
+}
+
+export class DoubaoHandler implements ApiHandler {
+	private options: DoubaoHandlerOptions
+	private client: OpenAI | undefined
+	constructor(options: DoubaoHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.doubaoApiKey) {
+				throw new Error("Doubao API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: "https://ark.cn-beijing.volces.com/api/v3/",
+					apiKey: this.options.doubaoApiKey,
+				})
+			} catch (error) {
+				throw new Error(`Error creating Doubao client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	getModel(): { id: DoubaoModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+		if (modelId && modelId in doubaoModels) {
+			const id = modelId as DoubaoModelId
+			return { id, info: doubaoModels[id] }
+		}
+		return {
+			id: doubaoDefaultModelId,
+			info: doubaoModels[doubaoDefaultModelId],
+		}
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[]): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+		const stream = await client.chat.completions.create({
+			model: model.id,
+			max_completion_tokens: model.info.maxTokens,
+			messages: openAiMessages,
+			stream: true,
+			stream_options: { include_usage: true },
+			temperature: 0,
+		})
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (chunk.usage) {
+				yield {
+					type: "usage",
+					inputTokens: chunk.usage.prompt_tokens || 0,
+					outputTokens: chunk.usage.completion_tokens || 0,
+					// @ts-expect-error-next-line
+					cacheReadTokens: chunk.usage.prompt_cache_hit_tokens || 0,
+					// @ts-expect-error-next-line
+					cacheWriteTokens: chunk.usage.prompt_cache_miss_tokens || 0,
+				}
+			}
+		}
+	}
+}

apps/vscode/src/core/api/providers/fireworks.ts

@@ -0,0 +1,116 @@
+import { FireworksModelId, fireworksDefaultModelId, fireworksModels, ModelInfo } from "@shared/api"
+import OpenAI from "openai"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from ".."
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+
+interface FireworksHandlerOptions extends CommonApiHandlerOptions {
+	fireworksApiKey?: string
+	fireworksModelId?: string
+	fireworksModelMaxCompletionTokens?: number
+	fireworksModelMaxTokens?: number
+}
+
+export class FireworksHandler implements ApiHandler {
+	private options: FireworksHandlerOptions
+	private client: OpenAI | undefined
+
+	constructor(options: FireworksHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.fireworksApiKey) {
+				throw new Error("Fireworks API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: "https://api.fireworks.ai/inference/v1",
+					apiKey: this.options.fireworksApiKey,
+				})
+			} catch (error) {
+				throw new Error(`Error creating Fireworks client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[]): ApiStream {
+		const client = this.ensureClient()
+		const modelId = this.options.fireworksModelId ?? ""
+
+		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+
+		const stream = await client.chat.completions.create({
+			model: modelId,
+			messages: openAiMessages,
+			stream: true,
+			stream_options: { include_usage: true },
+			temperature: 0,
+		})
+
+		let reasoning: string | null = null
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (reasoning || delta?.content?.includes("<think>")) {
+				reasoning = (reasoning || "") + (delta.content ?? "")
+			}
+
+			if (delta?.content && !reasoning) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (reasoning || (delta && "reasoning_content" in delta && delta.reasoning_content)) {
+				yield {
+					type: "reasoning",
+					reasoning: delta.content || ((delta as any).reasoning_content as string | undefined) || "",
+				}
+				if (reasoning?.includes("</think>")) {
+					// Reset so the next chunk is regular content
+					reasoning = null
+				}
+			}
+
+			if (chunk.usage) {
+				const usage = chunk.usage as OpenAI.CompletionUsage & {
+					prompt_cache_hit_tokens?: number
+					prompt_cache_miss_tokens?: number
+					prompt_tokens_details?: {
+						cached_tokens?: number
+					}
+				}
+				yield {
+					type: "usage",
+					inputTokens: usage.prompt_tokens || 0,
+					outputTokens: usage.completion_tokens || 0,
+					// Fireworks can return cache hits either as prompt_cache_hit_tokens or prompt_tokens_details.cached_tokens.
+					cacheReadTokens: usage.prompt_cache_hit_tokens ?? usage.prompt_tokens_details?.cached_tokens ?? 0,
+					cacheWriteTokens: usage.prompt_cache_miss_tokens || 0,
+				}
+			}
+		}
+	}
+
+	getModel(): { id: FireworksModelId; info: ModelInfo } {
+		const modelId = this.options.fireworksModelId
+		if (modelId && modelId in fireworksModels) {
+			const id = modelId as FireworksModelId
+			return { id, info: fireworksModels[id] }
+		}
+		return {
+			id: fireworksDefaultModelId,
+			info: fireworksModels[fireworksDefaultModelId],
+		}
+	}
+}

apps/vscode/src/core/api/providers/gemini-mock.test.ts

@@ -1,10 +1,4 @@
-// Mock for the `@google/genai` module to avoid ESM-in-CommonJS compatibility
-// issues in the VS Code integration test build (the `out/` tree runs as
-// CommonJS). Loaded by test-setup.js, which intercepts `require("@google/genai")`.
-//
-// Previously colocated as apps/vscode/src/core/api/providers/gemini-mock.test.ts
-// alongside the legacy Gemini provider; kept as a standalone test fixture after
-// the legacy provider handlers were removed.
+// Mock for @google/genai module to avoid ESM compatibility issues in tests
 
 export class GoogleGenAI {
 	constructor(_options: any) {

apps/vscode/src/core/api/providers/gemini.ts

@@ -0,0 +1,570 @@
+// Restore GenerateContentConfig import and add GenerateContentResponseUsageMetadata
+import {
+	ApiError,
+	FunctionCallingConfigMode,
+	type GenerateContentConfig,
+	type GenerateContentResponseUsageMetadata,
+	GoogleGenAI,
+	FunctionDeclaration as GoogleTool,
+	ThinkingLevel,
+} from "@google/genai"
+import { GeminiModelId, geminiDefaultModelId, geminiModels, ModelInfo } from "@shared/api"
+import { GEMINI_FLASH_MAX_OUTPUT_TOKENS, isGeminiFlashModel } from "@utils/model-utils"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { telemetryService } from "@/services/telemetry"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { Logger } from "@/shared/services/Logger"
+import { ApiHandler, CommonApiHandlerOptions } from "../"
+import { RetriableError, withRetry } from "../retry"
+import { convertAnthropicMessageToGemini } from "../transform/gemini-format"
+import { ApiStream } from "../transform/stream"
+
+const rateLimitPatterns = [/got status: 429/i, /429 Too Many Requests/i, /rate limit exceeded/i, /too many requests/i]
+
+interface GeminiHandlerOptions extends CommonApiHandlerOptions {
+	isVertex?: boolean
+	vertexProjectId?: string
+	vertexRegion?: string
+	geminiApiKey?: string
+	geminiBaseUrl?: string
+	thinkingBudgetTokens?: number
+	reasoningEffort?: string
+	apiModelId?: string
+	ulid?: string
+}
+
+function mapReasoningEffortToGeminiThinkingLevel(effort: string): ThinkingLevel {
+	switch (effort) {
+		case "low":
+		case "medium":
+			return ThinkingLevel.LOW
+		case "high":
+		case "xhigh":
+			return ThinkingLevel.HIGH
+		default:
+			return ThinkingLevel.LOW
+	}
+}
+
+function getGeminiMaxOutputTokens(modelId: string, modelMaxTokens?: number): number | undefined {
+	if (!isGeminiFlashModel(modelId)) {
+		return undefined
+	}
+
+	if (modelMaxTokens && modelMaxTokens > 0) {
+		return Math.min(modelMaxTokens, GEMINI_FLASH_MAX_OUTPUT_TOKENS)
+	}
+
+	return GEMINI_FLASH_MAX_OUTPUT_TOKENS
+}
+
+/**
+ * Handler for Google's Gemini API with optimized caching strategy and accurate cost accounting.
+ *
+ * Key features:
+ * - One cache per task: Creates a single cache per task and reuses it for subsequent turns
+ * - Stable cache keys: Uses ulid as a stable identifier for caches
+ * - Efficient cache updates: Only updates caches when there's new content to add
+ * - Split cost accounting: Separates immediate costs from ongoing cache storage costs
+ *
+ * Cost accounting approach:
+ * - Immediate costs (per message): Input tokens, output tokens, and cache read costs
+ * - Ongoing costs (per task): Cache storage costs for the TTL period
+ *
+ * Gemini's caching system is unique in that it charges for holding tokens in cache by the hour.
+ * This implementation optimizes for both performance and cost by:
+ * 1. Minimizing redundant cache creations
+ * 2. Properly accounting for cache costs in the billing calculations
+ * 3. Using a stable cache key to ensure cache reuse across turns
+ * 4. Separating immediate costs from ongoing costs to avoid double-counting
+ */
+export class GeminiHandler implements ApiHandler {
+	private options: GeminiHandlerOptions
+	private client: GoogleGenAI | undefined
+
+	constructor(options: GeminiHandlerOptions) {
+		// Store the options
+		this.options = options
+	}
+
+	private ensureClient(): GoogleGenAI {
+		if (!this.client) {
+			const options = this.options as GeminiHandlerOptions
+			const externalHeaders = buildExternalBasicHeaders()
+
+			if (options.isVertex) {
+				// Initialize with Vertex AI configuration
+				const project = this.options.vertexProjectId ?? "not-provided"
+				const location = this.options.vertexRegion ?? "not-provided"
+
+				try {
+					this.client = new GoogleGenAI({
+						vertexai: true,
+						project,
+						location,
+						httpOptions: {
+							headers: externalHeaders,
+						},
+					})
+				} catch (error) {
+					throw new Error(`Error creating Gemini Vertex AI client: ${error.message}`)
+				}
+			} else {
+				// Initialize with standard API key
+				if (!options.geminiApiKey) {
+					throw new Error("API key is required for Google Gemini when not using Vertex AI")
+				}
+
+				try {
+					this.client = new GoogleGenAI({
+						apiKey: options.geminiApiKey,
+						httpOptions: {
+							headers: externalHeaders,
+						},
+					})
+				} catch (error) {
+					throw new Error(`Error creating Gemini client: ${error.message}`)
+				}
+			}
+		}
+		return this.client
+	}
+
+	/**
+	 * Creates a message using the Gemini API with implicit caching.
+	 *
+	 * Cost accounting:
+	 * - Immediate costs (returned in the usage object): Input tokens, output tokens, cache read costs
+	 *
+	 * @param systemPrompt The system prompt to use for the message
+	 * @param messages The conversation history to include in the message
+	 * @returns An async generator that yields chunks of the response with accurate immediate costs
+	 */
+	@withRetry({
+		maxRetries: 4,
+		baseDelay: 2000,
+		maxDelay: 15000,
+	})
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: GoogleTool[]): ApiStream {
+		const client = this.ensureClient()
+		const { id: modelId, info } = this.getModel()
+		const contents = messages.map(convertAnthropicMessageToGemini)
+		// Gemini may emit multiple function calls under the same responseId and without functionCall.id.
+		// Track a local sequence so each emitted tool call has a stable unique ID.
+		const responseToolCallCount = new Map<string, number>()
+
+		// Configure thinking budget/level if supported
+		const _thinkingBudget = this.options.thinkingBudgetTokens ?? 0
+		const maxBudget = info.thinkingConfig?.maxBudget ?? 24576
+		const thinkingBudget = Math.min(_thinkingBudget, maxBudget)
+		// When ThinkingLevel is defined, thinking budget cannot be zero
+		// and only level is used to control thinking behavior.
+		// Only set thinkingLevel for models that support it
+		let thinkingLevel: ThinkingLevel | undefined
+		const rawReasoningEffort = (this.options.reasoningEffort || "").toLowerCase()
+		const normalizedReasoningEffort = !rawReasoningEffort || rawReasoningEffort === "none" ? "low" : rawReasoningEffort
+		if (info.thinkingConfig?.supportsThinkingLevel) {
+			thinkingLevel = mapReasoningEffortToGeminiThinkingLevel(normalizedReasoningEffort)
+		}
+
+		// Set up base generation config
+		const maxOutputTokens = getGeminiMaxOutputTokens(modelId, info.maxTokens)
+		const requestConfig: GenerateContentConfig = {
+			// Add base URL if configured
+			httpOptions: this.options.geminiBaseUrl ? { baseUrl: this.options.geminiBaseUrl } : undefined,
+			systemInstruction: systemPrompt,
+			// Set temperature (default to 0)
+			// Gemini 3 recommends 1.0
+			temperature: info.temperature ?? 1,
+			...(maxOutputTokens !== undefined ? { maxOutputTokens } : {}),
+		}
+
+		// Add thinking config only if the model supports it
+		if (info.thinkingConfig) {
+			requestConfig.thinkingConfig = {
+				// Turn off thinking:
+				// thinkingBudget: 0
+				// Turn on dynamic thinking:
+				// thinkingBudget: -1
+				// Turn on fixed thinking budget:
+				thinkingBudget: thinkingLevel ? undefined : thinkingBudget, // Use budget only if thinkingLevel is not set
+				thinkingLevel,
+				includeThoughts: thinkingBudget > 0 || !!thinkingLevel,
+			}
+		}
+
+		// Generate content using the configured parameters
+		const sdkCallStartTime = Date.now()
+		let responseId: string | undefined
+		let sdkFirstChunkTime: number | undefined
+		let ttftSdkMs: number | undefined
+		let apiSuccess = false
+		let apiError: string | undefined
+		let promptTokens = 0
+		let outputTokens = 0
+		let cacheReadTokens = 0
+		let thoughtsTokenCount = 0 // Initialize thought token counts
+		let lastUsageMetadata: GenerateContentResponseUsageMetadata | undefined
+
+		const isNativeToolCallsEnabled = tools?.length
+		if (isNativeToolCallsEnabled) {
+			requestConfig.tools = [{ functionDeclarations: tools }]
+			requestConfig.toolConfig = {
+				// Force the model to call 'any' function.
+				functionCallingConfig: {
+					mode: FunctionCallingConfigMode.ANY,
+				},
+			}
+		}
+
+		try {
+			const result = await client.models.generateContentStream({
+				model: modelId,
+				contents: contents,
+				config: {
+					...requestConfig,
+				},
+			})
+
+			let isFirstSdkChunk = true
+			for await (const chunk of result) {
+				const responseKey = chunk.responseId || "gemini-response"
+				if (isFirstSdkChunk) {
+					sdkFirstChunkTime = Date.now()
+					ttftSdkMs = sdkFirstChunkTime - sdkCallStartTime
+					isFirstSdkChunk = false
+				}
+
+				// Handle thinking content from Gemini's response
+				const parts = chunk?.candidates?.[0]?.content?.parts || []
+				for (const part of parts) {
+					if (part.thought && part.text) {
+						yield {
+							type: "reasoning",
+							id: chunk.responseId,
+							reasoning: part.text || "",
+							signature: part.thoughtSignature,
+						}
+					} else if (part.text) {
+						yield {
+							type: "text",
+							text: part.text,
+							id: chunk.responseId,
+							signature: part.thoughtSignature,
+						}
+					}
+					if (part.functionCall) {
+						const functionCall = part.functionCall
+						const args = Object.entries(functionCall.args || {}).filter(([_key, val]) => !!val)
+						if (functionCall.args && args.length > 0) {
+							const existingId = functionCall.id?.trim()
+							const toolCallId =
+								existingId ??
+								(() => {
+									const sequenceNumber = responseToolCallCount.get(responseKey) ?? 0
+									responseToolCallCount.set(responseKey, sequenceNumber + 1)
+									return `${responseKey}-tool-${sequenceNumber}`
+								})()
+							yield {
+								type: "tool_calls",
+								id: chunk.responseId,
+								tool_call: {
+									call_id: toolCallId,
+									function: {
+										id: toolCallId,
+										name: functionCall.name,
+										arguments: JSON.stringify(functionCall.args),
+									},
+								},
+								signature: part.thoughtSignature,
+							}
+						}
+					}
+				}
+
+				if (chunk.usageMetadata) {
+					responseId = chunk.responseId
+					lastUsageMetadata = chunk.usageMetadata
+					promptTokens = lastUsageMetadata.promptTokenCount ?? promptTokens
+					outputTokens = lastUsageMetadata.candidatesTokenCount ?? outputTokens
+					thoughtsTokenCount = lastUsageMetadata.thoughtsTokenCount ?? thoughtsTokenCount
+					cacheReadTokens = lastUsageMetadata.cachedContentTokenCount ?? cacheReadTokens
+				}
+			}
+			apiSuccess = true
+
+			if (lastUsageMetadata) {
+				const totalCost = this.calculateCost({
+					info,
+					inputTokens: promptTokens,
+					outputTokens,
+					thoughtsTokenCount,
+					cacheReadTokens,
+				})
+				yield {
+					type: "usage",
+					inputTokens: promptTokens - cacheReadTokens,
+					outputTokens,
+					thoughtsTokenCount,
+					cacheReadTokens,
+					cacheWriteTokens: 0,
+					totalCost,
+					id: responseId,
+				}
+			}
+		} catch (error) {
+			apiSuccess = false
+			// Let the error propagate to be handled by withRetry or Task.ts
+			// Telemetry will be sent in the finally block.
+			if (error instanceof Error) {
+				apiError = error.message
+
+				if (error instanceof ApiError) {
+					if (error.status === 429) {
+						// The API includes more details in the message
+						// https://github.com/googleapis/js-genai/blob/v1.11.0/src/_api_client.ts#L758
+						const response = this.attemptParse(error.message)
+
+						if (response && response.error) {
+							const responseBody = this.attemptParse(response.error.message)
+
+							if (responseBody.error) {
+								const detail = responseBody.error.details?.find(
+									(d: any) => d["@type"] === "type.googleapis.com/google.rpc.RetryInfo",
+								)
+
+								const detailedError = new RetriableError(
+									apiError,
+									this.parseRetryDelay(detail?.retryDelay) || undefined,
+									{
+										cause: error,
+									},
+								)
+								throw detailedError
+							}
+						}
+
+						throw new RetriableError(apiError, undefined, { cause: error })
+					}
+
+					// Fallback in case Gemini throws a rate limit error without a 429 status code
+					// https://github.com/cline/cline/pull/5205#discussion_r2311761559
+					const isRateLimit = rateLimitPatterns.some((pattern) => pattern.test(error.message))
+					if (isRateLimit) {
+						throw new RetriableError(apiError, undefined, { cause: error })
+					}
+				}
+			} else {
+				apiError = String(error)
+			}
+
+			throw error
+		} finally {
+			const sdkCallEndTime = Date.now()
+			const totalDurationSdkMs = sdkCallEndTime - sdkCallStartTime
+			const cacheHit = cacheReadTokens > 0
+			const cacheHitPercentage = promptTokens > 0 ? (cacheReadTokens / promptTokens) * 100 : undefined
+			const throughputTokensPerSecSdk =
+				totalDurationSdkMs > 0 && outputTokens > 0 ? outputTokens / (totalDurationSdkMs / 1000) : undefined
+
+			if (this.options.ulid) {
+				telemetryService.captureGeminiApiPerformance(this.options.ulid, modelId, {
+					ttftSec: ttftSdkMs !== undefined ? ttftSdkMs / 1000 : undefined,
+					totalDurationSec: totalDurationSdkMs / 1000,
+					promptTokens,
+					outputTokens,
+					cacheReadTokens,
+					cacheHit,
+					cacheHitPercentage,
+					apiSuccess,
+					apiError,
+					throughputTokensPerSec: throughputTokensPerSecSdk,
+				})
+			} else {
+				Logger.warn("GeminiHandler: ulid not available for telemetry in createMessage.")
+			}
+		}
+	}
+
+	/**
+	 * Calculate the immediate dollar cost of the API call based on token usage and model pricing.
+	 *
+	 * This method accounts for the immediate costs of the API call:
+	 * - Input token costs (for uncached tokens)
+	 * - Output token costs
+	 * - Cache read costs
+	 * - Gemini implicit caching has no write costs
+	 *
+	 */
+	public calculateCost({
+		info,
+		inputTokens,
+		outputTokens,
+		thoughtsTokenCount = 0,
+		cacheReadTokens = 0,
+	}: {
+		info: ModelInfo
+		inputTokens: number
+		outputTokens: number
+		thoughtsTokenCount: number
+		cacheReadTokens?: number
+	}) {
+		// Exit early if any required pricing information is missing
+		if (!info.inputPrice || !info.outputPrice) {
+			return undefined
+		}
+
+		let inputPrice = info.inputPrice
+		let outputPrice = info.outputPrice
+		// Right now, we only show the immediate costs of caching and not the ongoing costs of storing the cache
+		let cacheReadsPrice = info.cacheReadsPrice ?? 0
+
+		// If there's tiered pricing then adjust prices based on the input tokens used
+		if (info.tiers) {
+			const tier = info.tiers.find((tier) => inputTokens <= tier.contextWindow)
+			if (tier) {
+				inputPrice = tier.inputPrice ?? inputPrice
+				outputPrice = tier.outputPrice ?? outputPrice
+				cacheReadsPrice = tier.cacheReadsPrice ?? cacheReadsPrice
+			}
+		}
+
+		// Subtract the cached input tokens from the total input tokens
+		const uncachedInputTokens = inputTokens - (cacheReadTokens ?? 0)
+
+		// Calculate immediate costs only
+
+		// 1. Input token costs (for uncached tokens)
+		const inputTokensCost = inputPrice * (uncachedInputTokens / 1_000_000)
+
+		// 2. Output token costs
+		const responseTokensCost = outputPrice * ((outputTokens + thoughtsTokenCount) / 1_000_000)
+
+		// 3. Cache read costs (immediate)
+		const cacheReadCost = (cacheReadTokens ?? 0) > 0 ? cacheReadsPrice * ((cacheReadTokens ?? 0) / 1_000_000) : 0
+
+		// Calculate total immediate cost (excluding cache write/storage costs)
+		const totalCost = inputTokensCost + responseTokensCost + cacheReadCost
+
+		// Create the trace object for debugging
+		const trace: Record<string, { price: number; tokens: number; cost: number }> = {
+			input: { price: inputPrice, tokens: uncachedInputTokens, cost: inputTokensCost },
+			output: { price: outputPrice, tokens: outputTokens, cost: responseTokensCost },
+		}
+
+		// Only include cache read costs in the trace (cache write costs are tracked separately)
+		if ((cacheReadTokens ?? 0) > 0) {
+			trace.cacheRead = { price: cacheReadsPrice, tokens: cacheReadTokens ?? 0, cost: cacheReadCost }
+		}
+
+		// Logger.log(`[GeminiHandler] calculateCost -> ${totalCost}`, trace)
+		return totalCost
+	}
+
+	/**
+	 * Get the model ID and info for the current configuration
+	 */
+	getModel(): { id: GeminiModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+		if (modelId && modelId in geminiModels) {
+			const id = modelId as GeminiModelId
+			return { id, info: geminiModels[id] }
+		}
+		return {
+			id: geminiDefaultModelId,
+			info: geminiModels[geminiDefaultModelId],
+		}
+	}
+
+	/**
+	 * Count tokens in content using the Gemini API
+	 */
+	async countTokens(content: Array<any>): Promise<number> {
+		try {
+			const client = this.ensureClient()
+			const { id: model } = this.getModel()
+
+			// Convert content to Gemini format
+			const geminiContent = content.map((block) => {
+				if (typeof block === "string") {
+					return { text: block }
+				}
+				return { text: JSON.stringify(block) }
+			})
+
+			// Use Gemini's token counting API
+			const response = await client.models.countTokens({
+				model,
+				contents: [{ parts: geminiContent }],
+			})
+
+			if (response.totalTokens === undefined) {
+				Logger.warn("Gemini token counting returned undefined, using fallback")
+				return this.estimateTokens(content)
+			}
+
+			return response.totalTokens
+		} catch (error) {
+			Logger.warn("Gemini token counting failed, using fallback", error)
+			return this.estimateTokens(content)
+		}
+	}
+
+	/**
+	 * Fallback token estimation method
+	 */
+	private estimateTokens(content: Array<any>): number {
+		// Simple estimation: ~4 characters per token
+		const totalChars = content.reduce((total, block) => {
+			if (typeof block === "string") {
+				return total + block.length
+			}
+			if (block && typeof block === "object") {
+				// Safely stringify the object
+				try {
+					const jsonStr = JSON.stringify(block)
+					return total + jsonStr.length
+				} catch (e) {
+					Logger.warn("Failed to stringify block for token estimation", e)
+					return total
+				}
+			}
+			return total
+		}, 0)
+
+		return Math.ceil(totalChars / 4)
+	}
+
+	private parseRetryDelay(retryAfter?: string): number {
+		if (!retryAfter) {
+			return 0
+		}
+
+		const unit = retryAfter.at(-1)
+		const value = Number.parseInt(retryAfter, 10)
+
+		if (Number.isNaN(value)) {
+			return 0
+		}
+
+		if (unit === "s") {
+			return value
+		}
+		if (unit === "m") {
+			return value * 60 // Convert minutes to seconds
+		}
+		if (unit === "h") {
+			return value * 60 * 60 // Convert hours to seconds
+		}
+
+		return value
+	}
+
+	private attemptParse(str: string) {
+		try {
+			return JSON.parse(str)
+		} catch (_) {
+			return null
+		}
+	}
+}

apps/vscode/src/core/api/providers/groq.ts

@@ -0,0 +1,320 @@
+import { GroqModelId, groqDefaultModelId, groqModels, ModelInfo } from "@shared/api"
+import { calculateApiCostOpenAI } from "@utils/cost"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { fetch } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface GroqHandlerOptions extends CommonApiHandlerOptions {
+	groqApiKey?: string
+	groqModelId?: string
+	groqModelInfo?: ModelInfo
+	apiModelId?: string // For backward compatibility
+}
+
+// Enhanced usage interface to support Groq's cached token fields
+interface GroqUsage extends OpenAI.CompletionUsage {
+	prompt_tokens_details?: {
+		cached_tokens?: number
+	}
+}
+
+// Model family definitions for enhanced behavior
+interface GroqModelFamily {
+	name: string
+	supportedFeatures: {
+		streaming: boolean
+		temperature: boolean
+		vision: boolean
+		tools: boolean
+	}
+	maxTokensOverride?: number
+	specialParams?: Record<string, any>
+}
+
+const MODEL_FAMILIES: Record<string, GroqModelFamily> = {
+	// Moonshort 4 Family - Latest generation with vision support
+	"kimi-k2": {
+		name: "kimi-k2",
+		supportedFeatures: { streaming: true, temperature: true, vision: true, tools: true },
+		maxTokensOverride: 8192,
+	},
+	// Llama 4 Family - Latest generation with vision support
+	llama4: {
+		name: "Llama 4",
+		supportedFeatures: { streaming: true, temperature: true, vision: true, tools: true },
+		maxTokensOverride: 8192,
+	},
+	// Llama 3.3 Family - Balanced performance
+	"llama3.3": {
+		name: "Llama 3.3",
+		supportedFeatures: { streaming: true, temperature: true, vision: false, tools: true },
+		maxTokensOverride: 32768,
+	},
+	// Llama 3.1 Family - Fast inference
+	"llama3.1": {
+		name: "Llama 3.1",
+		supportedFeatures: { streaming: true, temperature: true, vision: false, tools: true },
+		maxTokensOverride: 131072,
+	},
+	// DeepSeek Family - Reasoning-optimized
+	deepseek: {
+		name: "DeepSeek",
+		supportedFeatures: { streaming: true, temperature: true, vision: false, tools: true },
+		maxTokensOverride: 8192,
+		specialParams: {
+			top_p: 0.95,
+			reasoning_format: "parsed",
+		},
+	},
+	// Qwen Family - Enhanced for Q&A
+	qwen: {
+		name: "Qwen",
+		supportedFeatures: { streaming: true, temperature: true, vision: false, tools: true },
+		maxTokensOverride: 32768,
+	},
+	// Compound Models - Hybrid architectures
+	compound: {
+		name: "Compound",
+		supportedFeatures: { streaming: true, temperature: true, vision: false, tools: true },
+		maxTokensOverride: 8192,
+	},
+}
+
+export class GroqHandler implements ApiHandler {
+	private options: GroqHandlerOptions
+	private client: OpenAI | undefined
+
+	constructor(options: GroqHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.groqApiKey) {
+				throw new Error("Groq API key is required")
+			}
+			try {
+				this.client = new OpenAI({
+					baseURL: "https://api.groq.com/openai/v1",
+					apiKey: this.options.groqApiKey,
+					defaultHeaders: buildExternalBasicHeaders(),
+					fetch, // Use configured fetch with proxy support
+				})
+			} catch (error) {
+				throw new Error(`Error creating Groq client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	private async *yieldUsage(info: ModelInfo, usage: GroqUsage | undefined): ApiStream {
+		const inputTokens = usage?.prompt_tokens || 0
+		const outputTokens = usage?.completion_tokens || 0
+
+		const cacheReadTokens = usage?.prompt_tokens_details?.cached_tokens || 0
+
+		// Groq does not track cache writes
+		const cacheWriteTokens = 0
+
+		// Calculate cost using OpenAI-compatible cost calculation
+		const totalCost = calculateApiCostOpenAI(info, inputTokens, outputTokens, cacheWriteTokens, cacheReadTokens)
+
+		// Calculate non-cached input tokens for proper reporting
+		const nonCachedInputTokens = Math.max(0, inputTokens - cacheReadTokens - cacheWriteTokens)
+
+		yield {
+			type: "usage",
+			inputTokens: nonCachedInputTokens,
+			outputTokens,
+			cacheWriteTokens,
+			cacheReadTokens,
+			totalCost,
+		}
+	}
+
+	/**
+	 * Detects the model family based on the model ID
+	 */
+	private detectModelFamily(modelId: string): GroqModelFamily {
+		if (modelId.includes("kimi-k2")) {
+			return MODEL_FAMILIES["kimi-k2"]
+		}
+		// Llama 4 variants
+		if (modelId.includes("llama-4") || modelId.includes("llama/llama-4")) {
+			return MODEL_FAMILIES.llama4
+		}
+		// Llama 3.3 variants
+		if (modelId.includes("llama-3.3")) {
+			return MODEL_FAMILIES["llama3.3"]
+		}
+		// Llama 3.1 variants
+		if (modelId.includes("llama-3.1")) {
+			return MODEL_FAMILIES["llama3.1"]
+		}
+		// DeepSeek variants
+		if (modelId.includes("deepseek")) {
+			return MODEL_FAMILIES.deepseek
+		}
+		// Qwen variants
+		if (modelId.includes("qwen")) {
+			return MODEL_FAMILIES.qwen
+		}
+		// Compound variants
+		if (modelId.includes("compound")) {
+			return MODEL_FAMILIES.compound
+		}
+
+		// Default fallback to Llama 3.3 behavior
+		return MODEL_FAMILIES["kimi-k2"]
+	}
+
+	/**
+	 * Gets the optimal max_tokens based on model family and capabilities
+	 */
+	private getOptimalMaxTokens(model: { id: string; info: ModelInfo }, modelFamily: GroqModelFamily): number {
+		// Use model-specific max tokens if available
+		if (model.info.maxTokens && model.info.maxTokens > 0) {
+			return model.info.maxTokens
+		}
+
+		// Use family override if available
+		if (modelFamily.maxTokensOverride) {
+			return modelFamily.maxTokensOverride
+		}
+
+		// Default fallback
+		return 8192
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+		const modelFamily = this.detectModelFamily(model.id)
+
+		// Optimize parameters based on model family
+		const temperature = 0
+		const maxTokens = this.getOptimalMaxTokens(model, modelFamily)
+
+		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+
+		// Build request parameters with model-specific optimizations
+		const requestParams: OpenAI.Chat.ChatCompletionCreateParamsStreaming & {
+			reasoning_format?: "parsed" | "raw" | "hidden"
+			top_p?: number
+		} = {
+			model: model.id,
+			max_tokens: maxTokens,
+			messages: openAiMessages,
+			stream: true,
+			stream_options: { include_usage: true },
+			temperature,
+			...getOpenAIToolParams(tools),
+		}
+
+		// Add any special parameters for specific model families
+		if (modelFamily.specialParams) {
+			Object.assign(requestParams, modelFamily.specialParams)
+		}
+
+		const toolCallProcessor = new ToolCallProcessor()
+		const stream = await client.chat.completions.create(requestParams)
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+
+			// Handle reasoning field if present (for reasoning models with parsed output)
+			if ((delta as any)?.reasoning) {
+				const reasoningContent = (delta as any).reasoning as string
+				yield {
+					type: "reasoning",
+					reasoning: reasoningContent,
+				}
+				continue
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			// Handle content field - trust the parsed output from Groq
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			// Handle usage information
+			if (chunk.usage) {
+				yield* this.yieldUsage(model.info, chunk.usage)
+			}
+		}
+	}
+
+	/**
+	 * Checks if the current model supports vision/images
+	 */
+	supportsImages(): boolean {
+		const model = this.getModel()
+		return model.info.supportsImages === true
+	}
+
+	/**
+	 * Checks if the current model supports tools
+	 */
+	supportsTools(): boolean {
+		const model = this.getModel()
+		const modelFamily = this.detectModelFamily(model.id)
+		return modelFamily.supportedFeatures.tools
+	}
+
+	/**
+	 * Gets model information with enhanced family detection
+	 */
+	getModel(): { id: string; info: ModelInfo } {
+		// First priority: groqModelId and groqModelInfo (like Requesty does)
+		const groqModelId = this.options.groqModelId
+		const groqModelInfo = this.options.groqModelInfo
+		if (groqModelId && groqModelInfo) {
+			return { id: groqModelId, info: groqModelInfo }
+		}
+
+		// Second priority: groqModelId with static model info
+		if (groqModelId && groqModelId in groqModels) {
+			const id = groqModelId as GroqModelId
+			return { id, info: groqModels[id] }
+		}
+
+		// Third priority: apiModelId (for backward compatibility)
+		const apiModelId = this.options.apiModelId
+		if (apiModelId && apiModelId in groqModels) {
+			const id = apiModelId as GroqModelId
+			return { id, info: groqModels[id] }
+		}
+
+		// Default fallback
+		return {
+			id: groqDefaultModelId,
+			info: groqModels[groqDefaultModelId],
+		}
+	}
+
+	/**
+	 * Gets model family information for debugging/introspection
+	 */
+	getModelFamily(): GroqModelFamily {
+		const model = this.getModel()
+		return this.detectModelFamily(model.id)
+	}
+}

apps/vscode/src/core/api/providers/hicap.ts

@@ -0,0 +1,104 @@
+import { hicapModelInfoSaneDefaults, ModelInfo } from "@shared/api"
+import OpenAI from "openai"
+import type { ChatCompletionReasoningEffort } from "openai/resources/chat/completions"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../index"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+
+interface OpenAiHandlerOptions extends CommonApiHandlerOptions {
+	hicapApiKey?: string
+	hicapModelId?: string
+}
+
+export class HicapHandler implements ApiHandler {
+	private options: OpenAiHandlerOptions
+	private client: OpenAI | undefined
+
+	constructor(options: OpenAiHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.hicapApiKey) {
+				throw new Error("Hicap API key is required")
+			}
+			if (!this.options.hicapModelId) {
+				throw new Error("Model ID is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: "https://api.hicap.ai/v2/openai",
+					apiKey: this.options.hicapApiKey,
+					defaultHeaders: {
+						"api-key": this.options.hicapApiKey,
+					},
+				})
+			} catch (error: any) {
+				throw new Error(`Error creating OpenAI client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[]): ApiStream {
+		const client = this.ensureClient()
+		const modelId = this.options.hicapModelId ?? ""
+
+		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+		const temperature: number = 1
+		let reasoningEffort: ChatCompletionReasoningEffort | undefined
+		let maxTokens: number | undefined
+
+		const stream = await client.chat.completions.create({
+			model: modelId,
+			messages: openAiMessages,
+			temperature,
+			max_tokens: maxTokens,
+			reasoning_effort: reasoningEffort,
+			stream: true,
+			stream_options: { include_usage: true },
+		})
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta && "reasoning_content" in delta && delta.reasoning_content) {
+				yield {
+					type: "reasoning",
+					reasoning: (delta.reasoning_content as string | undefined) || "",
+				}
+			}
+
+			if (chunk.usage) {
+				yield {
+					type: "usage",
+					inputTokens: chunk.usage.prompt_tokens || 0,
+					outputTokens: chunk.usage.completion_tokens || 0,
+					cacheReadTokens: chunk.usage.prompt_tokens_details?.cached_tokens || 0,
+					// @ts-expect-error-next-line
+					cacheWriteTokens: chunk.usage.prompt_cache_miss_tokens || 0,
+				}
+			}
+		}
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		return {
+			id: this.options.hicapModelId ?? "",
+			info: hicapModelInfoSaneDefaults,
+		}
+	}
+}

apps/vscode/src/core/api/providers/huawei-cloud-maas.ts

@@ -0,0 +1,142 @@
+import { HuaweiCloudMaasModelId, huaweiCloudMaasDefaultModelId, huaweiCloudMaasModels, ModelInfo } from "@shared/api"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from ".."
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface HuaweiCloudMaaSHandlerOptions extends CommonApiHandlerOptions {
+	huaweiCloudMaasApiKey?: string
+	huaweiCloudMaasModelId?: string
+	huaweiCloudMaasModelInfo?: ModelInfo
+}
+
+export class HuaweiCloudMaaSHandler implements ApiHandler {
+	private options: HuaweiCloudMaaSHandlerOptions
+	private client: OpenAI | undefined
+	constructor(options: HuaweiCloudMaaSHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.huaweiCloudMaasApiKey) {
+				throw new Error("Huawei Cloud MaaS API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: "https://api.modelarts-maas.com/v1/",
+					apiKey: this.options.huaweiCloudMaasApiKey,
+				})
+			} catch (error) {
+				throw new Error(`Error creating Huawei Cloud MaaS client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	getModel(): { id: HuaweiCloudMaasModelId; info: ModelInfo } {
+		// First priority: huaweiCloudMaasModelId and huaweiCloudMaasModelInfo (like Groq does)
+		const huaweiCloudMaasModelId = this.options.huaweiCloudMaasModelId
+		const huaweiCloudMaasModelInfo = this.options.huaweiCloudMaasModelInfo
+		if (huaweiCloudMaasModelId && huaweiCloudMaasModelInfo) {
+			return { id: huaweiCloudMaasModelId as HuaweiCloudMaasModelId, info: huaweiCloudMaasModelInfo }
+		}
+
+		// Second priority: huaweiCloudMaasModelId with static model info
+		if (huaweiCloudMaasModelId && huaweiCloudMaasModelId in huaweiCloudMaasModels) {
+			const id = huaweiCloudMaasModelId as HuaweiCloudMaasModelId
+			return { id, info: huaweiCloudMaasModels[id] }
+		}
+
+		// Default fallback
+		return {
+			id: huaweiCloudMaasDefaultModelId,
+			info: huaweiCloudMaasModels[huaweiCloudMaasDefaultModelId],
+		}
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+		const stream = await client.chat.completions.create({
+			model: model.id,
+			max_completion_tokens: model.info.maxTokens,
+			messages: openAiMessages,
+			stream: true,
+			stream_options: { include_usage: true },
+			temperature: 0,
+			...getOpenAIToolParams(tools),
+		})
+
+		let reasoning: string | null = null
+		let didOutputUsage: boolean = false
+		let finalUsage: any = null
+
+		const toolCallProcessor = new ToolCallProcessor()
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+
+			// Handle reasoning content detection
+			if (delta?.content) {
+				if (reasoning || delta.content.includes("<think>")) {
+					reasoning = (reasoning || "") + delta.content
+				} else if (!reasoning) {
+					yield {
+						type: "text",
+						text: delta.content,
+					}
+				}
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			// Handle reasoning output
+			if (reasoning || (delta && "reasoning_content" in delta && delta.reasoning_content)) {
+				const reasoningContent = delta?.content || ((delta as any)?.reasoning_content as string | undefined) || ""
+				if (reasoningContent.trim()) {
+					yield {
+						type: "reasoning",
+						reasoning: reasoningContent,
+					}
+				}
+
+				// Check if reasoning is complete
+				if (reasoning?.includes("</think>")) {
+					reasoning = null
+				}
+			}
+
+			// Store usage information for later output
+			if (chunk.usage) {
+				finalUsage = chunk.usage
+			}
+
+			// Output usage when stream is finished
+			if (!didOutputUsage && chunk.choices?.[0]?.finish_reason) {
+				if (finalUsage) {
+					yield {
+						type: "usage",
+						inputTokens: finalUsage.prompt_tokens || 0,
+						outputTokens: finalUsage.completion_tokens || 0,
+						cacheWriteTokens: 0,
+						cacheReadTokens: 0,
+					}
+				}
+				didOutputUsage = true
+			}
+		}
+	}
+}

apps/vscode/src/core/api/providers/huggingface.ts

@@ -0,0 +1,148 @@
+import { HuggingFaceModelId, huggingFaceDefaultModelId, huggingFaceModels, ModelInfo } from "@shared/api"
+import { calculateApiCostOpenAI } from "@utils/cost"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface HuggingFaceHandlerOptions extends CommonApiHandlerOptions {
+	huggingFaceApiKey?: string
+	huggingFaceModelId?: string
+	huggingFaceModelInfo?: ModelInfo
+}
+
+export class HuggingFaceHandler implements ApiHandler {
+	private options: HuggingFaceHandlerOptions
+	private client: OpenAI | undefined
+	private cachedModel: { id: HuggingFaceModelId; info: ModelInfo } | undefined
+
+	constructor(options: HuggingFaceHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.huggingFaceApiKey) {
+				throw new Error("Hugging Face API key is required")
+			}
+
+			try {
+				this.client = createOpenAIClient({
+					baseURL: "https://router.huggingface.co/v1",
+					apiKey: this.options.huggingFaceApiKey,
+				})
+			} catch (error: any) {
+				throw new Error(`Error creating Hugging Face client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	private async *yieldUsage(info: ModelInfo, usage: OpenAI.Completions.CompletionUsage | undefined): ApiStream {
+		if (!usage) {
+			return
+		}
+
+		const inputTokens = usage.prompt_tokens || 0
+		const outputTokens = usage.completion_tokens || 0
+		const totalCost = calculateApiCostOpenAI(info, inputTokens, outputTokens)
+
+		const usageData = {
+			type: "usage" as const,
+			inputTokens: inputTokens,
+			outputTokens: outputTokens,
+			cacheWriteTokens: 0,
+			cacheReadTokens: 0,
+			totalCost: totalCost,
+		}
+
+		yield usageData
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		try {
+			const client = this.ensureClient()
+			const model = this.getModel()
+
+			const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+				{ role: "system", content: systemPrompt },
+				...convertToOpenAiMessages(messages),
+			]
+
+			const requestParams = {
+				model: model.id,
+				max_tokens: model.info.maxTokens,
+				messages: openAiMessages,
+				stream: true,
+				stream_options: { include_usage: true },
+				temperature: 0,
+				...getOpenAIToolParams(tools),
+			}
+
+			const toolCallProcessor = new ToolCallProcessor()
+			const stream = (await client.chat.completions.create(requestParams)) as any
+
+			let _chunkCount = 0
+			let _totalContent = ""
+
+			for await (const chunk of stream) {
+				_chunkCount++
+				const delta = chunk.choices?.[0]?.delta
+				if (delta?.content) {
+					_totalContent += delta.content
+
+					yield {
+						type: "text",
+						text: delta.content,
+					}
+				}
+
+				if (delta?.tool_calls) {
+					yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+				}
+
+				if (chunk.usage) {
+					yield* this.yieldUsage(model.info, chunk.usage)
+				}
+			}
+		} catch (error: any) {
+			throw error
+		}
+	}
+
+	getModel(): { id: HuggingFaceModelId; info: ModelInfo } {
+		// Return cached model if available
+		if (this.cachedModel) {
+			return this.cachedModel
+		}
+
+		const modelId = this.options.huggingFaceModelId
+
+		// List all available models for debugging
+		const _availableModels = Object.keys(huggingFaceModels)
+		let result: { id: HuggingFaceModelId; info: ModelInfo }
+
+		if (modelId && modelId in huggingFaceModels) {
+			const id = modelId as HuggingFaceModelId
+			const modelInfo = huggingFaceModels[id]
+			result = { id, info: modelInfo }
+		} else {
+			const defaultInfo = huggingFaceModels[huggingFaceDefaultModelId]
+			result = {
+				id: huggingFaceDefaultModelId,
+				info: defaultInfo,
+			}
+		}
+
+		// Cache the result for future calls
+		this.cachedModel = result
+
+		return result
+	}
+}

apps/vscode/src/core/api/providers/litellm.ts

@@ -0,0 +1,403 @@
+import { Anthropic } from "@anthropic-ai/sdk"
+import { LiteLLMModelInfo, liteLlmDefaultModelId, liteLlmModelInfoSaneDefaults } from "@shared/api"
+import { isClaudeOpusAdaptiveThinkingModel, resolveClaudeOpusAdaptiveThinking } from "@shared/utils/reasoning-support"
+import OpenAI from "openai"
+import { StateManager } from "@/core/storage/StateManager"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient, fetch } from "@/shared/net"
+import { Logger } from "@/shared/services/Logger"
+import { isAnthropicModelId } from "@/utils/model-utils"
+import { ApiHandler, CommonApiHandlerOptions } from ".."
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+
+interface LiteLlmHandlerOptions extends CommonApiHandlerOptions {
+	liteLlmApiKey?: string
+	liteLlmBaseUrl?: string
+	liteLlmModelId?: string
+	liteLlmModelInfo?: LiteLLMModelInfo
+	reasoningEffort?: string
+	thinkingBudgetTokens?: number
+	liteLlmUsePromptCache?: boolean
+	ulid?: string
+}
+
+/**
+ * Extended chat completion parameters that include LiteLLM-specific options
+ * not present in the standard OpenAI SDK types
+ */
+interface LiteLlmChatCompletionCreateParams extends OpenAI.Chat.ChatCompletionCreateParamsStreaming {
+	drop_params?: boolean
+}
+
+export interface LiteLlmModelInfoResponse {
+	data: Array<{
+		model_name: string
+		litellm_params: {
+			model: string
+			[key: string]: any
+		}
+		model_info: {
+			input_cost_per_token: number
+			output_cost_per_token: number
+			cache_creation_input_token_cost?: number
+			cache_read_input_token_cost?: number
+			supports_prompt_caching?: boolean
+			[key: string]: any
+		}
+	}>
+}
+
+/**
+ * Exported utility function to fetch LiteLLM model info
+ * @param baseUrl The base URL for the LiteLLM API
+ * @param apiKey The API key for authentication
+ * @returns The model info response or undefined if fetch fails
+ */
+export async function fetchLiteLlmModelsInfo(baseUrl: string, apiKey: string): Promise<LiteLlmModelInfoResponse | undefined> {
+	// Handle base URLs that already include /v1 to avoid double /v1/v1/
+	const normalizedBaseUrl = baseUrl.endsWith("/v1") ? baseUrl : `${baseUrl}/v1`
+	const url = `${normalizedBaseUrl}/model/info`
+
+	try {
+		const response = await fetch(url, {
+			method: "GET",
+			headers: {
+				accept: "application/json",
+				"x-litellm-api-key": apiKey,
+				...buildExternalBasicHeaders(),
+			},
+		})
+
+		if (response.ok) {
+			const data: LiteLlmModelInfoResponse = await response.json()
+			return data
+		}
+		Logger.error("Failed to fetch LiteLLM model info:", response.statusText)
+		// Try with Authorization header instead
+		const retryResponse = await fetch(url, {
+			method: "GET",
+			headers: {
+				accept: "application/json",
+				Authorization: `Bearer ${apiKey}`,
+				...buildExternalBasicHeaders(),
+			},
+		})
+
+		if (retryResponse.ok) {
+			const data: LiteLlmModelInfoResponse = await retryResponse.json()
+			return data
+		}
+		Logger.error("Failed to fetch LiteLLM model info with Authorization header:", retryResponse.statusText)
+		throw new Error(`Failed to fetch LiteLLM model info: ${retryResponse.statusText}`)
+	} catch (error) {
+		Logger.error("Error fetching LiteLLM model info:", error)
+		throw error
+	}
+}
+
+export class LiteLlmHandler implements ApiHandler {
+	private options: LiteLlmHandlerOptions
+	private client: OpenAI | undefined
+	private modelInfoCache: LiteLlmModelInfoResponse | undefined
+	private modelInfoCacheTimestamp = 0
+	private readonly modelInfoCacheTTL = 5 * 60 * 1000 // 5 minutes
+
+	constructor(options: LiteLlmHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.liteLlmApiKey) {
+				throw new Error("LiteLLM API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: this.options.liteLlmBaseUrl || "http://localhost:4000",
+					apiKey: this.options.liteLlmApiKey || "noop",
+				})
+			} catch (error) {
+				throw new Error(`Error creating LiteLLM client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	private async modelInfo(publicModelName: string): Promise<LiteLlmModelInfoResponse["data"][number] | undefined> {
+		const modelInfo = await this.fetchModelsInfo()
+
+		if (!modelInfo?.data) {
+			return undefined
+		}
+
+		return modelInfo.data.find((model) => model.model_name === publicModelName)
+	}
+
+	private async fetchModelsInfo(): Promise<LiteLlmModelInfoResponse | undefined> {
+		// Check if cache is still valid
+		const now = Date.now()
+		if (this.modelInfoCache && now - this.modelInfoCacheTimestamp < this.modelInfoCacheTTL) {
+			return this.modelInfoCache
+		}
+
+		const client = this.ensureClient()
+		const data = await fetchLiteLlmModelsInfo(client.baseURL, this.options.liteLlmApiKey || "")
+
+		if (data) {
+			this.modelInfoCache = data
+			this.modelInfoCacheTimestamp = now
+		}
+
+		return data
+	}
+
+	private async getModelCostInfo(publicModelName: string): Promise<{
+		inputCostPerToken: number
+		outputCostPerToken: number
+		cacheCreationCostPerToken?: number
+		cacheReadCostPerToken?: number
+	}> {
+		try {
+			const matchingModel = await this.modelInfo(publicModelName)
+
+			if (matchingModel) {
+				return {
+					inputCostPerToken: matchingModel.model_info.input_cost_per_token || 0,
+					outputCostPerToken: matchingModel.model_info.output_cost_per_token || 0,
+					cacheCreationCostPerToken: matchingModel.model_info.cache_creation_input_token_cost,
+					cacheReadCostPerToken: matchingModel.model_info.cache_read_input_token_cost,
+				}
+			}
+		} catch (error) {
+			Logger.warn("Error getting LiteLLM model cost info:", error)
+		}
+
+		// Fallback to zero costs if we can't get the information
+		return {
+			inputCostPerToken: 0,
+			outputCostPerToken: 0,
+		}
+	}
+
+	async calculateCost(
+		prompt_tokens: number,
+		completion_tokens: number,
+		cache_creation_tokens?: number,
+		cache_read_tokens?: number,
+	): Promise<number | undefined> {
+		const publicModelId = this.options.liteLlmModelId || liteLlmDefaultModelId
+
+		try {
+			const costInfo = await this.getModelCostInfo(publicModelId)
+
+			// Calculate costs for different token types
+			const inputCost = Math.max(0, prompt_tokens - (cache_read_tokens || 0)) * costInfo.inputCostPerToken
+			const outputCost = completion_tokens * costInfo.outputCostPerToken
+			const cacheCreationCost = (cache_creation_tokens || 0) * (costInfo.cacheCreationCostPerToken || 0)
+			const cacheReadCost = (cache_read_tokens || 0) * (costInfo.cacheReadCostPerToken || 0)
+
+			const totalCost = inputCost + outputCost + cacheCreationCost + cacheReadCost
+
+			return totalCost
+		} catch (error) {
+			Logger.error("Error calculating spend:", error)
+			return undefined
+		}
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[]): ApiStream {
+		const client = this.ensureClient()
+
+		const formattedMessages = convertToOpenAiMessages(messages)
+		const systemMessage: OpenAI.Chat.ChatCompletionSystemMessageParam | Anthropic.Messages.TextBlockParam = {
+			role: "system",
+			content: systemPrompt,
+		}
+		const modelId = this.options.liteLlmModelId || liteLlmDefaultModelId
+		const isOminiModel = modelId.includes("o1-mini") || modelId.includes("o3-mini") || modelId.includes("o4-mini")
+		const isCodexModel = modelId.toLowerCase().includes("codex")
+
+		// Configuration for extended thinking
+		const budgetTokens = this.options.thinkingBudgetTokens || 0
+		const reasoningOn = budgetTokens !== 0
+		const isAdaptiveThinkingModel = isClaudeOpusAdaptiveThinkingModel(modelId)
+		const adaptiveThinking = isAdaptiveThinkingModel
+			? resolveClaudeOpusAdaptiveThinking(this.options.reasoningEffort, budgetTokens)
+			: undefined
+		const thinkingConfig = isAdaptiveThinkingModel
+			? adaptiveThinking?.enabled
+				? ({ type: "adaptive" } as any)
+				: undefined
+			: reasoningOn
+				? { type: "enabled", budget_tokens: budgetTokens }
+				: undefined
+
+		let temperature: number | undefined = this.options.liteLlmModelInfo?.temperature ?? 1
+
+		if (isAdaptiveThinkingModel) {
+			temperature = undefined
+		} else if ((isOminiModel || isAnthropicModelId(modelId)) && reasoningOn) {
+			temperature = undefined // OAI omni and Anthropic extended thinking mode doesn't support temperature
+		}
+
+		const modelInfo = await this.modelInfo(modelId)
+		// Automatically enable caching if the model supports it
+		const cacheControl =
+			(modelInfo?.model_info.supports_prompt_caching ?? false) ? { cache_control: { type: "ephemeral" } } : undefined
+
+		if (cacheControl) {
+			// Add cache_control to system message if enabled
+			// https://docs.litellm.ai/docs/providers/anthropic#caching---large-context-caching
+			systemMessage.content = [
+				{
+					text: systemPrompt,
+					type: "text",
+					...cacheControl,
+				},
+			] as Anthropic.Messages.TextBlockParam[]
+		}
+
+		// Find the last two user messages to apply caching
+		const userMsgIndices = formattedMessages.reduce(
+			(acc, msg, index) => (msg.role === "user" ? [...acc, index] : acc),
+			[] as number[],
+		)
+		const lastUserMsgIndex = userMsgIndices[userMsgIndices.length - 1] ?? -1
+		const secondLastUserMsgIndex = userMsgIndices[userMsgIndices.length - 2] ?? -1
+
+		// Apply cache_control to the last two user messages if enabled
+		// https://docs.litellm.ai/docs/providers/anthropic#caching---large-context-caching
+		const enhancedMessages: OpenAI.Chat.ChatCompletionMessageParam[] = formattedMessages.map(
+			(message, index): OpenAI.Chat.ChatCompletionMessageParam => {
+				if ((index === lastUserMsgIndex || index === secondLastUserMsgIndex) && cacheControl) {
+					// Handle both string and array content types
+					if (typeof message.content === "string") {
+						return {
+							...message,
+							content: [
+								{
+									type: "text",
+									text: message.content,
+									...cacheControl,
+								},
+							] as any,
+						}
+					}
+					if (Array.isArray(message.content)) {
+						// Apply cache control to the last content item in the array
+						return {
+							...message,
+							content: message.content.map((item, contentIndex) =>
+								contentIndex === (message.content?.length || 0) - 1
+									? {
+											...item,
+											...cacheControl,
+										}
+									: item,
+							) as any,
+						}
+					}
+
+					return {
+						...message,
+						...cacheControl,
+					}
+				}
+				return message
+			},
+		)
+
+		const stream = await client.chat.completions.create({
+			model: this.options.liteLlmModelId || liteLlmDefaultModelId,
+			messages: [systemMessage, ...enhancedMessages],
+			temperature,
+			stream: true,
+			drop_params: true,
+			...(!isCodexModel && { stream_options: { include_usage: true } }), // Codex models are only on the responses api, which doesn't take the stream_options parameter. we will need to migrate to the responses api for this to work
+			...(thinkingConfig && { thinking: thinkingConfig }), // Add thinking configuration when applicable
+			...(isAdaptiveThinkingModel && adaptiveThinking?.effort
+				? { output_config: { effort: adaptiveThinking.effort } }
+				: {}),
+			...(this.options.ulid && { litellm_session_id: `cline-${this.options.ulid}` }), // Add session ID for LiteLLM tracking
+		} as LiteLlmChatCompletionCreateParams)
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+
+			// Handle normal text content
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			// Handle reasoning events
+			// This is not in the standard types but may be in the response
+			interface ThinkingDelta {
+				reasoning_content?: string
+			}
+
+			if ((delta as ThinkingDelta)?.reasoning_content) {
+				yield {
+					type: "reasoning",
+					reasoning: (delta as ThinkingDelta).reasoning_content || "",
+				}
+			}
+
+			// Handle token usage information
+			if (chunk.usage) {
+				// Extract cache-related information if available
+				// Need to use type assertion since these properties are not in the standard OpenAI types
+				const usage = chunk.usage as {
+					prompt_tokens: number
+					completion_tokens: number
+					cache_creation_input_tokens?: number
+					prompt_cache_miss_tokens?: number
+					cache_read_input_tokens?: number
+					prompt_cache_hit_tokens?: number
+				}
+
+				const cacheWriteTokens = usage.cache_creation_input_tokens || usage.prompt_cache_miss_tokens || 0
+				const cacheReadTokens = usage.cache_read_input_tokens || usage.prompt_cache_hit_tokens || 0
+
+				// Calculate cost using the actual token usage including cache tokens
+				const totalCost =
+					(await this.calculateCost(
+						usage.prompt_tokens || 0,
+						usage.completion_tokens || 0,
+						cacheWriteTokens > 0 ? cacheWriteTokens : undefined,
+						cacheReadTokens > 0 ? cacheReadTokens : undefined,
+					)) || 0
+
+				yield {
+					type: "usage",
+					inputTokens: usage.prompt_tokens || 0,
+					outputTokens: usage.completion_tokens || 0,
+					cacheWriteTokens: cacheWriteTokens > 0 ? cacheWriteTokens : undefined,
+					cacheReadTokens: cacheReadTokens > 0 ? cacheReadTokens : undefined,
+					totalCost,
+				}
+			}
+		}
+	}
+
+	getModel() {
+		const modelId = this.options.liteLlmModelId || liteLlmDefaultModelId
+
+		// Try to get model info from StateManager cache first
+		const cachedModelInfo = StateManager.get().getModelInfo("liteLlm", modelId)
+
+		// Fall back to provided model info or defaults if not in cache
+		const modelInfo = cachedModelInfo || this.options.liteLlmModelInfo || liteLlmModelInfoSaneDefaults
+
+		return {
+			id: modelId,
+			info: modelInfo,
+		}
+	}
+}

apps/vscode/src/core/api/providers/lmstudio.ts

@@ -0,0 +1,109 @@
+import { type ModelInfo, openAiModelInfoSaneDefaults } from "@shared/api"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import type { ApiHandler, CommonApiHandlerOptions } from "../"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import type { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface LmStudioHandlerOptions extends CommonApiHandlerOptions {
+	lmStudioBaseUrl?: string
+	lmStudioModelId?: string
+	lmStudioMaxTokens?: string
+}
+
+export class LmStudioHandler implements ApiHandler {
+	private options: LmStudioHandlerOptions
+	private client: OpenAI | undefined
+
+	constructor(options: LmStudioHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			try {
+				this.client = createOpenAIClient({
+					// Docs on the new v0 api endpoint: https://lmstudio.ai/docs/app/api/endpoints/rest
+					baseURL: new URL("api/v0", this.options.lmStudioBaseUrl || "http://localhost:1234").toString(),
+					apiKey: "noop",
+				})
+			} catch (error) {
+				throw new Error(`Error creating LM Studio client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry({ retryAllErrors: true })
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+
+		try {
+			const stream = await client.chat.completions.create({
+				model: this.getModel().id,
+				messages: openAiMessages,
+				stream: true,
+				stream_options: { include_usage: true },
+				max_completion_tokens: this.options.lmStudioMaxTokens ? Number(this.options.lmStudioMaxTokens) : undefined,
+				...getOpenAIToolParams(tools),
+			})
+
+			const toolCallProcessor = new ToolCallProcessor()
+
+			for await (const chunk of stream) {
+				const choice = chunk.choices?.[0]
+				const delta = choice?.delta
+				if (delta?.content) {
+					yield {
+						type: "text",
+						text: delta.content,
+					}
+				}
+				if (delta && "reasoning_content" in delta && delta.reasoning_content) {
+					yield {
+						type: "reasoning",
+						reasoning: (delta.reasoning_content as string | undefined) || "",
+					}
+				}
+
+				if (delta?.tool_calls) {
+					yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+				}
+
+				if (chunk.usage) {
+					yield {
+						type: "usage",
+						inputTokens: chunk.usage.prompt_tokens || 0,
+						outputTokens: chunk.usage.completion_tokens || 0,
+						cacheReadTokens: chunk.usage.prompt_tokens_details?.cached_tokens || 0,
+					}
+				}
+			}
+		} catch {
+			// LM Studio doesn't return an error code/body for now
+			throw new Error(
+				"Please check the LM Studio developer logs to debug what went wrong. You may need to load the model with a larger context length to work with Cline's prompts. Alternatively, try enabling Compact Prompt in your settings when working with a limited context window.",
+			)
+		}
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		const info = { ...openAiModelInfoSaneDefaults }
+		const maxTokens = Number(this.options.lmStudioMaxTokens)
+		if (!Number.isNaN(maxTokens)) {
+			info.contextWindow = maxTokens
+		}
+		return {
+			id: this.options.lmStudioModelId || "",
+			info,
+		}
+	}
+}

apps/vscode/src/core/api/providers/minimax.ts

@@ -0,0 +1,208 @@
+import { Anthropic } from "@anthropic-ai/sdk"
+import { Tool as AnthropicTool } from "@anthropic-ai/sdk/resources/index"
+import { Stream as AnthropicStream } from "@anthropic-ai/sdk/streaming"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { MinimaxModelId, ModelInfo, minimaxDefaultModelId, minimaxModels } from "@/shared/api"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { fetch } from "@/shared/net"
+import { ClineTool } from "@/shared/tools"
+import { ApiHandler, CommonApiHandlerOptions } from "../index"
+import { withRetry } from "../retry"
+import { ApiStream } from "../transform/stream"
+
+interface MinimaxHandlerOptions extends CommonApiHandlerOptions {
+	minimaxApiKey?: string
+	minimaxApiLine?: string
+	apiModelId?: string
+	thinkingBudgetTokens?: number
+}
+
+export class MinimaxHandler implements ApiHandler {
+	private options: MinimaxHandlerOptions
+	private client: Anthropic | undefined
+
+	constructor(options: MinimaxHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): Anthropic {
+		if (!this.client) {
+			if (!this.options.minimaxApiKey) {
+				throw new Error("MiniMax API key is required")
+			}
+			try {
+				const externalHeaders = buildExternalBasicHeaders()
+				this.client = new Anthropic({
+					apiKey: this.options.minimaxApiKey,
+					baseURL:
+						this.options.minimaxApiLine === "china"
+							? "https://api.minimaxi.com/anthropic"
+							: "https://api.minimax.io/anthropic",
+					defaultHeaders: externalHeaders,
+					fetch, // Use configured fetch with proxy support
+				})
+			} catch (error) {
+				throw new Error(`Error creating MiniMax client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: ClineTool[]): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+
+		// Tools are available only when native tools are enabled
+		const nativeToolsOn = tools?.length && tools?.length > 0
+
+		const budget_tokens = this.options.thinkingBudgetTokens || 0
+		const reasoningOn = (model.info.supportsReasoning ?? false) && budget_tokens !== 0
+
+		// MiniMax M2 uses Anthropic API format
+		const stream: AnthropicStream<Anthropic.RawMessageStreamEvent> = await client.messages.create({
+			model: model.id,
+			max_tokens: model.info.maxTokens || 8192,
+			system: [{ text: systemPrompt, type: "text" }],
+			messages,
+			stream: true,
+			tools: nativeToolsOn ? (tools as AnthropicTool[]) : undefined,
+			thinking: reasoningOn ? { type: "enabled", budget_tokens: budget_tokens } : undefined,
+			// "Thinking isn't compatible with temperature, top_p, or top_k modifications"
+			temperature: reasoningOn ? undefined : 1.0, // MiniMax recommends 1.0, range is (0.0, 1.0]
+			// NOTE: Forcing tool use when tools are provided will result in error when thinking is also enabled.
+			tool_choice: nativeToolsOn && !reasoningOn ? { type: "any" } : undefined,
+		})
+
+		const lastStartedToolCall = { id: "", name: "", arguments: "" }
+
+		for await (const chunk of stream) {
+			switch (chunk?.type) {
+				case "message_start": {
+					// tells us cache reads/writes/input/output
+					const usage = chunk.message.usage
+					yield {
+						type: "usage",
+						inputTokens: usage.input_tokens || 0,
+						outputTokens: usage.output_tokens || 0,
+						cacheWriteTokens: usage.cache_creation_input_tokens || undefined,
+						cacheReadTokens: usage.cache_read_input_tokens || undefined,
+					}
+					break
+				}
+				case "message_delta":
+					// tells us stop_reason, stop_sequence, and output tokens along the way and at the end of the message
+					yield {
+						type: "usage",
+						inputTokens: 0,
+						outputTokens: chunk.usage.output_tokens || 0,
+					}
+					break
+				case "message_stop":
+					// no usage data, just an indicator that the message is done
+					break
+				case "content_block_start":
+					switch (chunk.content_block.type) {
+						case "thinking":
+							yield {
+								type: "reasoning",
+								reasoning: chunk.content_block.thinking || "",
+								signature: chunk.content_block.signature,
+							}
+							break
+						case "redacted_thinking":
+							// Content is encrypted, and we don't want to pass placeholder text back to the API
+							yield {
+								type: "reasoning",
+								reasoning: "[Redacted thinking block]",
+								redacted_data: chunk.content_block.data,
+							}
+							break
+						case "tool_use":
+							if (chunk.content_block.id && chunk.content_block.name) {
+								// Store tool call information for streaming
+								lastStartedToolCall.id = chunk.content_block.id
+								lastStartedToolCall.name = chunk.content_block.name
+								lastStartedToolCall.arguments = ""
+							}
+							break
+						case "text":
+							// we may receive multiple text blocks, in which case just insert a line break between them
+							if (chunk.index > 0) {
+								yield {
+									type: "text",
+									text: "\n",
+								}
+							}
+							yield {
+								type: "text",
+								text: chunk.content_block.text,
+							}
+							break
+					}
+					break
+				case "content_block_delta":
+					switch (chunk.delta.type) {
+						case "thinking_delta":
+							// 'reasoning' type just displays in the UI, but reasoning with signature will be used to send the thinking traces back to the API
+							yield {
+								type: "reasoning",
+								reasoning: chunk.delta.thinking,
+							}
+							break
+						case "signature_delta":
+							// It's used when sending the thinking block back to the API
+							// API expects this in completed form, not as array of deltas
+							if (chunk.delta.signature) {
+								yield {
+									type: "reasoning",
+									reasoning: "",
+									signature: chunk.delta.signature,
+								}
+							}
+							break
+						case "text_delta":
+							yield {
+								type: "text",
+								text: chunk.delta.text,
+							}
+							break
+						case "input_json_delta":
+							if (lastStartedToolCall.id && lastStartedToolCall.name && chunk.delta.partial_json) {
+								// Convert Anthropic tool_use to OpenAI-compatible format for internal processing
+								yield {
+									type: "tool_calls",
+									tool_call: {
+										...lastStartedToolCall,
+										function: {
+											...lastStartedToolCall,
+											id: lastStartedToolCall.id,
+											name: lastStartedToolCall.name,
+											arguments: chunk.delta.partial_json,
+										},
+									},
+								}
+							}
+							break
+					}
+					break
+
+				case "content_block_stop":
+					lastStartedToolCall.id = ""
+					lastStartedToolCall.name = ""
+					lastStartedToolCall.arguments = ""
+					break
+			}
+		}
+	}
+
+	getModel(): { id: MinimaxModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+
+		if (modelId && modelId in minimaxModels) {
+			const id = modelId as MinimaxModelId
+			return { id, info: minimaxModels[id] }
+		}
+		return { id: minimaxDefaultModelId, info: minimaxModels[minimaxDefaultModelId] }
+	}
+}

apps/vscode/src/core/api/providers/mistral.ts

@@ -0,0 +1,154 @@
+import { Mistral } from "@mistralai/mistralai"
+import { HTTPClient } from "@mistralai/mistralai/lib/http"
+import { Tool as MistralTool } from "@mistralai/mistralai/models/components/tool"
+import { MistralModelId, ModelInfo, mistralDefaultModelId, mistralModels } from "@shared/api"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { fetch } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../"
+import { withRetry } from "../retry"
+import { convertToMistralMessages } from "../transform/mistral-format"
+import { ApiStream } from "../transform/stream"
+
+interface MistralHandlerOptions extends CommonApiHandlerOptions {
+	mistralApiKey?: string
+	apiModelId?: string
+}
+
+export class MistralHandler implements ApiHandler {
+	private options: MistralHandlerOptions
+	private client: Mistral | undefined
+
+	constructor(options: MistralHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): Mistral {
+		if (!this.client) {
+			if (!this.options.mistralApiKey) {
+				throw new Error("Mistral API key is required")
+			}
+			try {
+				const externalHeaders = buildExternalBasicHeaders()
+				// Create HTTP client with custom fetch for proxy support
+				// The Mistral SDK's HTTPClient passes a Request object to the fetcher,
+				// but we need to extract the URL and init options to pass to our fetch wrapper
+				// which properly handles proxy configuration in standalone mode (JetBrains/CLI)
+				const httpClient = new HTTPClient({
+					fetcher: async (input: RequestInfo | URL, init?: RequestInit) => {
+						// Handle both string/URL and Request object inputs
+						if (input instanceof Request) {
+							Object.keys(externalHeaders).forEach((key) => {
+								if (!input.headers.has(key)) {
+									input.headers.set(key, externalHeaders[key])
+								}
+							})
+							return fetch(input.url, {
+								method: input.method,
+								headers: input.headers,
+								body: input.body,
+								redirect: input.redirect,
+								signal: input.signal,
+								// duplex is required when sending a body stream in Node.js/undici
+								duplex: input.body ? "half" : undefined,
+								...init,
+							} as RequestInit)
+						}
+
+						// Merge external headers with existing headers
+						const mergedInit = {
+							...init,
+							headers: {
+								...externalHeaders,
+								...(init?.headers || {}),
+							},
+						}
+						return fetch(input, mergedInit)
+					},
+				})
+
+				this.client = new Mistral({
+					apiKey: this.options.mistralApiKey,
+					httpClient,
+				})
+			} catch (error) {
+				throw new Error(`Error creating Mistral client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const stream = await client.chat
+			.stream({
+				model: this.getModel().id,
+				// max_completion_tokens: this.getModel().info.maxTokens,
+				temperature: 0,
+				messages: [{ role: "system", content: systemPrompt }, ...convertToMistralMessages(messages)],
+				stream: true,
+				tools: tools?.length ? (tools as MistralTool[]) : undefined,
+				toolChoice: tools?.length ? "any" : undefined,
+			})
+			.catch((err) => {
+				// The Mistal SDK uses statusCode instead of status
+				// However, if they introduce status for something, I don't want to override it
+				if ("statusCode" in err && !("status" in err)) {
+					err.status = err.statusCode
+				}
+
+				throw err
+			})
+
+		for await (const chunk of stream) {
+			const delta = chunk.data.choices[0]?.delta
+			if (delta.toolCalls) {
+				for (const toolCall of delta.toolCalls) {
+					yield {
+						type: "tool_calls",
+						tool_call: {
+							function: {
+								id: toolCall.id,
+								name: toolCall.function.name,
+								arguments: JSON.stringify(toolCall.function.arguments),
+							},
+						},
+					}
+				}
+			} else if (delta?.content) {
+				let content: string = ""
+				if (typeof delta.content === "string") {
+					content = delta.content
+				} else if (Array.isArray(delta.content)) {
+					content = delta.content.map((c) => (c.type === "text" ? c.text : "")).join("")
+				}
+				yield {
+					type: "text",
+					text: content,
+				}
+			}
+
+			if (chunk.data.usage) {
+				yield {
+					type: "usage",
+					inputTokens: chunk.data.usage.promptTokens || 0,
+					outputTokens: chunk.data.usage.completionTokens || 0,
+				}
+			}
+		}
+	}
+
+	getModel(): { id: MistralModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+		if (modelId && modelId in mistralModels) {
+			const id = modelId as MistralModelId
+			return { id, info: mistralModels[id] }
+		}
+		return {
+			id: mistralDefaultModelId,
+			info: mistralModels[mistralDefaultModelId],
+		}
+	}
+}

apps/vscode/src/core/api/providers/moonshot.ts

@@ -0,0 +1,110 @@
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { ModelInfo, MoonshotModelId, moonshotDefaultModelId, moonshotModels } from "@/shared/api"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../index"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface MoonshotHandlerOptions extends CommonApiHandlerOptions {
+	moonshotApiKey?: string
+	moonshotApiLine?: string
+	apiModelId?: string
+}
+
+// Enhanced usage interface to support Moonshot's cached token field
+interface MoonshotUsage extends OpenAI.CompletionUsage {
+	cached_tokens?: number
+}
+
+export class MoonshotHandler implements ApiHandler {
+	private client: OpenAI | undefined
+
+	constructor(private readonly options: MoonshotHandlerOptions) {}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.moonshotApiKey) {
+				throw new Error("Moonshot API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL:
+						this.options.moonshotApiLine === "china" ? "https://api.moonshot.cn/v1" : "https://api.moonshot.ai/v1",
+					apiKey: this.options.moonshotApiKey,
+				})
+			} catch (error) {
+				throw new Error(`Error creating Moonshot client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+
+		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+
+		const stream = await client.chat.completions.create({
+			model: model.id,
+			messages: openAiMessages,
+			temperature: model.info.temperature,
+			max_tokens: model.info.maxTokens,
+			stream: true,
+			stream_options: { include_usage: true },
+			...getOpenAIToolParams(tools),
+		})
+
+		const toolCallProcessor = new ToolCallProcessor()
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			if (delta && "reasoning_content" in delta && delta.reasoning_content) {
+				yield {
+					type: "reasoning",
+					reasoning: (delta.reasoning_content as string | undefined) || "",
+				}
+			}
+
+			if (chunk.usage) {
+				const usage = chunk.usage as MoonshotUsage
+				yield {
+					type: "usage",
+					cacheWriteTokens: 0,
+					cacheReadTokens: usage.cached_tokens ?? 0,
+					inputTokens: (usage.prompt_tokens || 0) - (usage.cached_tokens ?? 0),
+					outputTokens: usage.completion_tokens || 0,
+				}
+			}
+		}
+	}
+
+	getModel(): { id: MoonshotModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+
+		if (modelId && modelId in moonshotModels) {
+			const id = modelId as MoonshotModelId
+			return { id, info: moonshotModels[id] }
+		}
+		return { id: moonshotDefaultModelId, info: moonshotModels[moonshotDefaultModelId] }
+	}
+}

apps/vscode/src/core/api/providers/nebius.ts

@@ -0,0 +1,96 @@
+import { type ModelInfo, type NebiusModelId, nebiusDefaultModelId, nebiusModels } from "@shared/api"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../index"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { convertToR1Format } from "../transform/r1-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface NebiusHandlerOptions extends CommonApiHandlerOptions {
+	nebiusApiKey?: string
+	apiModelId?: string
+}
+
+export class NebiusHandler implements ApiHandler {
+	private client: OpenAI | undefined
+
+	constructor(private readonly options: NebiusHandlerOptions) {}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.nebiusApiKey) {
+				throw new Error("Nebius API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: "https://api.studio.nebius.ai/v1",
+					apiKey: this.options.nebiusApiKey,
+				})
+			} catch (error) {
+				throw new Error(`Error creating Nebius client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+
+		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = model.id.includes("DeepSeek-R1")
+			? convertToR1Format([{ role: "user", content: systemPrompt }, ...messages])
+			: [{ role: "system", content: systemPrompt }, ...convertToOpenAiMessages(messages)]
+
+		const stream = await client.chat.completions.create({
+			model: model.id,
+			messages: openAiMessages,
+			temperature: 0,
+			stream: true,
+			stream_options: { include_usage: true },
+			...getOpenAIToolParams(tools),
+		})
+		const toolCallProcessor = new ToolCallProcessor()
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			if (delta && "reasoning_content" in delta && delta.reasoning_content) {
+				yield {
+					type: "reasoning",
+					reasoning: (delta.reasoning_content as string | undefined) || "",
+				}
+			}
+
+			if (chunk.usage) {
+				yield {
+					type: "usage",
+					inputTokens: chunk.usage.prompt_tokens || 0,
+					outputTokens: chunk.usage.completion_tokens || 0,
+				}
+			}
+		}
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+
+		if (modelId !== undefined && modelId in nebiusModels) {
+			return { id: modelId, info: nebiusModels[modelId as NebiusModelId] }
+		}
+		return { id: nebiusDefaultModelId, info: nebiusModels[nebiusDefaultModelId] }
+	}
+}

apps/vscode/src/core/api/providers/nousresearch.ts

@@ -0,0 +1,93 @@
+import { ModelInfo, NousResearchModelId, nousResearchDefaultModelId, nousResearchModels } from "@shared/api"
+import OpenAI from "openai"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../index"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+
+interface NousResearchHandlerOptions extends CommonApiHandlerOptions {
+	nousResearchApiKey?: string
+	apiModelId?: string
+}
+
+export class NousResearchHandler implements ApiHandler {
+	private options: NousResearchHandlerOptions
+	private client: OpenAI | undefined
+
+	constructor(options: NousResearchHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.nousResearchApiKey) {
+				throw new Error("NousResearch API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: "https://inference-api.nousResearch.com/v1",
+					apiKey: this.options.nousResearchApiKey,
+				})
+			} catch (error: any) {
+				throw new Error(`Error creating NousResearch client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[]): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+
+		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+
+		const stream = await client.chat.completions.create({
+			model: model.id,
+			messages: openAiMessages,
+			temperature: 0,
+			stream: true,
+			stream_options: { include_usage: true },
+		})
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta && "reasoning_content" in delta && delta.reasoning_content) {
+				yield {
+					type: "reasoning",
+					reasoning: (delta.reasoning_content as string | undefined) || "",
+				}
+			}
+
+			if (chunk.usage) {
+				yield {
+					type: "usage",
+					inputTokens: chunk.usage.prompt_tokens || 0,
+					outputTokens: chunk.usage.completion_tokens || 0,
+				}
+			}
+		}
+	}
+
+	getModel(): { id: NousResearchModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+
+		if (modelId && modelId in nousResearchModels) {
+			const id = modelId as NousResearchModelId
+			return { id, info: nousResearchModels[id] }
+		}
+		return { id: nousResearchDefaultModelId, info: nousResearchModels[nousResearchDefaultModelId] }
+	}
+}

apps/vscode/src/core/api/providers/oca.ts

@@ -0,0 +1,473 @@
+import { Anthropic, APIError as AnthropicAPIError } from "@anthropic-ai/sdk"
+import { liteLlmDefaultModelId, liteLlmModelInfoSaneDefaults, ModelInfo } from "@shared/api"
+import OpenAI, { APIError as OpenAIAPIError, OpenAIError } from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { OcaAuthService } from "@/services/auth/oca/OcaAuthService"
+import {
+	DEFAULT_EXTERNAL_OCA_BASE_URL,
+	DEFAULT_INTERNAL_OCA_BASE_URL,
+	OCI_HEADER_OPC_REQUEST_ID,
+} from "@/services/auth/oca/utils/constants"
+import { createOcaHeaders } from "@/services/auth/oca/utils/utils"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { OcaModelInfo } from "@/shared/api"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { fetch } from "@/shared/net"
+import { ApiFormat } from "@/shared/proto/index.cline"
+import { Logger } from "@/shared/services/Logger"
+import { ApiHandler, type CommonApiHandlerOptions } from ".."
+import { withRetry } from "../retry"
+import { sanitizeAnthropicMessages } from "../transform/anthropic-format"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { convertToOpenAIResponsesInput } from "../transform/openai-response-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+import { convertOpenAIToolsToAnthropicTools, handleAnthropicMessagesApiStreamResponse } from "../utils/messages_api_support"
+import { handleResponsesApiStreamResponse } from "../utils/responses_api_support"
+
+export interface OcaHandlerOptions extends CommonApiHandlerOptions {
+	ocaBaseUrl?: string
+	ocaModelId?: string
+	ocaModelInfo?: OcaModelInfo
+	ocaReasoningEffort?: string
+	thinkingBudgetTokens?: number
+	ocaUsePromptCache?: boolean
+	taskId?: string
+	ocaMode?: string // "internal" or "external"
+}
+
+export class OcaHandler implements ApiHandler {
+	protected options: OcaHandlerOptions
+	protected openAIClient: OpenAI | undefined
+	protected anthropicClient: Anthropic | undefined
+	protected externalHeaders: Record<string, string> = {}
+
+	constructor(options: OcaHandlerOptions) {
+		this.options = options
+	}
+
+	protected initializeOpenAIClient(options: OcaHandlerOptions): OpenAI {
+		const externalHeaders = buildExternalBasicHeaders()
+		return new (class OCIOpenAI extends OpenAI {
+			protected override async prepareOptions(opts: any): Promise<void> {
+				const token = await OcaAuthService.getInstance().getAuthToken()
+				if (!token) {
+					throw new OpenAIError("Unable to handle auth, Oracle Code Assist (OCA) access token is not available")
+				}
+				opts.headers ??= {}
+				// OCA Headers
+				const ociHeaders = await createOcaHeaders(token, options.taskId!)
+				opts.headers = { ...opts.headers, ...externalHeaders, ...ociHeaders }
+				Logger.log(`Making request with customer opc-request-id: ${opts.headers?.["opc-request-id"]}`)
+				return super.prepareOptions(opts)
+			}
+
+			protected override makeStatusError(
+				status: number | undefined,
+				error: Object | undefined,
+				message: string | undefined,
+				headers: any | undefined,
+			): OpenAIAPIError {
+				interface OciError {
+					code?: string
+					message?: string
+				}
+				let ociErrorMessage = message
+				if (typeof error === "object" && error !== null) {
+					try {
+						ociErrorMessage = JSON.stringify(error)
+						const ociErr = error as OciError
+						if (ociErr.code !== undefined && ociErr.message !== undefined) {
+							ociErrorMessage = `${ociErr.code}: ${ociErr.message}`
+						}
+					} catch {}
+				}
+				const opcRequestId = headers?.[OCI_HEADER_OPC_REQUEST_ID]
+				if (opcRequestId) {
+					ociErrorMessage += `\n(${OCI_HEADER_OPC_REQUEST_ID}: ${opcRequestId})`
+				}
+				const statusCode = typeof status === "number" ? status : 500
+				return super.makeStatusError(statusCode, error ?? {}, ociErrorMessage, headers)
+			}
+		})({
+			baseURL:
+				options.ocaBaseUrl ||
+				(options.ocaMode === "internal" ? DEFAULT_INTERNAL_OCA_BASE_URL : DEFAULT_EXTERNAL_OCA_BASE_URL),
+			apiKey: "noop",
+			fetch, // Use configured fetch with proxy support
+		})
+	}
+
+	protected initializeAnthropicClient(options: OcaHandlerOptions): Anthropic {
+		const externalHeaders = buildExternalBasicHeaders()
+		return new (class OCIAnthropic extends Anthropic {
+			protected override async prepareOptions(opts: any): Promise<void> {
+				const token = await OcaAuthService.getInstance().getAuthToken()
+				if (!token) {
+					throw new OpenAIError("Unable to handle auth, Oracle Code Assist (OCA) access token is not available")
+				}
+				opts.headers ??= {}
+				// OCA Headers
+				const ociHeaders = await createOcaHeaders(token, options.taskId!)
+				opts.headers = { ...opts.headers, ...externalHeaders, ...ociHeaders }
+				Logger.log(`Making request with customer opc-request-id: ${opts.headers?.["opc-request-id"]}`)
+				return super.prepareOptions(opts)
+			}
+
+			protected override makeStatusError(
+				status: number | undefined,
+				error: Object | undefined,
+				message: string | undefined,
+				headers: any | undefined,
+			): AnthropicAPIError {
+				interface OciError {
+					code?: string
+					message?: string
+				}
+				let ociErrorMessage = message
+				if (typeof error === "object" && error !== null) {
+					try {
+						ociErrorMessage = JSON.stringify(error)
+						const ociErr = error as OciError
+						if (ociErr.code !== undefined && ociErr.message !== undefined) {
+							ociErrorMessage = `${ociErr.code}: ${ociErr.message}`
+						}
+					} catch {}
+				}
+				const opcRequestId = headers?.[OCI_HEADER_OPC_REQUEST_ID]
+				if (opcRequestId) {
+					ociErrorMessage += `\n(${OCI_HEADER_OPC_REQUEST_ID}: ${opcRequestId})`
+				}
+				const statusCode = typeof status === "number" ? status : 500
+				return super.makeStatusError(statusCode, error ?? {}, ociErrorMessage, headers)
+			}
+		})({
+			baseURL:
+				options.ocaBaseUrl ||
+				(options.ocaMode === "internal" ? DEFAULT_INTERNAL_OCA_BASE_URL : DEFAULT_EXTERNAL_OCA_BASE_URL),
+			apiKey: "noop",
+			fetch, // Use configured fetch with proxy support
+		})
+	}
+
+	protected ensureOpenAIClient(): OpenAI {
+		if (!this.openAIClient) {
+			if (!this.options.ocaModelId) {
+				throw new Error("Oracle Code Assist (OCA) model is not selected")
+			}
+			try {
+				this.openAIClient = this.initializeOpenAIClient(this.options)
+			} catch (error) {
+				throw new Error(`Error creating Oracle Code Assist (OCA) client: ${error.message}`)
+			}
+		}
+		return this.openAIClient
+	}
+
+	protected ensureAnthropicClient(): Anthropic {
+		if (!this.anthropicClient) {
+			if (!this.options.ocaModelId) {
+				throw new Error("Oracle Code Assist (OCA) model is not selected")
+			}
+			try {
+				this.anthropicClient = this.initializeAnthropicClient(this.options)
+			} catch (error) {
+				throw new Error(`Error creating Oracle Code Assist (OCA) client: ${error.message}`)
+			}
+		}
+		return this.anthropicClient
+	}
+
+	async getApiCosts(prompt_tokens: number, completion_tokens: number): Promise<number | undefined> {
+		// Reference: https://github.com/BerriAI/litellm/blob/122ee634f434014267af104814022af1d9a0882f/litellm/proxy/spend_tracking/spend_management_endpoints.py#L1473
+		const client = this.ensureOpenAIClient()
+		const modelId = this.options.ocaModelId || liteLlmDefaultModelId
+		const token = await OcaAuthService.getInstance().getAuthToken()
+		if (!token) {
+			throw new OpenAIError("Unable to handle auth, Oracle Code Assist (OCA) access token is not available")
+		}
+		const externalHeaders = buildExternalBasicHeaders()
+		const ociHeaders = await createOcaHeaders(token, this.options.taskId!)
+		Logger.log(`Making calculate cost request with customer opc-request-id: ${ociHeaders["opc-request-id"]}`)
+		try {
+			const response = await fetch(`${client.baseURL}/spend/calculate`, {
+				method: "POST",
+				headers: { ...externalHeaders, ...ociHeaders },
+				body: JSON.stringify({
+					completion_response: {
+						model: modelId,
+						usage: {
+							prompt_tokens,
+							completion_tokens,
+						},
+					},
+				}),
+			})
+
+			if (response.ok) {
+				const data: { cost: number } = await response.json()
+				return data.cost
+			}
+			Logger.error("Error calculating spend:", response.statusText)
+			return undefined
+		} catch (error) {
+			Logger.error("Error calculating spend:", error)
+			return undefined
+		}
+	}
+
+	async calculateCost(
+		modelInfo: ModelInfo,
+		inputTokens: number,
+		outputTokens: number,
+		_cacheWriteTokens?: number,
+		_cacheReadTokens?: number,
+	) {
+		const inputCost = (await this.getApiCosts(1e6, 0)) || 0
+		const outputCost = (await this.getApiCosts(0, 1e6)) || 0
+		const totalCost = (inputCost * inputTokens) / 1e6 + (outputCost * outputTokens) / 1e6
+		return totalCost
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		if (this.options.ocaModelInfo?.apiFormat == ApiFormat.OPENAI_RESPONSES) {
+			yield* this.createMessageResponsesApi(systemPrompt, messages, tools)
+		} else if (this.options.ocaModelInfo?.apiFormat == ApiFormat.ANTHROPIC_CHAT) {
+			yield* this.createMessageMessagesApi(systemPrompt, messages, tools)
+		} else {
+			yield* this.createMessageChatApi(systemPrompt, messages, tools)
+		}
+	}
+
+	async *createMessageChatApi(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureOpenAIClient()
+		const formattedMessages = convertToOpenAiMessages(messages)
+		const systemMessage: OpenAI.Chat.ChatCompletionSystemMessageParam = {
+			role: "system",
+			content: systemPrompt,
+		}
+		const modelId = this.options.ocaModelId || liteLlmDefaultModelId
+		const isOminiModel = modelId.includes("o1-mini") || modelId.includes("o3-mini") || modelId.includes("o4-mini")
+
+		// Configuration for extended thinking
+		const budgetTokens = this.options.thinkingBudgetTokens || 0
+		const reasoningOn = budgetTokens !== 0
+		const thinkingConfig = reasoningOn ? { type: "enabled", budget_tokens: budgetTokens } : undefined
+
+		let temperature: number | undefined = this.options.ocaModelInfo?.temperature ?? 0
+		const maxTokens: number | undefined = this.options.ocaModelInfo?.maxTokens
+
+		if (isOminiModel && reasoningOn) {
+			temperature = undefined // Thinking mode doesn't support temperature
+		}
+
+		// Define cache control object if prompt caching is enabled
+		const cacheControl = this.options.ocaUsePromptCache ? { cache_control: { type: "ephemeral" } } : undefined
+
+		// Add cache_control to system message if enabled
+		const enhancedSystemMessage = {
+			...systemMessage,
+			...(cacheControl && cacheControl),
+		}
+
+		// Find the last two user messages to apply caching
+		const userMsgIndices = formattedMessages.reduce((acc, msg, index) => {
+			if (msg.role === "user") {
+				acc.push(index)
+			}
+			return acc
+		}, [] as number[])
+		const lastUserMsgIndex = userMsgIndices[userMsgIndices.length - 1] ?? -1
+		const secondLastUserMsgIndex = userMsgIndices[userMsgIndices.length - 2] ?? -1
+
+		// Apply cache_control to the last two user messages if enabled
+		const enhancedMessages = formattedMessages.map((message, index) => {
+			if ((index === lastUserMsgIndex || index === secondLastUserMsgIndex) && cacheControl) {
+				return {
+					...message,
+					...cacheControl,
+				}
+			}
+			return message
+		})
+
+		const toolCallProcessor = new ToolCallProcessor()
+
+		const chatCompletionsParams: OpenAI.Chat.Completions.ChatCompletionCreateParamsStreaming = {
+			model: this.options.ocaModelId || liteLlmDefaultModelId,
+			messages: [enhancedSystemMessage, ...enhancedMessages],
+			temperature,
+			stream: true,
+			max_completion_tokens: maxTokens,
+			max_tokens: maxTokens,
+			stream_options: { include_usage: true },
+			...(thinkingConfig && { thinking: thinkingConfig }), // Add thinking configuration when applicable
+			...(this.options.taskId && {
+				litellm_session_id: `cline-${this.options.taskId}`,
+				...getOpenAIToolParams(tools),
+			}), // Add session ID for LiteLLM tracking
+		}
+
+		if (this.options.ocaModelInfo?.supportsReasoningEffort) {
+			chatCompletionsParams["reasoning_effort"] = this.options.ocaReasoningEffort || ("medium" as any)
+		}
+
+		const stream = await client.chat.completions.create(chatCompletionsParams)
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+
+			// Handle normal text content
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			// Handle reasoning events (thinking)
+			// Thinking is not in the standard types but may be in the response
+			interface ThinkingDelta {
+				thinking?: string
+			}
+
+			if ((delta as ThinkingDelta)?.thinking) {
+				yield {
+					type: "reasoning",
+					reasoning: (delta as ThinkingDelta).thinking || "",
+				}
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			// Handle token usage information
+			if (chunk.usage) {
+				const totalCost = await this.calculateCost(
+					this.options.ocaModelInfo!,
+					chunk.usage.prompt_tokens,
+					chunk.usage.completion_tokens,
+				)
+
+				// Extract cache-related information if available
+				// Need to use type assertion since these properties are not in the standard OpenAI types
+				const usage = chunk.usage as {
+					prompt_tokens: number
+					completion_tokens: number
+					cache_creation_input_tokens?: number
+					prompt_cache_miss_tokens?: number
+					cache_read_input_tokens?: number
+					prompt_cache_hit_tokens?: number
+				}
+
+				const cacheWriteTokens = usage.cache_creation_input_tokens || usage.prompt_cache_miss_tokens || 0
+				const cacheReadTokens = usage.cache_read_input_tokens || usage.prompt_cache_hit_tokens || 0
+
+				yield {
+					type: "usage",
+					inputTokens: usage.prompt_tokens || 0,
+					outputTokens: usage.completion_tokens || 0,
+					cacheWriteTokens: cacheWriteTokens > 0 ? cacheWriteTokens : undefined,
+					cacheReadTokens: cacheReadTokens > 0 ? cacheReadTokens : undefined,
+					totalCost,
+				}
+			}
+		}
+	}
+
+	async *createMessageResponsesApi(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureOpenAIClient()
+		const inputMessages = convertToOpenAIResponsesInput(messages, { usePreviousResponseId: false }).input
+		// Convert messages to Responses API input format
+		const input: OpenAI.Responses.ResponseInputItem[] = [{ role: "system", content: systemPrompt }, ...inputMessages]
+
+		// Convert ChatCompletion tools to Responses API format if provided
+		const responseTools = tools
+			?.filter((tool) => tool?.type === "function")
+			.map((tool: any) => ({
+				type: "function" as const,
+				name: tool.function.name,
+				description: tool.function.description,
+				parameters: tool.function.parameters,
+				strict: tool.function.strict ?? true, // Responses API defaults to strict mode
+			}))
+
+		let temperature: number | undefined = this.options.ocaModelInfo?.temperature ?? 0
+		const maxOutputTokens: number | undefined = this.options.ocaModelInfo?.maxTokens
+
+		const ocaModelInfo = this.options.ocaModelInfo
+		if (!ocaModelInfo) {
+			throw new Error("Oracle Code Assist (OCA) model info is required for Responses API")
+		}
+
+		const reasoningOn = !!ocaModelInfo.supportsReasoning
+		if (reasoningOn) {
+			temperature = undefined
+		}
+
+		const responsesParams: OpenAI.Responses.ResponseCreateParamsStreaming = {
+			model: this.options.ocaModelId || liteLlmDefaultModelId,
+			input,
+			stream: true,
+			tools: responseTools,
+			...(typeof temperature === "number" ? { temperature } : {}),
+			...(typeof maxOutputTokens === "number" && maxOutputTokens > 0 ? { max_output_tokens: maxOutputTokens } : {}),
+		}
+
+		if (reasoningOn) {
+			responsesParams.reasoning = { effort: this.options.ocaReasoningEffort as any, summary: "auto" }
+		}
+
+		// Create the response using Responses API
+		const stream = await client.responses.create(responsesParams)
+
+		yield* handleResponsesApiStreamResponse(stream, ocaModelInfo, this.calculateCost.bind(this))
+	}
+
+	async *createMessageMessagesApi(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureAnthropicClient()
+
+		const modelId = this.options.ocaModelId || liteLlmDefaultModelId
+
+		const budgetTokens = this.options.thinkingBudgetTokens || 0
+		const reasoningOn = this.options.ocaModelInfo?.supportsReasoning && budgetTokens !== 0
+
+		let temperature: number | undefined = this.options.ocaModelInfo?.temperature ?? 0
+		const maxTokens: number | undefined = this.options.ocaModelInfo?.maxTokens || 8192
+
+		if (reasoningOn) {
+			temperature = 0
+		}
+
+		const anthropicTools = convertOpenAIToolsToAnthropicTools(tools)
+		const anthropicMessages = sanitizeAnthropicMessages(messages, this.options.ocaUsePromptCache ?? false)
+
+		const stream = await client.messages.create({
+			model: modelId,
+			max_tokens: maxTokens,
+			temperature: reasoningOn ? undefined : temperature,
+			system: [
+				{
+					text: systemPrompt,
+					type: "text",
+					cache_control: this.options.ocaUsePromptCache ? { type: "ephemeral" } : undefined,
+				},
+			],
+			messages: anthropicMessages,
+			stream: true,
+			tools: anthropicTools,
+			thinking: reasoningOn ? { type: "enabled", budget_tokens: budgetTokens } : undefined,
+		})
+
+		yield* handleAnthropicMessagesApiStreamResponse(stream)
+	}
+
+	getModel() {
+		return {
+			id: this.options.ocaModelId || liteLlmDefaultModelId,
+			info: this.options.ocaModelInfo || liteLlmModelInfoSaneDefaults,
+		}
+	}
+}

apps/vscode/src/core/api/providers/ollama.ts

@@ -0,0 +1,159 @@
+import { type ModelInfo, openAiModelInfoSaneDefaults } from "@shared/api"
+import { type Config, type Message, Ollama } from "ollama"
+import type { ChatCompletionTool } from "openai/resources/chat/completions"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { fetch } from "@/shared/net"
+import { Logger } from "@/shared/services/Logger"
+import type { ApiHandler, CommonApiHandlerOptions } from "../"
+import { withRetry } from "../retry"
+import { convertToOllamaMessages } from "../transform/ollama-format"
+import type { ApiStream } from "../transform/stream"
+import { ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface OllamaHandlerOptions extends CommonApiHandlerOptions {
+	ollamaBaseUrl?: string
+	ollamaApiKey?: string
+	ollamaModelId?: string
+	ollamaApiOptionsCtxNum?: string
+	requestTimeoutMs?: number
+}
+
+const DEFAULT_CONTEXT_WINDOW = 32768
+
+export class OllamaHandler implements ApiHandler {
+	private options: OllamaHandlerOptions
+	private client: Ollama | undefined
+
+	constructor(options: OllamaHandlerOptions) {
+		const ollamaApiOptionsCtxNum = (options.ollamaApiOptionsCtxNum ?? DEFAULT_CONTEXT_WINDOW).toString()
+		this.options = { ...options, ollamaApiOptionsCtxNum }
+	}
+
+	private ensureClient(): Ollama {
+		if (!this.client) {
+			try {
+				const externalHeaders = buildExternalBasicHeaders()
+				const clientOptions: Partial<Config> = {
+					host: this.options.ollamaBaseUrl,
+					fetch,
+					headers: externalHeaders,
+				}
+
+				// Add API key if provided (for Ollama cloud or authenticated instances)
+				if (this.options.ollamaApiKey) {
+					clientOptions.headers = {
+						...clientOptions.headers,
+						Authorization: `Bearer ${this.options.ollamaApiKey}`,
+					}
+				}
+
+				this.client = new Ollama(clientOptions)
+			} catch (error) {
+				throw new Error(`Error creating Ollama client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry({ retryAllErrors: true })
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: ChatCompletionTool[]): ApiStream {
+		const client = this.ensureClient()
+		const ollamaMessages: Message[] = [{ role: "system", content: systemPrompt }, ...convertToOllamaMessages(messages)]
+
+		try {
+			// Create a promise that rejects after timeout
+			const timeoutMs = this.options.requestTimeoutMs || 30000
+			const timeoutPromise = new Promise<never>((_, reject) => {
+				setTimeout(() => reject(new Error(`Ollama request timed out after ${timeoutMs / 1000} seconds`)), timeoutMs)
+			})
+
+			// Create the actual API request promise
+			const apiPromise = client.chat({
+				model: this.getModel().id,
+				messages: ollamaMessages,
+				stream: true,
+				options: {
+					num_ctx: Number(this.options.ollamaApiOptionsCtxNum),
+				},
+				tools: tools as any,
+			})
+
+			const toolCallProcessor = new ToolCallProcessor()
+
+			// Race the API request against the timeout
+			const stream = (await Promise.race([apiPromise, timeoutPromise])) as Awaited<typeof apiPromise>
+
+			try {
+				for await (const chunk of stream) {
+					Logger.debug("[OllamaHandler] Message Chunk" + JSON.stringify(chunk))
+
+					const delta = chunk.message
+
+					if (delta?.tool_calls) {
+						Logger.debug(`[OllamaHandler] Tool Calls Detected: ${JSON.stringify(delta.tool_calls)}`)
+						yield* toolCallProcessor.processToolCallDeltas(
+							delta.tool_calls?.map((tc, inx) => ({
+								index: inx,
+								id: `ollama-tool-${inx}`,
+								function: {
+									name: tc.function.name,
+									arguments:
+										typeof tc.function.arguments === "string"
+											? tc.function.arguments
+											: JSON.stringify(tc.function.arguments),
+								},
+								type: "function",
+							})),
+						)
+					}
+
+					if (typeof delta.content === "string") {
+						yield {
+							type: "text",
+							text: delta.content,
+						}
+					}
+					// Handle token usage if available
+					if (chunk.eval_count !== undefined || chunk.prompt_eval_count !== undefined) {
+						yield {
+							type: "usage",
+							inputTokens: chunk.prompt_eval_count || 0,
+							outputTokens: chunk.eval_count || 0,
+						}
+					}
+				}
+			} catch (streamError: any) {
+				Logger.error("Error processing Ollama stream:", streamError)
+				throw new Error(`Ollama stream processing error: ${streamError.message || "Unknown error"}`)
+			}
+		} catch (error) {
+			// Check if it's a timeout error
+			if (error?.message?.includes("timed out")) {
+				const timeoutMs = this.options.requestTimeoutMs || 30000
+				throw new Error(`Ollama request timed out after ${timeoutMs / 1000} seconds`)
+			}
+
+			// Enhance error reporting
+			const statusCode = error.status || error.statusCode
+			const errorMessage = error.message || "Unknown error"
+
+			Logger.error(`Ollama API error (${statusCode || "unknown"}): ${errorMessage}`)
+			throw error
+		}
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		return {
+			id: this.options.ollamaModelId || "",
+			info: {
+				...openAiModelInfoSaneDefaults,
+				contextWindow: Number(this.options.ollamaApiOptionsCtxNum),
+			},
+		}
+	}
+
+	abort(): void {
+		this.client?.abort()
+	}
+}

apps/vscode/src/core/api/providers/openai-codex.ts

@@ -0,0 +1,732 @@
+import { ModelInfo, OpenAiCodexModelId, openAiCodexDefaultModelId, openAiCodexModels } from "@shared/api"
+import { normalizeOpenaiReasoningEffort } from "@shared/storage/types"
+import OpenAI from "openai"
+import type { ChatCompletionTool } from "openai/resources/chat/completions"
+import * as os from "os"
+import { MessageEvent as UndiciMessageEvent, WebSocket as UndiciWebSocket } from "undici"
+import { v7 as uuidv7 } from "uuid"
+import { openAiCodexOAuthManager } from "@/integrations/openai-codex/oauth"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { featureFlagsService } from "@/services/feature-flags"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { fetch } from "@/shared/net"
+import { ApiFormat } from "@/shared/proto/cline/models"
+import { FeatureFlag } from "@/shared/services/feature-flags/feature-flags"
+import { Logger } from "@/shared/services/Logger"
+import { ApiHandler, CommonApiHandlerOptions } from "../"
+import { convertToOpenAIResponsesInput } from "../transform/openai-response-format"
+import { ApiStream, ApiStreamUsageChunk } from "../transform/stream"
+
+/**
+ * OpenAI Codex base URL for API requests
+ * Routes to chatgpt.com/backend-api/codex
+ */
+const CODEX_API_BASE_URL = "https://chatgpt.com/backend-api/codex"
+const CODEX_RESPONSES_WEBSOCKET_URL = "wss://chatgpt.com/backend-api/codex/responses"
+
+interface OpenAiCodexHandlerOptions extends CommonApiHandlerOptions {
+	reasoningEffort?: string
+	apiModelId?: string
+}
+
+/**
+ * OpenAiCodexHandler - Uses OpenAI Responses API with OAuth authentication
+ *
+ * Key differences from OpenAiNativeHandler:
+ * - Uses OAuth Bearer tokens instead of API keys
+ * - Routes requests to Codex backend (chatgpt.com/backend-api/codex)
+ * - Subscription-based pricing (no per-token costs)
+ * - Limited model subset
+ * - Custom headers for Codex backend
+ */
+export class OpenAiCodexHandler implements ApiHandler {
+	private options: OpenAiCodexHandlerOptions
+	private client?: OpenAI
+	private responsesWs: UndiciWebSocket | undefined
+	private websocketRequestInFlight = false
+	// Session ID for the Codex API (persists for the lifetime of the handler)
+	private readonly sessionId: string
+	// Abort controller for cancelling ongoing requests
+	private abortController?: AbortController
+	// Track tool call identity for streaming
+	private pendingToolCallId: string | undefined
+	private pendingToolCallName: string | undefined
+
+	constructor(options: OpenAiCodexHandlerOptions) {
+		this.options = options
+		this.sessionId = uuidv7()
+	}
+
+	private normalizeUsage(usage: any, _model: { id: string; info: ModelInfo }): ApiStreamUsageChunk | undefined {
+		if (!usage) {
+			return undefined
+		}
+
+		const inputDetails = usage.input_tokens_details ?? usage.prompt_tokens_details
+
+		const hasCachedTokens = typeof inputDetails?.cached_tokens === "number"
+		const hasCacheMissTokens = typeof inputDetails?.cache_miss_tokens === "number"
+		const cachedFromDetails = hasCachedTokens ? inputDetails.cached_tokens : 0
+		const missFromDetails = hasCacheMissTokens ? inputDetails.cache_miss_tokens : 0
+
+		let totalInputTokens = usage.input_tokens ?? usage.prompt_tokens ?? 0
+		if (totalInputTokens === 0 && inputDetails && (cachedFromDetails > 0 || missFromDetails > 0)) {
+			totalInputTokens = cachedFromDetails + missFromDetails
+		}
+
+		const totalOutputTokens = usage.output_tokens ?? usage.completion_tokens ?? 0
+		const cacheWriteTokens = usage.cache_creation_input_tokens ?? usage.cache_write_tokens ?? 0
+		const cacheReadTokens =
+			usage.cache_read_input_tokens ?? usage.cache_read_tokens ?? usage.cached_tokens ?? cachedFromDetails ?? 0
+
+		const reasoningTokens =
+			typeof usage.output_tokens_details?.reasoning_tokens === "number"
+				? usage.output_tokens_details.reasoning_tokens
+				: undefined
+
+		// Subscription-based: no per-token costs
+		const out: ApiStreamUsageChunk = {
+			type: "usage",
+			inputTokens: totalInputTokens,
+			outputTokens: totalOutputTokens,
+			cacheWriteTokens,
+			cacheReadTokens,
+			...(typeof reasoningTokens === "number" ? { reasoningTokens } : {}),
+			totalCost: 0, // Subscription-based pricing
+		}
+		return out
+	}
+
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: ChatCompletionTool[]): ApiStream {
+		const model = this.getModel()
+
+		// Reset state for this request
+		this.pendingToolCallId = undefined
+		this.pendingToolCallName = undefined
+
+		// Get access token from OAuth manager
+		let accessToken = await openAiCodexOAuthManager.getAccessToken()
+		if (!accessToken) {
+			throw new Error("Not authenticated with OpenAI Codex. Please sign in using the OpenAI Codex OAuth flow in settings.")
+		}
+		const useWebsocketMode = this.useWebsocketMode(model.info.apiFormat)
+		const { input, previousResponseId } = convertToOpenAIResponsesInput(messages, { usePreviousResponseId: useWebsocketMode })
+		const usePreviousResponseId = useWebsocketMode && !!previousResponseId
+
+		// Build request body
+		const requestBody = this.buildRequestBody(model, input, systemPrompt, tools, previousResponseId)
+		const fallbackRequestBody = this.buildRequestBody(model, input, systemPrompt, tools)
+
+		// Make the request with retry on auth failure
+		for (let attempt = 0; attempt < 2; attempt++) {
+			try {
+				yield* this.executeRequest(requestBody, fallbackRequestBody, model, accessToken, usePreviousResponseId)
+				return
+			} catch (error) {
+				const message = error instanceof Error ? error.message : String(error)
+				const isAuthFailure = /unauthorized|invalid token|not authenticated|authentication|401/i.test(message)
+
+				if (attempt === 0 && isAuthFailure) {
+					// Force refresh the token for retry
+					const refreshed = await openAiCodexOAuthManager.forceRefreshAccessToken()
+					if (!refreshed) {
+						throw new Error(
+							"Not authenticated with OpenAI Codex. Please sign in using the OpenAI Codex OAuth flow in settings.",
+						)
+					}
+					accessToken = refreshed
+					continue
+				}
+				throw error
+			}
+		}
+	}
+
+	private useWebsocketMode(apiFormat?: ApiFormat): boolean {
+		if (featureFlagsService.getBooleanFlagEnabled(FeatureFlag.OPENAI_RESPONSES_WEBSOCKET_MODE)) {
+			return apiFormat === ApiFormat.OPENAI_RESPONSES_WEBSOCKET_MODE
+		}
+		return false
+	}
+
+	private buildRequestBody(
+		model: { id: string; info: ModelInfo },
+		formattedInput: any,
+		systemPrompt: string,
+		tools?: ChatCompletionTool[],
+		previousResponseId?: string,
+	): any {
+		// Determine reasoning effort
+		const reasoningEffort = normalizeOpenaiReasoningEffort(this.options.reasoningEffort)
+		const includeReasoning = reasoningEffort !== "none"
+
+		const body: any = {
+			model: model.id,
+			input: formattedInput,
+			stream: true,
+			store: false,
+			instructions: systemPrompt,
+			...(previousResponseId ? { previous_response_id: previousResponseId } : {}),
+			...(includeReasoning ? { include: ["reasoning.encrypted_content"] } : {}),
+			...(includeReasoning
+				? {
+						reasoning: {
+							effort: reasoningEffort,
+							summary: "auto",
+						},
+					}
+				: {}),
+		}
+
+		// Add tools if provided
+		// Pass through strict value from tool (MCP/custom tools have strict: false, built-in tools default to true)
+		if (tools && tools.length > 0) {
+			body.tools = tools
+				.filter((tool: any) => tool?.type === "function")
+				.map((tool: any) => ({
+					type: "function",
+					name: tool.function.name,
+					description: tool.function.description,
+					parameters: tool.function.parameters,
+					strict: tool.function.strict ?? true,
+				}))
+		}
+
+		return body
+	}
+
+	private async *executeRequest(
+		requestBody: any,
+		fallbackRequestBody: any,
+		model: { id: string; info: ModelInfo },
+		accessToken: string,
+		useWebsocketMode: boolean,
+	): ApiStream {
+		// Create AbortController for cancellation
+		this.abortController = new AbortController()
+
+		try {
+			// Get ChatGPT account ID for organization subscriptions
+			const accountId = await openAiCodexOAuthManager.getAccountId()
+
+			// Build Codex-specific headers
+			const codexHeaders: Record<string, string> = {
+				originator: "cline",
+				session_id: this.sessionId,
+				"User-Agent": `cline/${process.env.npm_package_version || "1.0.0"} (${os.platform()} ${os.release()}; ${os.arch()}) node/${process.version.slice(1)}`,
+				...(accountId ? { "ChatGPT-Account-Id": accountId } : {}),
+				...buildExternalBasicHeaders(),
+			}
+
+			if (useWebsocketMode) {
+				try {
+					yield* this.createResponseStreamWebsocket(requestBody, fallbackRequestBody, accessToken, codexHeaders, model)
+					return
+				} catch (error) {
+					Logger.error("OpenAI Codex websocket mode failed, falling back to HTTP Responses API:", error)
+					this.closeResponsesWebsocket()
+				}
+			}
+
+			// Try using OpenAI SDK first
+			try {
+				const client =
+					this.client ??
+					new OpenAI({
+						apiKey: accessToken,
+						baseURL: CODEX_API_BASE_URL,
+						defaultHeaders: codexHeaders,
+						fetch, // Use shared fetch for proxy support
+					})
+
+				const stream = (await (client as any).responses.create(requestBody, {
+					signal: this.abortController.signal,
+					headers: codexHeaders,
+				})) as AsyncIterable<any>
+
+				if (typeof (stream as any)?.[Symbol.asyncIterator] !== "function") {
+					throw new Error("OpenAI SDK did not return an AsyncIterable")
+				}
+
+				for await (const event of stream) {
+					if (this.abortController.signal.aborted) {
+						break
+					}
+
+					for await (const outChunk of this.processEvent(event, model)) {
+						yield outChunk
+					}
+				}
+			} catch (_sdkErr) {
+				// Fallback to manual SSE via fetch
+				yield* this.makeCodexRequest(requestBody, model, accessToken)
+			}
+		} finally {
+			this.abortController = undefined
+		}
+	}
+
+	private async *createResponseStreamWebsocket(
+		primaryParams: OpenAI.Responses.ResponseCreateParamsStreaming,
+		fallbackParams: OpenAI.Responses.ResponseCreateParamsStreaming,
+		accessToken: string,
+		codexHeaders: Record<string, string>,
+		model: { id: string; info: ModelInfo },
+	): ApiStream {
+		try {
+			for await (const event of this.createResponseEventsViaWebsocket(primaryParams, accessToken, codexHeaders)) {
+				if (this.abortController?.signal.aborted) {
+					return
+				}
+				yield* this.processEvent(event, model)
+			}
+		} catch (error) {
+			if (this.shouldRetryWebsocketWithFullContext(error, !!primaryParams.previous_response_id)) {
+				Logger.log(
+					"Retrying Codex websocket response with full context after previous_response_not_found or socket reset",
+				)
+				this.closeResponsesWebsocket()
+				for await (const event of this.createResponseEventsViaWebsocket(fallbackParams, accessToken, codexHeaders)) {
+					if (this.abortController?.signal.aborted) {
+						return
+					}
+					yield* this.processEvent(event, model)
+				}
+				return
+			}
+			throw error
+		}
+	}
+
+	private shouldRetryWebsocketWithFullContext(error: unknown, hadPreviousResponseId: boolean): boolean {
+		const errorCode =
+			typeof error === "object" && error && "code" in error && typeof (error as { code: unknown }).code === "string"
+				? (error as { code: string }).code
+				: undefined
+
+		if (hadPreviousResponseId && errorCode === "previous_response_not_found") {
+			return true
+		}
+		if (errorCode === "websocket_closed" || errorCode === "websocket_error") {
+			return true
+		}
+		return false
+	}
+
+	private async ensureResponsesWebsocket(accessToken: string, codexHeaders: Record<string, string>): Promise<UndiciWebSocket> {
+		if (this.responsesWs && this.responsesWs.readyState === UndiciWebSocket.OPEN) {
+			return this.responsesWs
+		}
+
+		this.closeResponsesWebsocket()
+
+		const ws = new UndiciWebSocket(CODEX_RESPONSES_WEBSOCKET_URL, {
+			headers: {
+				Authorization: `Bearer ${accessToken}`,
+				"OpenAI-Beta": "responses_websockets=2026-02-06",
+				...codexHeaders,
+			},
+		})
+
+		await new Promise<void>((resolve, reject) => {
+			const cleanup = () => {
+				ws.removeEventListener("open", handleOpen)
+				ws.removeEventListener("error", handleError)
+				ws.removeEventListener("close", handleClose)
+			}
+			const handleOpen = () => {
+				cleanup()
+				resolve()
+			}
+			const handleError = () => {
+				cleanup()
+				reject(new Error("Failed to open Codex Responses websocket"))
+			}
+			const handleClose = () => {
+				cleanup()
+				reject(new Error("Codex Responses websocket closed before opening"))
+			}
+			ws.addEventListener("open", handleOpen)
+			ws.addEventListener("error", handleError)
+			ws.addEventListener("close", handleClose)
+		})
+
+		this.responsesWs = ws
+		return ws
+	}
+
+	private closeResponsesWebsocket() {
+		if (this.responsesWs) {
+			try {
+				this.responsesWs.close()
+			} catch {}
+			this.responsesWs = undefined
+		}
+	}
+
+	private async *createResponseEventsViaWebsocket(
+		params: OpenAI.Responses.ResponseCreateParamsStreaming,
+		accessToken: string,
+		codexHeaders: Record<string, string>,
+	): AsyncGenerator<OpenAI.Responses.ResponseStreamEvent> {
+		if (this.websocketRequestInFlight) {
+			const error: Error & { code?: string } = new Error("Websocket response.create is already in progress")
+			error.code = "websocket_concurrency_limit"
+			throw error
+		}
+
+		const ws = await this.ensureResponsesWebsocket(accessToken, codexHeaders)
+		this.websocketRequestInFlight = true
+
+		const eventQueue: OpenAI.Responses.ResponseStreamEvent[] = []
+		let resolver: (() => void) | undefined
+		let completed = false
+		let failure: (Error & { code?: string }) | undefined
+
+		const wake = () => {
+			const next = resolver
+			resolver = undefined
+			next?.()
+		}
+
+		const handleMessage = (evt: UndiciMessageEvent) => {
+			try {
+				let raw = ""
+				if (typeof evt.data === "string") {
+					raw = evt.data
+				} else if (evt.data instanceof ArrayBuffer) {
+					raw = new TextDecoder().decode(new Uint8Array(evt.data))
+				} else if (ArrayBuffer.isView(evt.data)) {
+					raw = new TextDecoder().decode(new Uint8Array(evt.data.buffer, evt.data.byteOffset, evt.data.byteLength))
+				} else {
+					raw = String(evt.data)
+				}
+
+				const parsed = JSON.parse(raw)
+				if (parsed?.type === "error" && parsed?.error) {
+					const error: Error & { code?: string } = new Error(parsed.error.message || "Codex Responses websocket error")
+					error.code = parsed.error.code
+					failure = error
+					completed = true
+					wake()
+					return
+				}
+
+				eventQueue.push(parsed as OpenAI.Responses.ResponseStreamEvent)
+				if (parsed?.type === "response.completed" || parsed?.type === "response.failed") {
+					completed = true
+				}
+				wake()
+			} catch (error) {
+				const parseError: Error & { code?: string } = new Error(
+					`Failed to parse websocket event: ${error instanceof Error ? error.message : String(error)}`,
+				)
+				parseError.code = "websocket_parse_error"
+				failure = parseError
+				completed = true
+				wake()
+			}
+		}
+
+		const handleError = () => {
+			const error: Error & { code?: string } = new Error("Codex Responses websocket emitted an error event")
+			error.code = "websocket_error"
+			failure = error
+			completed = true
+			wake()
+		}
+
+		const handleClose = () => {
+			if (!completed) {
+				const error: Error & { code?: string } = new Error("Codex Responses websocket closed during response stream")
+				error.code = "websocket_closed"
+				failure = error
+				completed = true
+				wake()
+			}
+		}
+
+		ws.addEventListener("message", handleMessage)
+		ws.addEventListener("error", handleError)
+		ws.addEventListener("close", handleClose)
+
+		try {
+			ws.send(
+				JSON.stringify({
+					type: "response.create",
+					...params,
+				}),
+			)
+
+			while (!completed || eventQueue.length > 0) {
+				if (eventQueue.length === 0) {
+					await new Promise<void>((resolve) => {
+						resolver = resolve
+					})
+					continue
+				}
+
+				const event = eventQueue.shift()
+				if (event) {
+					yield event
+				}
+			}
+
+			if (failure) {
+				throw failure
+			}
+		} finally {
+			ws.removeEventListener("message", handleMessage)
+			ws.removeEventListener("error", handleError)
+			ws.removeEventListener("close", handleClose)
+			this.websocketRequestInFlight = false
+		}
+	}
+
+	private async *makeCodexRequest(requestBody: any, model: { id: string; info: ModelInfo }, accessToken: string): ApiStream {
+		const url = `${CODEX_API_BASE_URL}/responses`
+
+		// Get ChatGPT account ID for organization subscriptions
+		const accountId = await openAiCodexOAuthManager.getAccountId()
+
+		// Build headers with required Codex-specific fields
+		const headers: Record<string, string> = {
+			"Content-Type": "application/json",
+			Authorization: `Bearer ${accessToken}`,
+			originator: "cline",
+			session_id: this.sessionId,
+			"User-Agent": `cline/${process.env.npm_package_version || "1.0.0"} (${os.platform()} ${os.release()}; ${os.arch()}) node/${process.version.slice(1)}`,
+		}
+
+		// Add ChatGPT-Account-Id if available
+		if (accountId) {
+			headers["ChatGPT-Account-Id"] = accountId
+		}
+
+		try {
+			const response = await fetch(url, {
+				method: "POST",
+				headers,
+				body: JSON.stringify(requestBody),
+				signal: this.abortController?.signal,
+			})
+
+			if (!response.ok) {
+				const errorText = await response.text()
+				let errorMessage = `Codex API request failed: ${response.status}`
+
+				try {
+					const errorJson = JSON.parse(errorText)
+					if (errorJson.error?.message) {
+						errorMessage = errorJson.error.message
+					} else if (errorJson.message) {
+						errorMessage = errorJson.message
+					}
+				} catch {
+					if (errorText) {
+						errorMessage += ` - ${errorText}`
+					}
+				}
+
+				throw new Error(errorMessage)
+			}
+
+			if (!response.body) {
+				throw new Error("No response body from Codex API")
+			}
+
+			yield* this.handleStreamResponse(response.body, model)
+		} catch (error) {
+			if (error instanceof Error) {
+				throw new Error(`Codex API error: ${error.message}`)
+			}
+			throw new Error("Unexpected error connecting to Codex API")
+		}
+	}
+
+	private async *handleStreamResponse(body: ReadableStream<Uint8Array>, model: { id: string; info: ModelInfo }): ApiStream {
+		const reader = body.getReader()
+		const decoder = new TextDecoder()
+		let buffer = ""
+
+		try {
+			while (true) {
+				if (this.abortController?.signal.aborted) {
+					break
+				}
+
+				const { done, value } = await reader.read()
+				if (done) {
+					break
+				}
+
+				buffer += decoder.decode(value, { stream: true })
+				const lines = buffer.split("\n")
+				buffer = lines.pop() || ""
+
+				for (const line of lines) {
+					if (line.startsWith("data: ")) {
+						const data = line.slice(6).trim()
+						if (data === "[DONE]") {
+							continue
+						}
+
+						try {
+							const parsed = JSON.parse(data)
+
+							for await (const outChunk of this.processEvent(parsed, model)) {
+								yield outChunk
+							}
+						} catch (e) {
+							if (!(e instanceof SyntaxError)) {
+								throw e
+							}
+						}
+					}
+				}
+			}
+		} finally {
+			reader.releaseLock()
+		}
+	}
+
+	private async *processEvent(event: any, model: { id: string; info: ModelInfo }): ApiStream {
+		// Handle text deltas
+		if (event?.type === "response.text.delta" || event?.type === "response.output_text.delta") {
+			if (event?.delta) {
+				yield { type: "text", text: event.delta }
+			}
+			return
+		}
+
+		// Handle reasoning deltas
+		if (
+			event?.type === "response.reasoning.delta" ||
+			event?.type === "response.reasoning_text.delta" ||
+			event?.type === "response.reasoning_summary.delta" ||
+			event?.type === "response.reasoning_summary_text.delta"
+		) {
+			if (event?.delta) {
+				yield { type: "reasoning", reasoning: event.delta }
+			}
+			return
+		}
+
+		// Handle refusal deltas
+		if (event?.type === "response.refusal.delta") {
+			if (event?.delta) {
+				yield { type: "text", text: `[Refusal] ${event.delta}` }
+			}
+			return
+		}
+
+		// Handle tool/function call deltas
+		if (event?.type === "response.tool_call_arguments.delta" || event?.type === "response.function_call_arguments.delta") {
+			const callId = event.call_id || event.tool_call_id || event.id || this.pendingToolCallId
+			const name = event.name || event.function_name || this.pendingToolCallName
+			const args = event.delta || event.arguments
+
+			if (typeof callId === "string" && callId.length > 0 && typeof name === "string" && name.length > 0) {
+				yield {
+					type: "tool_calls",
+					tool_call: {
+						call_id: callId,
+						function: {
+							id: callId,
+							name,
+							arguments: typeof args === "string" ? args : "",
+						},
+					},
+				}
+			}
+			return
+		}
+
+		// Handle output item events
+		if (event?.type === "response.output_item.added" || event?.type === "response.output_item.done") {
+			const item = event?.item
+			if (item) {
+				// Capture tool identity for subsequent argument deltas
+				if (item.type === "function_call" || item.type === "tool_call") {
+					const callId = item.call_id || item.tool_call_id || item.id
+					const name = item.name || item.function?.name || item.function_name
+					if (typeof callId === "string" && callId.length > 0) {
+						this.pendingToolCallId = callId
+						this.pendingToolCallName = typeof name === "string" ? name : undefined
+					}
+				}
+
+				if (item.type === "text" && item.text) {
+					yield { type: "text", text: item.text }
+				} else if (item.type === "reasoning" && item.text) {
+					yield { type: "reasoning", reasoning: item.text }
+				} else if (item.type === "message" && Array.isArray(item.content)) {
+					for (const content of item.content) {
+						if ((content?.type === "text" || content?.type === "output_text") && content?.text) {
+							yield { type: "text", text: content.text }
+						}
+					}
+				} else if (
+					(item.type === "function_call" || item.type === "tool_call") &&
+					event.type === "response.output_item.done"
+				) {
+					const callId = item.call_id || item.tool_call_id || item.id
+					if (callId) {
+						const args = item.arguments || item.function?.arguments || item.function_arguments
+						yield {
+							type: "tool_calls",
+							id: callId,
+							tool_call: {
+								call_id: callId,
+								function: {
+									id: callId,
+									name: item.name || item.function?.name || item.function_name || "",
+									arguments: typeof args === "string" ? args : "{}",
+								},
+							},
+						}
+					}
+				}
+			}
+			return
+		}
+
+		// Handle completion events
+		if (event?.type === "response.done" || event?.type === "response.completed") {
+			const usage = event?.response?.usage || event?.usage || undefined
+			const usageData = this.normalizeUsage(usage, model)
+			if (usageData) {
+				yield usageData
+			}
+			return
+		}
+
+		// Fallbacks for legacy formats
+		if (event?.choices?.[0]?.delta?.content) {
+			yield { type: "text", text: event.choices[0].delta.content }
+			return
+		}
+
+		if (event?.usage) {
+			const usageData = this.normalizeUsage(event.usage, model)
+			if (usageData) {
+				yield usageData
+			}
+		}
+	}
+
+	abort(): void {
+		this.closeResponsesWebsocket()
+		this.abortController?.abort()
+	}
+
+	getModel(): { id: OpenAiCodexModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+
+		const id = modelId && modelId in openAiCodexModels ? (modelId as OpenAiCodexModelId) : openAiCodexDefaultModelId
+
+		const info: ModelInfo = openAiCodexModels[id]
+
+		return { id, info }
+	}
+}

apps/vscode/src/core/api/providers/openai-native.ts

@@ -0,0 +1,702 @@
+import {
+	ModelInfo,
+	OpenAiCompatibleModelInfo,
+	OpenAiNativeModelId,
+	openAiNativeDefaultModelId,
+	openAiNativeModels,
+} from "@shared/api"
+import { normalizeOpenaiReasoningEffort } from "@shared/storage/types"
+import { calculateApiCostOpenAI } from "@utils/cost"
+import OpenAI from "openai"
+import type {
+	ChatCompletionFunctionTool,
+	ChatCompletionReasoningEffort,
+	ChatCompletionTool,
+} from "openai/resources/chat/completions"
+import { MessageEvent as UndiciMessageEvent, WebSocket as UndiciWebSocket } from "undici"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { featureFlagsService } from "@/services/feature-flags"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { ApiFormat } from "@/shared/proto/cline/models"
+import { FeatureFlag } from "@/shared/services/feature-flags/feature-flags"
+import { Logger } from "@/shared/services/Logger"
+import { isGPT5ModelFamily } from "@/utils/model-utils"
+import { ApiHandler, CommonApiHandlerOptions } from "../"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { convertToOpenAIResponsesInput } from "../transform/openai-response-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface OpenAiNativeHandlerOptions extends CommonApiHandlerOptions {
+	openAiNativeApiKey?: string
+	reasoningEffort?: string
+	thinkingBudgetTokens?: number
+	apiModelId?: string
+	openAiNativeUseResponsesWebsocket?: boolean
+}
+
+export class OpenAiNativeHandler implements ApiHandler {
+	private options: OpenAiNativeHandlerOptions
+	private client: OpenAI | undefined
+	private responsesWs: UndiciWebSocket | undefined
+	private responsesWsReadyPromise: Promise<UndiciWebSocket> | undefined
+	private websocketRequestInFlight = false
+	private abortController?: AbortController
+
+	constructor(options: OpenAiNativeHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.openAiNativeApiKey) {
+				throw new Error("OpenAI API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					apiKey: this.options.openAiNativeApiKey,
+				})
+			} catch (error) {
+				throw new Error(`Error creating OpenAI client: ${error instanceof Error ? error.message : String(error)}`)
+			}
+		}
+		return this.client
+	}
+
+	private async *yieldUsage(info: ModelInfo, usage: OpenAI.Completions.CompletionUsage | undefined): ApiStream {
+		const inputTokens = usage?.prompt_tokens || 0 // sum of cache hits and misses
+		const outputTokens = usage?.completion_tokens || 0
+		const cacheReadTokens = usage?.prompt_tokens_details?.cached_tokens || 0
+		const cacheWriteTokens = 0
+		const totalCost = calculateApiCostOpenAI(info, inputTokens, outputTokens, cacheWriteTokens, cacheReadTokens)
+		const nonCachedInputTokens = Math.max(0, inputTokens - cacheReadTokens - cacheWriteTokens)
+		yield {
+			type: "usage",
+			inputTokens: nonCachedInputTokens,
+			outputTokens: outputTokens,
+			cacheWriteTokens: cacheWriteTokens,
+			cacheReadTokens: cacheReadTokens,
+			totalCost: totalCost,
+		}
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: ChatCompletionTool[]): ApiStream {
+		// Responses API requires tool format to be set to OPENAI_RESPONSES with native tools calling enabled
+		const apiFormat = this.getModel()?.info?.apiFormat
+		if (apiFormat === ApiFormat.OPENAI_RESPONSES || apiFormat === ApiFormat.OPENAI_RESPONSES_WEBSOCKET_MODE) {
+			if (!tools?.length) {
+				throw new Error("Native Tool Call must be enabled in your setting for OpenAI Responses API")
+			}
+			yield* this.createResponseStream(systemPrompt, messages, tools)
+		} else {
+			yield* this.createCompletionStream(systemPrompt, messages, tools)
+		}
+	}
+
+	private async *createCompletionStream(
+		systemPrompt: string,
+		messages: ClineStorageMessage[],
+		tools?: ChatCompletionTool[],
+	): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+		const toolCallProcessor = new ToolCallProcessor()
+		this.abortController = new AbortController()
+
+		// Handle o1 models separately as they don't support streaming
+		if (model.info.supportsStreaming === false) {
+			const response = await client.chat.completions.create(
+				{
+					model: model.id,
+					messages: [{ role: "user", content: systemPrompt }, ...convertToOpenAiMessages(messages, "openai-native")],
+				},
+				{ signal: this.abortController?.signal },
+			)
+			yield {
+				type: "text",
+				text: response.choices[0]?.message.content || "",
+			}
+			yield* this.yieldUsage(model.info, response.usage)
+			return
+		}
+
+		const systemRole = model.info.systemRole ?? "system"
+		const includeReasoning = model.info.supportsReasoningEffort
+		const includeTools = model.info.supportsTools ?? true
+		const requestedEffort = normalizeOpenaiReasoningEffort(this.options.reasoningEffort)
+		const reasoningEffort =
+			includeReasoning && requestedEffort !== "none" ? (requestedEffort as ChatCompletionReasoningEffort) : undefined
+
+		const stream = await client.chat.completions.create({
+			model: model.id,
+			messages: [{ role: systemRole, content: systemPrompt }, ...convertToOpenAiMessages(messages, "openai-native")],
+			stream: true,
+			stream_options: { include_usage: true },
+			reasoning_effort: reasoningEffort,
+			...(model.info.temperature !== undefined ? { temperature: model.info.temperature } : {}),
+			...(includeTools ? getOpenAIToolParams(tools, isGPT5ModelFamily(model.id)) : {}),
+		})
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta?.tool_calls) {
+				try {
+					yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+				} catch (error) {
+					Logger.error("Error processing tool call delta:", error, delta.tool_calls)
+				}
+			}
+
+			if (chunk.usage) {
+				// Only last chunk contains usage
+				yield* this.yieldUsage(model.info, chunk.usage)
+			}
+		}
+	}
+
+	private async *createResponseStream(
+		systemPrompt: string,
+		messages: ClineStorageMessage[],
+		tools: ChatCompletionTool[],
+	): ApiStream {
+		const model = this.getModel()
+		const usePreviousResponseId = this.useWebsocketMode(model.info.apiFormat)
+
+		// Warm websocket connection early in websocket mode so the first response.create avoids handshake latency.
+		if (usePreviousResponseId) {
+			this.preconnectResponsesWebsocket()
+		}
+
+		const { input, previousResponseId } = convertToOpenAIResponsesInput(messages, { usePreviousResponseId })
+		const responseTools = this.mapResponseTools(tools)
+		this.abortController = new AbortController()
+
+		const params = this.buildResponseCreateParams({
+			modelId: model.id,
+			systemPrompt,
+			input,
+			previousResponseId,
+			tools: responseTools,
+		})
+
+		const fallbackParams = this.buildResponseCreateParams({
+			modelId: model.id,
+			systemPrompt,
+			input,
+			tools: responseTools,
+		})
+
+		if (usePreviousResponseId && previousResponseId) {
+			try {
+				yield* this.createResponseStreamWebsocket(model.info, params, fallbackParams)
+				return
+			} catch (error) {
+				Logger.error("OpenAI websocket mode failed, falling back to HTTP Responses API:", error)
+				this.closeResponsesWebsocket()
+			}
+		}
+
+		yield* this.createResponseStreamHttp(model.info, params)
+	}
+
+	private preconnectResponsesWebsocket(): void {
+		void this.ensureResponsesWebsocket().catch((error) => {
+			Logger.debug("OpenAI websocket preconnect failed:", error)
+			this.closeResponsesWebsocket()
+		})
+	}
+
+	private useWebsocketMode(apiFormat?: ApiFormat): boolean {
+		if (featureFlagsService.getBooleanFlagEnabled(FeatureFlag.OPENAI_RESPONSES_WEBSOCKET_MODE)) {
+			return apiFormat === ApiFormat.OPENAI_RESPONSES_WEBSOCKET_MODE
+		}
+		return false
+	}
+
+	private mapResponseTools(tools: ChatCompletionTool[]): OpenAI.Responses.Tool[] {
+		return tools
+			?.filter((tool): tool is ChatCompletionFunctionTool => tool?.type === "function")
+			.map((tool) => ({
+				type: "function" as const,
+				name: tool.function.name,
+				description: tool.function.description,
+				parameters: tool.function.parameters ?? null,
+				strict: tool.function.strict ?? true,
+			}))
+	}
+
+	private buildResponseCreateParams(args: {
+		modelId: string
+		systemPrompt: string
+		input: OpenAI.Responses.ResponseInput
+		tools: OpenAI.Responses.Tool[]
+		previousResponseId?: string
+	}): OpenAI.Responses.ResponseCreateParamsStreaming {
+		const requestedEffort = normalizeOpenaiReasoningEffort(this.options.reasoningEffort)
+		const reasoning: { effort: ChatCompletionReasoningEffort; summary: "auto" } | undefined =
+			requestedEffort === "none"
+				? undefined
+				: {
+						effort: requestedEffort,
+						summary: "auto",
+					}
+
+		return {
+			model: args.modelId,
+			instructions: args.systemPrompt,
+			input: args.input,
+			stream: true,
+			tools: args.tools,
+			store: !args.previousResponseId, // Do not use store when websocket mode is enabled.
+			...(args.previousResponseId ? { previous_response_id: args.previousResponseId } : {}),
+			...(reasoning ? { reasoning } : {}),
+		}
+	}
+
+	private async *createResponseStreamHttp(
+		modelInfo: ModelInfo,
+		params: OpenAI.Responses.ResponseCreateParamsStreaming,
+	): ApiStream {
+		const client = this.ensureClient()
+		Logger.debug(`OpenAI Responses Input (HTTP): ${JSON.stringify(params.input)}`)
+		const stream = await client.responses.create(params, { signal: this.abortController?.signal })
+		yield* this.processResponsesEvents(stream, modelInfo)
+	}
+
+	private async *createResponseStreamWebsocket(
+		modelInfo: ModelInfo,
+		primaryParams: OpenAI.Responses.ResponseCreateParamsStreaming,
+		fallbackParams: OpenAI.Responses.ResponseCreateParamsStreaming,
+	): ApiStream {
+		Logger.debug(`OpenAI Responses Input (WebSocket): ${JSON.stringify(primaryParams.input)}`)
+		try {
+			yield* this.processResponsesEvents(this.createResponseEventsViaWebsocket(primaryParams), modelInfo)
+		} catch (error) {
+			if (this.shouldRetryWebsocketWithFullContext(error, !!primaryParams.previous_response_id)) {
+				Logger.log("Retrying websocket response with full context after previous_response_not_found or socket reset")
+				this.closeResponsesWebsocket()
+				yield* this.processResponsesEvents(this.createResponseEventsViaWebsocket(fallbackParams), modelInfo)
+				return
+			}
+			throw error
+		}
+	}
+
+	private shouldRetryWebsocketWithFullContext(error: unknown, hadPreviousResponseId: boolean): boolean {
+		const errorCode =
+			typeof error === "object" && error && "code" in error && typeof (error as { code: unknown }).code === "string"
+				? (error as { code: string }).code
+				: undefined
+
+		if (hadPreviousResponseId && errorCode === "previous_response_not_found") {
+			return true
+		}
+		if (errorCode === "websocket_closed" || errorCode === "websocket_error") {
+			return true
+		}
+		return false
+	}
+
+	private async ensureResponsesWebsocket(): Promise<UndiciWebSocket> {
+		if (this.responsesWs && this.responsesWs.readyState === UndiciWebSocket.OPEN) {
+			return this.responsesWs
+		}
+
+		if (this.responsesWsReadyPromise) {
+			return this.responsesWsReadyPromise
+		}
+
+		this.closeResponsesWebsocket()
+
+		if (!this.options.openAiNativeApiKey) {
+			throw new Error("OpenAI API key is required")
+		}
+
+		const ws = new UndiciWebSocket("wss://api.openai.com/v1/responses", {
+			headers: {
+				Authorization: `Bearer ${this.options.openAiNativeApiKey}`,
+				"OpenAI-Beta": "responses_websockets=2026-02-06",
+				...buildExternalBasicHeaders(),
+			},
+		})
+
+		this.responsesWs = ws
+		const readyPromise = new Promise<UndiciWebSocket>((resolve, reject) => {
+			const cleanup = () => {
+				ws.removeEventListener("open", handleOpen)
+				ws.removeEventListener("error", handleError)
+				ws.removeEventListener("close", handleClose)
+			}
+			const handleOpen = () => {
+				cleanup()
+				resolve(ws)
+			}
+			const handleError = () => {
+				cleanup()
+				reject(new Error("Failed to open Responses websocket"))
+			}
+			const handleClose = () => {
+				cleanup()
+				reject(new Error("Responses websocket closed before opening"))
+			}
+			ws.addEventListener("open", handleOpen)
+			ws.addEventListener("error", handleError)
+			ws.addEventListener("close", handleClose)
+		})
+
+		this.responsesWsReadyPromise = readyPromise
+
+		try {
+			return await readyPromise
+		} catch (error) {
+			if (this.responsesWs === ws) {
+				this.responsesWs = undefined
+			}
+			throw error
+		} finally {
+			if (this.responsesWsReadyPromise === readyPromise) {
+				this.responsesWsReadyPromise = undefined
+			}
+		}
+	}
+
+	private closeResponsesWebsocket() {
+		this.responsesWsReadyPromise = undefined
+		if (this.responsesWs) {
+			try {
+				this.responsesWs.close()
+			} catch {}
+			this.responsesWs = undefined
+		}
+	}
+
+	private async *createResponseEventsViaWebsocket(
+		params: OpenAI.Responses.ResponseCreateParamsStreaming,
+	): AsyncGenerator<OpenAI.Responses.ResponseStreamEvent> {
+		if (this.websocketRequestInFlight) {
+			const error: Error & { code?: string } = new Error("Websocket response.create is already in progress")
+			error.code = "websocket_concurrency_limit"
+			throw error
+		}
+
+		const ws = await this.ensureResponsesWebsocket()
+		this.websocketRequestInFlight = true
+
+		const eventQueue: OpenAI.Responses.ResponseStreamEvent[] = []
+		let resolver: (() => void) | undefined
+		let completed = false
+		let failure: (Error & { code?: string }) | undefined
+
+		const wake = () => {
+			const next = resolver
+			resolver = undefined
+			next?.()
+		}
+
+		const handleMessage = (evt: UndiciMessageEvent) => {
+			try {
+				let raw = ""
+				if (typeof evt.data === "string") {
+					raw = evt.data
+				} else if (evt.data instanceof ArrayBuffer) {
+					raw = new TextDecoder().decode(new Uint8Array(evt.data))
+				} else if (ArrayBuffer.isView(evt.data)) {
+					raw = new TextDecoder().decode(new Uint8Array(evt.data.buffer, evt.data.byteOffset, evt.data.byteLength))
+				} else {
+					raw = String(evt.data)
+				}
+				const parsed = JSON.parse(raw)
+
+				if (parsed?.type === "error" && parsed?.error) {
+					const error: Error & { code?: string } = new Error(parsed.error.message || "Responses websocket error")
+					error.code = parsed.error.code
+					failure = error
+					completed = true
+					wake()
+					return
+				}
+
+				eventQueue.push(parsed as OpenAI.Responses.ResponseStreamEvent)
+				if (parsed?.type === "response.completed" || parsed?.type === "response.failed") {
+					completed = true
+				}
+				wake()
+			} catch (error) {
+				const parseError: Error & { code?: string } = new Error(
+					`Failed to parse websocket event: ${error instanceof Error ? error.message : String(error)}`,
+				)
+				parseError.code = "websocket_parse_error"
+				failure = parseError
+				completed = true
+				wake()
+			}
+		}
+
+		const handleError = () => {
+			const error: Error & { code?: string } = new Error("Responses websocket emitted an error event")
+			error.code = "websocket_error"
+			failure = error
+			completed = true
+			wake()
+		}
+
+		const handleClose = () => {
+			if (!completed) {
+				const error: Error & { code?: string } = new Error("Responses websocket closed during response stream")
+				error.code = "websocket_closed"
+				failure = error
+				completed = true
+				wake()
+			}
+		}
+
+		ws.addEventListener("message", handleMessage)
+		ws.addEventListener("error", handleError)
+		ws.addEventListener("close", handleClose)
+
+		try {
+			ws.send(
+				JSON.stringify({
+					type: "response.create",
+					...params,
+				}),
+			)
+
+			while (!completed || eventQueue.length > 0) {
+				if (eventQueue.length === 0) {
+					await new Promise<void>((resolve) => {
+						resolver = resolve
+					})
+					continue
+				}
+
+				const event = eventQueue.shift()
+				if (event) {
+					yield event
+				}
+			}
+
+			if (failure) {
+				throw failure
+			}
+		} finally {
+			ws.removeEventListener("message", handleMessage)
+			ws.removeEventListener("error", handleError)
+			ws.removeEventListener("close", handleClose)
+			this.websocketRequestInFlight = false
+		}
+	}
+
+	private async *processResponsesEvents(
+		stream: AsyncIterable<OpenAI.Responses.ResponseStreamEvent>,
+		modelInfo: ModelInfo,
+	): ApiStream {
+		const functionCallByItemId = new Map<string, { call_id?: string; name?: string; id?: string }>()
+
+		for await (const chunk of stream) {
+			Logger.debug(`OpenAI Responses Chunk: ${JSON.stringify(chunk)}`)
+
+			if (chunk.type === "response.output_item.added") {
+				const item = chunk.item
+				if (item.type === "function_call" && item.id) {
+					functionCallByItemId.set(item.id, { call_id: item.call_id, name: item.name, id: item.id })
+					yield {
+						type: "tool_calls",
+						id: item.id,
+						tool_call: {
+							call_id: item.call_id,
+							function: {
+								id: item.id,
+								name: item.name,
+								arguments: item.arguments,
+							},
+						},
+					}
+				}
+				if (item.type === "reasoning" && item.encrypted_content && item.id) {
+					yield {
+						type: "reasoning",
+						id: item.id,
+						reasoning: "",
+						redacted_data: item.encrypted_content,
+					}
+				}
+			}
+			if (chunk.type === "response.output_item.done") {
+				const item = chunk.item
+				if (item.type === "function_call") {
+					if (item.id) {
+						functionCallByItemId.set(item.id, { call_id: item.call_id, name: item.name, id: item.id })
+					}
+					yield {
+						type: "tool_calls",
+						id: item.id || item.call_id,
+						tool_call: {
+							call_id: item.call_id,
+							function: {
+								id: item.id,
+								name: item.name,
+								arguments: item.arguments,
+							},
+						},
+					}
+				}
+				if (item.type === "reasoning") {
+					yield {
+						type: "reasoning",
+						id: item.id,
+						details: item.summary,
+						reasoning: "",
+					}
+				}
+			}
+			if (chunk.type === "response.reasoning_summary_part.added") {
+				yield {
+					type: "reasoning",
+					id: chunk.item_id,
+					reasoning: chunk.part.text,
+				}
+			}
+			if (chunk.type === "response.reasoning_summary_text.delta") {
+				yield {
+					type: "reasoning",
+					id: chunk.item_id,
+					reasoning: chunk.delta,
+				}
+			}
+			if (chunk.type === "response.reasoning_summary_part.done") {
+				yield {
+					type: "reasoning",
+					id: chunk.item_id,
+					details: chunk.part,
+					reasoning: "",
+				}
+			}
+			if (chunk.type === "response.output_text.delta") {
+				if (chunk.delta) {
+					yield {
+						id: chunk.item_id,
+						type: "text",
+						text: chunk.delta,
+					}
+				}
+			}
+			if (chunk.type === "response.reasoning_text.delta") {
+				if (chunk.delta) {
+					yield {
+						id: chunk.item_id,
+						type: "reasoning",
+						reasoning: chunk.delta,
+					}
+				}
+			}
+			if (chunk.type === "response.function_call_arguments.delta") {
+				const pendingCall = functionCallByItemId.get(chunk.item_id)
+				const callId = pendingCall?.call_id
+				const functionName = pendingCall?.name
+				const functionId = pendingCall?.id || chunk.item_id
+
+				yield {
+					type: "tool_calls",
+					tool_call: {
+						call_id: callId,
+						function: {
+							id: functionId,
+							name: functionName,
+							arguments: chunk.delta,
+						},
+					},
+				}
+			}
+			if (chunk.type === "response.function_call_arguments.done") {
+				if (chunk.item_id && chunk.name && chunk.arguments) {
+					const pendingCall = functionCallByItemId.get(chunk.item_id)
+					const callId = pendingCall?.call_id
+					const functionId = pendingCall?.id || chunk.item_id
+
+					yield {
+						type: "tool_calls",
+						tool_call: {
+							call_id: callId,
+							function: {
+								id: functionId,
+								name: chunk.name,
+								arguments: chunk.arguments,
+							},
+						},
+					}
+				}
+			}
+
+			if (
+				chunk.type === "response.incomplete" &&
+				chunk.response?.status === "incomplete" &&
+				chunk.response?.incomplete_details?.reason === "max_output_tokens"
+			) {
+				if (chunk.response?.output_text?.length > 0) {
+					Logger.log("Partial output:", chunk.response.output_text)
+				} else {
+					Logger.log("Ran out of tokens during reasoning")
+				}
+			}
+
+			if (chunk.type === "response.completed" && chunk.response?.usage) {
+				const usage = chunk.response.usage
+				const inputTokens = usage.input_tokens || 0
+				const outputTokens = usage.output_tokens || 0
+				const cacheReadTokens = usage.input_tokens_details?.cached_tokens || 0
+				const cacheWriteTokens = 0
+				const reasoningTokens = usage.output_tokens_details?.reasoning_tokens || 0
+				const totalTokens = usage.total_tokens || 0
+				Logger.log(`Total tokens from Responses API usage: ${totalTokens}`)
+				const totalCost = calculateApiCostOpenAI(
+					modelInfo,
+					inputTokens,
+					outputTokens + reasoningTokens,
+					cacheWriteTokens,
+					cacheReadTokens,
+				)
+				const nonCachedInputTokens = Math.max(0, inputTokens - cacheReadTokens - cacheWriteTokens)
+				yield {
+					type: "usage",
+					inputTokens: nonCachedInputTokens,
+					outputTokens: outputTokens,
+					cacheWriteTokens: cacheWriteTokens,
+					cacheReadTokens: cacheReadTokens,
+					thoughtsTokenCount: reasoningTokens,
+					totalCost: totalCost,
+					id: chunk.response.id,
+				}
+			}
+		}
+	}
+
+	abort(): void {
+		this.closeResponsesWebsocket()
+		this.abortController?.abort()
+		this.abortController = undefined
+	}
+
+	getModel(): { id: OpenAiNativeModelId; info: OpenAiCompatibleModelInfo } {
+		const modelId = this.options.apiModelId
+		if (modelId && modelId in openAiNativeModels) {
+			const id = modelId as OpenAiNativeModelId
+			const info = openAiNativeModels[id]
+			return { id, info: { ...info } }
+		}
+		return {
+			id: openAiNativeDefaultModelId,
+			info: { ...openAiNativeModels[openAiNativeDefaultModelId] },
+		}
+	}
+}

apps/vscode/src/core/api/providers/openai.ts

@@ -0,0 +1,188 @@
+import { DefaultAzureCredential, getBearerTokenProvider } from "@azure/identity"
+import { azureOpenAiDefaultApiVersion, ModelInfo, OpenAiCompatibleModelInfo, openAiModelInfoSaneDefaults } from "@shared/api"
+import { normalizeOpenaiReasoningEffort } from "@shared/storage/types"
+import OpenAI, { AzureOpenAI } from "openai"
+import type { ChatCompletionReasoningEffort, ChatCompletionTool } from "openai/resources/chat/completions"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient, fetch } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../index"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { convertToR1Format } from "../transform/r1-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface OpenAiHandlerOptions extends CommonApiHandlerOptions {
+	openAiApiKey?: string
+	openAiBaseUrl?: string
+	azureApiVersion?: string
+	azureIdentity?: boolean
+	openAiHeaders?: Record<string, string>
+	openAiModelId?: string
+	openAiModelInfo?: OpenAiCompatibleModelInfo
+	reasoningEffort?: string
+}
+
+export class OpenAiHandler implements ApiHandler {
+	private options: OpenAiHandlerOptions
+	private client: OpenAI | undefined
+
+	constructor(options: OpenAiHandlerOptions) {
+		this.options = options
+	}
+
+	private getAzureAudienceScope(baseUrl?: string): string {
+		const url = baseUrl?.toLowerCase() ?? ""
+		if (url.includes("azure.us")) return "https://cognitiveservices.azure.us/.default"
+		if (url.includes("azure.com")) return "https://cognitiveservices.azure.com/.default"
+		return "https://cognitiveservices.azure.com/.default"
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.openAiApiKey && !this.options.azureIdentity) {
+				throw new Error("OpenAI API key or Azure Identity Authentication is required")
+			}
+			try {
+				const baseUrl = this.options.openAiBaseUrl?.toLowerCase() ?? ""
+				const isAzureDomain = baseUrl.includes("azure.com") || baseUrl.includes("azure.us")
+				const externalHeaders = buildExternalBasicHeaders()
+				// Azure API shape slightly differs from the core API shape...
+				if (
+					this.options.azureApiVersion ||
+					(isAzureDomain && !this.options.openAiModelId?.toLowerCase().includes("deepseek"))
+				) {
+					if (this.options.azureIdentity) {
+						this.client = new AzureOpenAI({
+							baseURL: this.options.openAiBaseUrl,
+							azureADTokenProvider: getBearerTokenProvider(
+								new DefaultAzureCredential(),
+								this.getAzureAudienceScope(this.options.openAiBaseUrl),
+							),
+							apiVersion: this.options.azureApiVersion || azureOpenAiDefaultApiVersion,
+							defaultHeaders: {
+								...externalHeaders,
+								...this.options.openAiHeaders,
+							},
+							fetch,
+						})
+					} else {
+						this.client = new AzureOpenAI({
+							baseURL: this.options.openAiBaseUrl,
+							apiKey: this.options.openAiApiKey,
+							apiVersion: this.options.azureApiVersion || azureOpenAiDefaultApiVersion,
+							defaultHeaders: {
+								...externalHeaders,
+								...this.options.openAiHeaders,
+							},
+							fetch,
+						})
+					}
+				} else {
+					this.client = createOpenAIClient({
+						baseURL: this.options.openAiBaseUrl,
+						apiKey: this.options.openAiApiKey,
+						defaultHeaders: this.options.openAiHeaders,
+					})
+				}
+			} catch (error: any) {
+				throw new Error(`Error creating OpenAI client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: ChatCompletionTool[]): ApiStream {
+		const client = this.ensureClient()
+		const modelId = this.options.openAiModelId ?? ""
+		const isDeepseekReasoner = modelId.includes("deepseek-reasoner")
+		const isR1FormatRequired = this.options.openAiModelInfo?.isR1FormatRequired ?? false
+		const isReasoningModelFamily =
+			["o1", "o3", "o4", "gpt-5"].some((prefix) => modelId.includes(prefix)) && !modelId.includes("chat")
+
+		let openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+		let temperature: number | undefined
+		if (this.options.openAiModelInfo?.temperature !== undefined) {
+			const tempValue = Number(this.options.openAiModelInfo.temperature)
+			temperature = tempValue === 0 ? undefined : tempValue
+		} else {
+			temperature = openAiModelInfoSaneDefaults.temperature
+		}
+		let reasoningEffort: ChatCompletionReasoningEffort | undefined
+		let maxTokens: number | undefined
+
+		if (this.options.openAiModelInfo?.maxTokens && this.options.openAiModelInfo.maxTokens > 0) {
+			maxTokens = Number(this.options.openAiModelInfo.maxTokens)
+		} else {
+			maxTokens = undefined
+		}
+
+		if (isDeepseekReasoner || isR1FormatRequired) {
+			openAiMessages = convertToR1Format([{ role: "user", content: systemPrompt }, ...messages])
+		}
+
+		if (isReasoningModelFamily) {
+			openAiMessages = [{ role: "developer", content: systemPrompt }, ...convertToOpenAiMessages(messages)]
+			temperature = undefined // does not support temperature
+			const requestedEffort = normalizeOpenaiReasoningEffort(this.options.reasoningEffort)
+			reasoningEffort = requestedEffort === "none" ? undefined : (requestedEffort as ChatCompletionReasoningEffort)
+		}
+
+		const stream = await client.chat.completions.create({
+			model: modelId,
+			messages: openAiMessages,
+			temperature,
+			max_tokens: maxTokens,
+			reasoning_effort: reasoningEffort,
+			stream: true,
+			stream_options: { include_usage: true },
+			...getOpenAIToolParams(tools),
+		})
+
+		const toolCallProcessor = new ToolCallProcessor()
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta && "reasoning_content" in delta && delta.reasoning_content) {
+				yield {
+					type: "reasoning",
+					reasoning: (delta.reasoning_content as string | undefined) || "",
+				}
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			if (chunk.usage) {
+				yield {
+					type: "usage",
+					inputTokens: chunk.usage.prompt_tokens || 0,
+					outputTokens: chunk.usage.completion_tokens || 0,
+					cacheReadTokens: chunk.usage.prompt_tokens_details?.cached_tokens || 0,
+					// @ts-expect-error-next-line
+					cacheWriteTokens: chunk.usage.prompt_cache_miss_tokens || 0,
+				}
+			}
+		}
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		return {
+			id: this.options.openAiModelId ?? "",
+			info: this.options.openAiModelInfo ?? openAiModelInfoSaneDefaults,
+		}
+	}
+}

apps/vscode/src/core/api/providers/openrouter.ts

@@ -0,0 +1,235 @@
+import { setTimeout as setTimeoutPromise } from "node:timers/promises"
+import { StateManager } from "@core/storage/StateManager"
+import { ModelInfo, openRouterDefaultModelId, openRouterDefaultModelInfo } from "@shared/api"
+import { shouldSkipReasoningForModel } from "@utils/model-utils"
+import axios from "axios"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient, getAxiosSettings } from "@/shared/net"
+import { Logger } from "@/shared/services/Logger"
+import { ApiHandler, CommonApiHandlerOptions } from "../"
+import { withRetry } from "../retry"
+import { createOpenRouterStream } from "../transform/openrouter-stream"
+import { ApiStream, ApiStreamUsageChunk } from "../transform/stream"
+import { ToolCallProcessor } from "../transform/tool-call-processor"
+import { OpenRouterErrorResponse } from "./types"
+
+interface OpenRouterHandlerOptions extends CommonApiHandlerOptions {
+	openRouterApiKey?: string
+	openRouterModelId?: string
+	openRouterModelInfo?: ModelInfo
+	openRouterProviderSorting?: string
+	reasoningEffort?: string
+	thinkingBudgetTokens?: number
+	enableParallelToolCalling?: boolean
+}
+
+export class OpenRouterHandler implements ApiHandler {
+	private options: OpenRouterHandlerOptions
+	private client: OpenAI | undefined
+	lastGenerationId?: string
+
+	constructor(options: OpenRouterHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.openRouterApiKey) {
+				throw new Error("OpenRouter API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: "https://openrouter.ai/api/v1",
+					apiKey: this.options.openRouterApiKey,
+					defaultHeaders: {
+						"HTTP-Referer": "https://cline.bot", // Optional, for including your app on openrouter.ai rankings.
+						"X-Title": "Cline", // Optional. Shows in rankings on openrouter.ai.
+					},
+				})
+			} catch (error: any) {
+				throw new Error(`Error creating OpenRouter client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		this.lastGenerationId = undefined
+
+		const stream = await createOpenRouterStream(
+			client,
+			systemPrompt,
+			messages,
+			this.getModel(),
+			this.options.reasoningEffort,
+			this.options.thinkingBudgetTokens,
+			this.options.openRouterProviderSorting,
+			tools,
+			this.options.enableParallelToolCalling,
+		)
+
+		let didOutputUsage = false
+		const toolCallProcessor = new ToolCallProcessor()
+
+		for await (const chunk of stream) {
+			// openrouter returns an error object instead of the openai sdk throwing an error
+			// Check for error field directly on chunk
+			if ("error" in chunk) {
+				const error = chunk.error as OpenRouterErrorResponse["error"]
+				Logger.error(`OpenRouter API Error: ${error?.code} - ${error?.message}`)
+				// Include metadata in the error message if available
+				const metadataStr = error.metadata ? `\nMetadata: ${JSON.stringify(error.metadata, null, 2)}` : ""
+				throw new Error(`OpenRouter API Error ${error.code}: ${error.message}${metadataStr}`)
+			}
+
+			// Check for error in choices[0].finish_reason
+			// OpenRouter may return errors in a non-standard way within choices
+			const choice = chunk.choices?.[0]
+			// Use type assertion since OpenRouter uses non-standard "error" finish_reason
+			if ((choice?.finish_reason as string) === "error") {
+				// Use type assertion since OpenRouter adds non-standard error property
+				const choiceWithError = choice as any
+				if (choiceWithError.error) {
+					const error = choiceWithError.error
+					Logger.error(
+						`OpenRouter Mid-Stream Error: ${error?.code || "Unknown"} - ${error?.message || "Unknown error"}`,
+					)
+					// Format error details
+					const errorDetails = typeof error === "object" ? JSON.stringify(error, null, 2) : String(error)
+					throw new Error(`OpenRouter Mid-Stream Error: ${errorDetails}`)
+				}
+				// Fallback if error details are not available
+				throw new Error(`OpenRouter Mid-Stream Error: Stream terminated with error status but no error details provided`)
+			}
+
+			if (!this.lastGenerationId && chunk.id) {
+				this.lastGenerationId = chunk.id
+			}
+
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			// Reasoning tokens are returned separately from the content
+			// Skip reasoning content for Grok 4 models since it only displays "thinking" without providing useful information
+			if (
+				delta &&
+				"reasoning" in delta &&
+				delta.reasoning &&
+				!shouldSkipReasoningForModel(this.options.openRouterModelId)
+			) {
+				yield {
+					type: "reasoning",
+					reasoning: typeof delta.reasoning === "string" ? delta.reasoning : JSON.stringify(delta.reasoning),
+				}
+			}
+
+			// OpenRouter passes reasoning details that we can pass back unmodified in api requests to preserve reasoning traces for model
+			// See: https://openrouter.ai/docs/use-cases/reasoning-tokens#preserving-reasoning-blocks
+			if (
+				delta &&
+				"reasoning_details" in delta &&
+				delta.reasoning_details &&
+				// @ts-expect-error-next-line
+				delta.reasoning_details.length && // exists and non-0
+				!shouldSkipReasoningForModel(this.options.openRouterModelId)
+			) {
+				yield {
+					type: "reasoning",
+					reasoning: "",
+					details: delta.reasoning_details,
+				}
+			}
+
+			if (!didOutputUsage && chunk.usage) {
+				// @ts-expect-error-next-line -- OpenRouter returns cache_write_tokens for Anthropic models
+				const cacheWriteTokens = chunk.usage.prompt_tokens_details?.cache_write_tokens || 0
+				yield {
+					type: "usage",
+					cacheWriteTokens,
+					cacheReadTokens: chunk.usage.prompt_tokens_details?.cached_tokens || 0,
+					inputTokens:
+						(chunk.usage.prompt_tokens || 0) -
+						(chunk.usage.prompt_tokens_details?.cached_tokens || 0) -
+						(cacheWriteTokens || 0),
+					outputTokens: chunk.usage.completion_tokens || 0,
+					// @ts-expect-error-next-line
+					totalCost: (chunk.usage.cost || 0) + (chunk.usage.cost_details?.upstream_inference_cost || 0),
+				}
+				didOutputUsage = true
+			}
+		}
+
+		// Fallback to generation endpoint if usage chunk not returned
+		if (!didOutputUsage) {
+			const apiStreamUsage = await this.getApiStreamUsage()
+			if (apiStreamUsage) {
+				yield apiStreamUsage
+			}
+		}
+	}
+
+	async getApiStreamUsage(): Promise<ApiStreamUsageChunk | undefined> {
+		if (this.lastGenerationId) {
+			await setTimeoutPromise(500) // FIXME: necessary delay to ensure generation endpoint is ready
+			try {
+				const generationIterator = this.fetchGenerationDetails(this.lastGenerationId)
+				const generation = (await generationIterator.next()).value
+				// Logger.log("OpenRouter generation details:", generation)
+				return {
+					type: "usage",
+					cacheWriteTokens: generation?.native_tokens_cache_write || 0,
+					cacheReadTokens: generation?.native_tokens_cached || 0,
+					// openrouter generation endpoint fails often
+					inputTokens: (generation?.native_tokens_prompt || 0) - (generation?.native_tokens_cached || 0),
+					outputTokens: generation?.native_tokens_completion || 0,
+					totalCost: generation?.total_cost || 0,
+				}
+			} catch (error) {
+				// ignore if fails
+				Logger.error("Error fetching OpenRouter generation details:", error)
+			}
+		}
+		return undefined
+	}
+
+	@withRetry({ maxRetries: 4, baseDelay: 250, maxDelay: 1000, retryAllErrors: true })
+	async *fetchGenerationDetails(genId: string) {
+		// Logger.log("Fetching generation details for:", genId)
+		try {
+			const response = await axios.get(`https://openrouter.ai/api/v1/generation?id=${genId}`, {
+				headers: {
+					Authorization: `Bearer ${this.options.openRouterApiKey}`,
+				},
+				timeout: 15_000, // this request hangs sometimes
+				...getAxiosSettings(),
+			})
+			yield response.data?.data
+		} catch (error) {
+			// ignore if fails
+			Logger.error("Error fetching OpenRouter generation details:", error)
+			throw error
+		}
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		const modelId = this.options.openRouterModelId || openRouterDefaultModelId
+		const cachedModelInfo = StateManager.get().getModelInfo("openRouter", modelId)
+		return {
+			id: modelId,
+			info: cachedModelInfo || openRouterDefaultModelInfo,
+		}
+	}
+}

apps/vscode/src/core/api/providers/qwen-code.ts

@@ -0,0 +1,286 @@
+import { promises as fs } from "node:fs"
+import { ModelInfo, QwenCodeModelId, qwenCodeDefaultModelId, qwenCodeModels } from "@shared/api"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import * as os from "os"
+import * as path from "path"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { fetch } from "@/shared/net"
+import { Logger } from "@/shared/services/Logger"
+import { ApiHandler, CommonApiHandlerOptions } from "../"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+// --- Constants for Qwen OAuth2 ---
+const QWEN_OAUTH_BASE_URL = "https://chat.qwen.ai"
+const QWEN_OAUTH_TOKEN_ENDPOINT = `${QWEN_OAUTH_BASE_URL}/api/v1/oauth2/token`
+const QWEN_OAUTH_CLIENT_ID = "f0304373b74a44d2b584a3fb70ca9e56"
+const QWEN_DIR = ".qwen"
+const QWEN_CREDENTIAL_FILENAME = "oauth_creds.json"
+
+interface QwenOAuthCredentials {
+	access_token: string
+	refresh_token: string
+	token_type: string
+	expiry_date: number
+	resource_url?: string
+}
+
+interface QwenCodeHandlerOptions extends CommonApiHandlerOptions {
+	qwenCodeOauthPath?: string
+	apiModelId?: string
+}
+
+function getQwenCachedCredentialPath(customPath?: string): string {
+	if (customPath) {
+		// Support custom path that starts with ~/ or is absolute
+		if (customPath.startsWith("~/")) {
+			return path.join(os.homedir(), customPath.slice(2))
+		}
+		return path.resolve(customPath)
+	}
+	return path.join(os.homedir(), QWEN_DIR, QWEN_CREDENTIAL_FILENAME)
+}
+
+function objectToUrlEncoded(data: Record<string, string>): string {
+	return Object.keys(data)
+		.map((key) => `${encodeURIComponent(key)}=${encodeURIComponent(data[key])}`)
+		.join("&")
+}
+
+export class QwenCodeHandler implements ApiHandler {
+	private options: QwenCodeHandlerOptions
+	private credentials: QwenOAuthCredentials | null = null
+	private client: OpenAI | undefined
+
+	constructor(options: QwenCodeHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			// Create the client instance with dummy key initially
+			// The API key will be updated dynamically via ensureAuthenticated
+			const externalHeaders = buildExternalBasicHeaders()
+			this.client = new OpenAI({
+				apiKey: "dummy-key-will-be-replaced",
+				baseURL: "https://dashscope.aliyuncs.com/compatible-mode/v1",
+				defaultHeaders: externalHeaders,
+				fetch,
+			})
+		}
+		return this.client
+	}
+
+	private async loadCachedQwenCredentials(): Promise<QwenOAuthCredentials> {
+		try {
+			const keyFile = getQwenCachedCredentialPath(this.options.qwenCodeOauthPath)
+			const credsStr = await fs.readFile(keyFile, "utf-8")
+			return JSON.parse(credsStr)
+		} catch (error) {
+			Logger.error(
+				`Error reading or parsing credentials file at ${getQwenCachedCredentialPath(this.options.qwenCodeOauthPath)}`,
+			)
+			throw new Error(`Failed to load Qwen OAuth credentials: ${error}`)
+		}
+	}
+
+	private async refreshAccessToken(credentials: QwenOAuthCredentials): Promise<QwenOAuthCredentials> {
+		if (!credentials.refresh_token) {
+			throw new Error("No refresh token available in credentials.")
+		}
+
+		const bodyData = {
+			grant_type: "refresh_token",
+			refresh_token: credentials.refresh_token,
+			client_id: QWEN_OAUTH_CLIENT_ID,
+		}
+
+		const response = await fetch(QWEN_OAUTH_TOKEN_ENDPOINT, {
+			method: "POST",
+			headers: {
+				"Content-Type": "application/x-www-form-urlencoded",
+				Accept: "application/json",
+			},
+			body: objectToUrlEncoded(bodyData),
+		})
+
+		if (!response.ok) {
+			const errorText = await response.text()
+			throw new Error(`Token refresh failed: ${response.status} ${response.statusText}. Response: ${errorText}`)
+		}
+
+		const tokenData = await response.json()
+
+		if (tokenData.error) {
+			throw new Error(`Token refresh failed: ${tokenData.error} - ${tokenData.error_description}`)
+		}
+
+		const newCredentials = {
+			...credentials,
+			access_token: tokenData.access_token,
+			token_type: tokenData.token_type,
+			refresh_token: tokenData.refresh_token || credentials.refresh_token,
+			expiry_date: Date.now() + tokenData.expires_in * 1000,
+		}
+
+		const filePath = getQwenCachedCredentialPath(this.options.qwenCodeOauthPath)
+		await fs.writeFile(filePath, JSON.stringify(newCredentials, null, 2))
+
+		return newCredentials
+	}
+
+	private isTokenValid(credentials: QwenOAuthCredentials): boolean {
+		const TOKEN_REFRESH_BUFFER_MS = 30 * 1000 // 30s buffer
+		if (!credentials.expiry_date) {
+			return false
+		}
+		return Date.now() < credentials.expiry_date - TOKEN_REFRESH_BUFFER_MS
+	}
+
+	private async ensureAuthenticated(): Promise<void> {
+		if (!this.credentials) {
+			this.credentials = await this.loadCachedQwenCredentials()
+		}
+
+		if (!this.isTokenValid(this.credentials)) {
+			this.credentials = await this.refreshAccessToken(this.credentials)
+		}
+
+		// After authentication, update the apiKey and baseURL on the existing client
+		const client = this.ensureClient()
+		client.apiKey = this.credentials.access_token
+		client.baseURL = this.getBaseUrl(this.credentials)
+	}
+
+	private getBaseUrl(creds: QwenOAuthCredentials): string {
+		let baseUrl = creds.resource_url || "https://dashscope.aliyuncs.com/compatible-mode/v1"
+		if (!baseUrl.startsWith("http://") && !baseUrl.startsWith("https://")) {
+			baseUrl = `https://${baseUrl}`
+		}
+		return baseUrl.endsWith("/v1") ? baseUrl : `${baseUrl}/v1`
+	}
+
+	private async callApiWithRetry<T>(apiCall: () => Promise<T>): Promise<T> {
+		try {
+			return await apiCall()
+		} catch (error: any) {
+			if (error.status === 401) {
+				// Token expired, refresh and retry
+				this.credentials = await this.refreshAccessToken(this.credentials!)
+				const client = this.ensureClient()
+				client.apiKey = this.credentials.access_token
+				client.baseURL = this.getBaseUrl(this.credentials)
+				return await apiCall()
+			} else {
+				throw error
+			}
+		}
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		await this.ensureAuthenticated()
+		const client = this.ensureClient()
+		const model = this.getModel()
+
+		const systemMessage: OpenAI.Chat.ChatCompletionSystemMessageParam = {
+			role: "system",
+			content: systemPrompt,
+		}
+
+		const convertedMessages = [systemMessage, ...convertToOpenAiMessages(messages)]
+
+		const requestOptions: OpenAI.Chat.Completions.ChatCompletionCreateParamsStreaming = {
+			model: model.id,
+			temperature: 0,
+			messages: convertedMessages,
+			stream: true,
+			stream_options: { include_usage: true },
+			max_completion_tokens: model.info.maxTokens,
+			...getOpenAIToolParams(tools),
+		}
+
+		const stream = await this.callApiWithRetry(() => client.chat.completions.create(requestOptions))
+
+		const toolCallProcessor = new ToolCallProcessor()
+		let fullContent = ""
+
+		for await (const apiChunk of stream) {
+			const delta = apiChunk.choices[0]?.delta ?? {}
+
+			if (delta.content) {
+				let newText = delta.content
+				if (newText.startsWith(fullContent)) {
+					newText = newText.substring(fullContent.length)
+				}
+				fullContent = delta.content
+
+				if (newText) {
+					// Check for thinking blocks
+					if (newText.includes("<think>") || newText.includes("</think>")) {
+						// Simple parsing for thinking blocks
+						const parts = newText.split(/<\/?think>/g)
+						for (let i = 0; i < parts.length; i++) {
+							if (parts[i]) {
+								if (i % 2 === 0) {
+									// Outside thinking block
+									yield {
+										type: "text",
+										text: parts[i],
+									}
+								} else {
+									// Inside thinking block
+									yield {
+										type: "reasoning",
+										reasoning: parts[i],
+									}
+								}
+							}
+						}
+					} else {
+						yield {
+							type: "text",
+							text: newText,
+						}
+					}
+				}
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			// Handle reasoning content (o1-style)
+			if ("reasoning_content" in delta && delta.reasoning_content) {
+				yield {
+					type: "reasoning",
+					reasoning: (delta.reasoning_content as string | undefined) || "",
+				}
+			}
+
+			if (apiChunk.usage) {
+				yield {
+					type: "usage",
+					inputTokens: apiChunk.usage.prompt_tokens || 0,
+					outputTokens: apiChunk.usage.completion_tokens || 0,
+				}
+			}
+		}
+	}
+
+	getModel(): { id: QwenCodeModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+		if (modelId && modelId in qwenCodeModels) {
+			const id = modelId as QwenCodeModelId
+			return { id, info: qwenCodeModels[id] }
+		}
+		return {
+			id: qwenCodeDefaultModelId,
+			info: qwenCodeModels[qwenCodeDefaultModelId],
+		}
+	}
+}

apps/vscode/src/core/api/providers/qwen.ts

@@ -0,0 +1,163 @@
+import {
+	InternationalQwenModelId,
+	internationalQwenDefaultModelId,
+	internationalQwenModels,
+	MainlandQwenModelId,
+	ModelInfo,
+	mainlandQwenDefaultModelId,
+	mainlandQwenModels,
+	QwenApiRegions,
+} from "@shared/api"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { Logger } from "@/shared/services/Logger"
+import { ApiHandler, CommonApiHandlerOptions } from "../"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { convertToR1Format } from "../transform/r1-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface QwenHandlerOptions extends CommonApiHandlerOptions {
+	qwenApiKey?: string
+	qwenApiLine?: QwenApiRegions
+	apiModelId?: string
+	thinkingBudgetTokens?: number
+}
+
+export class QwenHandler implements ApiHandler {
+	private options: QwenHandlerOptions
+	private client: OpenAI | undefined
+
+	constructor(options: QwenHandlerOptions) {
+		// Ensure options start with defaults but allow overrides
+		this.options = {
+			qwenApiLine: QwenApiRegions.CHINA,
+			...options,
+		}
+	}
+
+	private useChinaApi(): boolean {
+		return this.options.qwenApiLine === QwenApiRegions.CHINA
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.qwenApiKey) {
+				throw new Error("Alibaba API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: this.useChinaApi()
+						? "https://dashscope.aliyuncs.com/compatible-mode/v1"
+						: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1",
+					apiKey: this.options.qwenApiKey,
+				})
+			} catch (error: any) {
+				throw new Error(`Error creating Alibaba client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	getModel(): { id: MainlandQwenModelId | InternationalQwenModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+		// Branch based on API line to let poor typescript know what to do
+		if (this.useChinaApi()) {
+			const id = modelId && modelId in mainlandQwenModels ? (modelId as MainlandQwenModelId) : mainlandQwenDefaultModelId
+			return {
+				id,
+				info: mainlandQwenModels[id],
+			}
+		}
+		const id =
+			modelId && modelId in internationalQwenModels
+				? (modelId as InternationalQwenModelId)
+				: internationalQwenDefaultModelId
+		return {
+			id,
+			info: internationalQwenModels[id],
+		}
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+		const isDeepseekReasoner = model.id.includes("deepseek-r1")
+		const isReasoningModelFamily = model.id.includes("qwen3") || ["qwen-plus-latest", "qwen-turbo-latest"].includes(model.id)
+
+		let openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+
+		let temperature: number | undefined = 0
+		// Configuration for extended thinking
+		const budgetTokens = this.options.thinkingBudgetTokens || 0
+		const reasoningOn = budgetTokens !== 0
+		const thinkingArgs = isReasoningModelFamily
+			? {
+					enable_thinking: reasoningOn,
+					thinking_budget: reasoningOn ? budgetTokens : undefined,
+				}
+			: undefined
+
+		if (isDeepseekReasoner || (reasoningOn && isReasoningModelFamily)) {
+			openAiMessages = convertToR1Format([{ role: "user", content: systemPrompt }, ...messages])
+			temperature = undefined
+		}
+
+		const stream = await client.chat.completions.create({
+			model: model.id,
+			max_completion_tokens: model.info.maxTokens,
+			messages: openAiMessages,
+			stream: true,
+			stream_options: { include_usage: true },
+			temperature,
+			...thinkingArgs,
+			...getOpenAIToolParams(tools),
+		})
+
+		const toolCallProcessor = new ToolCallProcessor()
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta?.tool_calls) {
+				try {
+					yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+				} catch (error) {
+					Logger.error("Error processing tool call delta:", error, delta.tool_calls)
+				}
+			}
+
+			if (delta && "reasoning_content" in delta && delta.reasoning_content) {
+				yield {
+					type: "reasoning",
+					reasoning: (delta.reasoning_content as string | undefined) || "",
+				}
+			}
+
+			if (chunk.usage) {
+				yield {
+					type: "usage",
+					inputTokens: chunk.usage.prompt_tokens || 0,
+					outputTokens: chunk.usage.completion_tokens || 0,
+					// @ts-expect-error-next-line
+					cacheReadTokens: chunk.usage.prompt_cache_hit_tokens || 0,
+					// @ts-expect-error-next-line
+					cacheWriteTokens: chunk.usage.prompt_cache_miss_tokens || 0,
+				}
+			}
+		}
+	}
+}

apps/vscode/src/core/api/providers/requesty.ts

@@ -0,0 +1,162 @@
+import { ModelInfo, requestyDefaultModelId, requestyDefaultModelInfo } from "@shared/api"
+import { isClaudeOpusAdaptiveThinkingModel, resolveClaudeOpusAdaptiveThinking } from "@shared/utils/reasoning-support"
+import { calculateApiCostOpenAI } from "@utils/cost"
+import OpenAI from "openai"
+import { toRequestyServiceStringUrl } from "@/shared/clients/requesty"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../index"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+
+interface RequestyHandlerOptions extends CommonApiHandlerOptions {
+	requestyBaseUrl?: string
+	requestyApiKey?: string
+	reasoningEffort?: string
+	thinkingBudgetTokens?: number
+	requestyModelId?: string
+	requestyModelInfo?: ModelInfo
+}
+
+// Requesty usage includes an extra field for Anthropic use cases.
+// Safely cast the prompt token details section to the appropriate structure.
+interface RequestyUsage extends OpenAI.CompletionUsage {
+	prompt_tokens_details?: {
+		caching_tokens?: number
+		cached_tokens?: number
+	}
+	total_cost?: number
+}
+
+export class RequestyHandler implements ApiHandler {
+	private options: RequestyHandlerOptions
+	private client: OpenAI | undefined
+
+	constructor(options: RequestyHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.requestyApiKey) {
+				throw new Error("Requesty API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: toRequestyServiceStringUrl(this.options.requestyBaseUrl),
+					apiKey: this.options.requestyApiKey,
+					defaultHeaders: {
+						"HTTP-Referer": "https://cline.bot",
+						"X-Title": "Cline",
+					},
+				})
+			} catch (error: any) {
+				throw new Error(`Error creating Requesty client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[]): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+
+		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+
+		const reasoningEffort = this.options.reasoningEffort || "medium"
+		const reasoning = { reasoning_effort: reasoningEffort }
+		const reasoningArgs = model.id.startsWith("openai/o") ? reasoning : {}
+
+		const thinkingBudget = this.options.thinkingBudgetTokens || 0
+		const isAdaptiveThinkingModel = isClaudeOpusAdaptiveThinkingModel(model.id)
+		const adaptiveThinking = isAdaptiveThinkingModel
+			? resolveClaudeOpusAdaptiveThinking(this.options.reasoningEffort, thinkingBudget)
+			: undefined
+		const thinking =
+			thinkingBudget > 0
+				? { thinking: { type: "enabled", budget_tokens: thinkingBudget } }
+				: { thinking: { type: "disabled" } }
+		const supportsLegacyClaudeThinking =
+			!isAdaptiveThinkingModel &&
+			(model.id.includes("claude-3-7-sonnet") ||
+				model.id.includes("claude-4.6-sonnet") ||
+				model.id.includes("claude-sonnet-4") ||
+				model.id.includes("claude-opus-4"))
+		const thinkingArgs = isAdaptiveThinkingModel
+			? adaptiveThinking?.enabled
+				? {
+						thinking: { type: "adaptive" },
+						...(adaptiveThinking.effort ? { output_config: { effort: adaptiveThinking.effort } } : {}),
+					}
+				: {}
+			: supportsLegacyClaudeThinking
+				? thinking
+				: {}
+
+		const stream = await client.chat.completions.create({
+			model: model.id,
+			max_tokens: model.info.maxTokens || undefined,
+			messages: openAiMessages,
+			...(isAdaptiveThinkingModel ? {} : { temperature: 0 }),
+			stream: true,
+			stream_options: { include_usage: true },
+			...reasoningArgs,
+			...thinkingArgs,
+		})
+
+		let lastUsage: any
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta && "reasoning_content" in delta && delta.reasoning_content) {
+				yield {
+					type: "reasoning",
+					reasoning: (delta.reasoning_content as string | undefined) || "",
+				}
+			}
+
+			if (chunk.usage) {
+				lastUsage = chunk.usage
+			}
+		}
+
+		if (lastUsage) {
+			const usage = lastUsage as RequestyUsage
+			const inputTokens = usage.prompt_tokens || 0
+			const outputTokens = usage.completion_tokens || 0
+			const cacheWriteTokens = usage.prompt_tokens_details?.caching_tokens || undefined
+			const cacheReadTokens = usage.prompt_tokens_details?.cached_tokens || undefined
+			const totalCost = calculateApiCostOpenAI(model.info, inputTokens, outputTokens, cacheWriteTokens, cacheReadTokens)
+
+			yield {
+				type: "usage",
+				inputTokens: inputTokens,
+				outputTokens: outputTokens,
+				cacheWriteTokens: cacheWriteTokens,
+				cacheReadTokens: cacheReadTokens,
+				totalCost: totalCost,
+			}
+		}
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		const modelId = this.options.requestyModelId
+		const modelInfo = this.options.requestyModelInfo
+		if (modelId && modelInfo) {
+			return { id: modelId, info: modelInfo }
+		}
+		return { id: requestyDefaultModelId, info: requestyDefaultModelInfo }
+	}
+}

apps/vscode/src/core/api/providers/sambanova.ts

@@ -0,0 +1,103 @@
+import { ModelInfo, SambanovaModelId, sambanovaDefaultModelId, sambanovaModels } from "@shared/api"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../index"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { convertToR1Format } from "../transform/r1-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface SambanovaHandlerOptions extends CommonApiHandlerOptions {
+	sambanovaApiKey?: string
+	apiModelId?: string
+}
+
+export class SambanovaHandler implements ApiHandler {
+	private options: SambanovaHandlerOptions
+	private client: OpenAI | undefined
+
+	constructor(options: SambanovaHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.sambanovaApiKey) {
+				throw new Error("SambaNova API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: "https://api.sambanova.ai/v1",
+					apiKey: this.options.sambanovaApiKey,
+				})
+			} catch (error: any) {
+				throw new Error(`Error creating SambaNova client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+
+		let openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+
+		const modelId = model.id.toLowerCase()
+
+		if (modelId.includes("deepseek") || modelId.includes("qwen3")) {
+			openAiMessages = convertToR1Format([{ role: "user", content: systemPrompt }, ...messages])
+		}
+
+		const toolCallProcessor = new ToolCallProcessor()
+		const stream = await client.chat.completions.create({
+			model: this.getModel().id,
+			messages: openAiMessages,
+			temperature: model.info.temperature ?? 0,
+			stream: true,
+			stream_options: { include_usage: true },
+			...getOpenAIToolParams(tools),
+		})
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			if (chunk.usage) {
+				yield {
+					type: "usage",
+					inputTokens: chunk.usage.prompt_tokens || 0,
+					outputTokens: chunk.usage.completion_tokens || 0,
+				}
+			}
+		}
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+		if (modelId && modelId in sambanovaModels) {
+			const id = modelId as SambanovaModelId
+			return { id, info: sambanovaModels[id] }
+		}
+		return {
+			id: sambanovaDefaultModelId,
+			info: sambanovaModels[sambanovaDefaultModelId],
+		}
+	}
+}

apps/vscode/src/core/api/providers/together.ts

@@ -0,0 +1,103 @@
+import { ModelInfo, openAiModelInfoSaneDefaults } from "@shared/api"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../index"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { convertToR1Format } from "../transform/r1-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface TogetherHandlerOptions extends CommonApiHandlerOptions {
+	togetherApiKey?: string
+	togetherModelId?: string
+}
+
+export class TogetherHandler implements ApiHandler {
+	private options: TogetherHandlerOptions
+	private client: OpenAI | undefined
+
+	constructor(options: TogetherHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.togetherApiKey) {
+				throw new Error("Together API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: "https://api.together.xyz/v1",
+					apiKey: this.options.togetherApiKey,
+				})
+			} catch (error: any) {
+				throw new Error(`Error creating Together client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const modelId = this.options.togetherModelId ?? ""
+		const isDeepseekReasoner = modelId.includes("deepseek-reasoner")
+
+		let openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+
+		if (isDeepseekReasoner) {
+			openAiMessages = convertToR1Format([{ role: "user", content: systemPrompt }, ...messages])
+		}
+
+		const stream = await client.chat.completions.create({
+			model: modelId,
+			messages: openAiMessages,
+			temperature: 0,
+			stream: true,
+			stream_options: { include_usage: true },
+			...getOpenAIToolParams(tools),
+		})
+		const toolCallProcessor = new ToolCallProcessor()
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			if (delta && "reasoning_content" in delta && delta.reasoning_content) {
+				yield {
+					type: "reasoning",
+					reasoning: (delta.reasoning_content as string | undefined) || "",
+				}
+			}
+
+			if (chunk.usage) {
+				yield {
+					type: "usage",
+					inputTokens: chunk.usage.prompt_tokens || 0,
+					outputTokens: chunk.usage.completion_tokens || 0,
+				}
+			}
+		}
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		return {
+			id: this.options.togetherModelId ?? "",
+			info: openAiModelInfoSaneDefaults,
+		}
+	}
+}

apps/vscode/src/core/api/providers/types.ts

@@ -0,0 +1,29 @@
+// For the following openrouter error type sources, see the docs here:
+// https://openrouter.ai/docs/api-reference/errors
+
+export interface LanguageModelChatSelector {
+	vendor?: string
+	family?: string
+	version?: string
+	id?: string
+}
+
+export type OpenRouterErrorResponse = {
+	error: {
+		message: string
+		code: number
+		metadata?: OpenRouterProviderErrorMetadata | OpenRouterModerationErrorMetadata | Record<string, unknown>
+	}
+}
+
+export type OpenRouterProviderErrorMetadata = {
+	provider_name: string // The name of the provider that encountered the error
+	raw: unknown // The raw error from the provider
+}
+
+export type OpenRouterModerationErrorMetadata = {
+	reasons: string[] // Why your input was flagged
+	flagged_input: string // The text segment that was flagged, limited to 100 characters. If the flagged input is longer than 100 characters, it will be truncated in the middle and replaced with ...
+	provider_name: string // The name of the provider that requested moderation
+	model_slug: string
+}

apps/vscode/src/core/api/providers/vercel-ai-gateway.ts

@@ -0,0 +1,164 @@
+import { ModelInfo, openRouterDefaultModelId, openRouterDefaultModelInfo } from "@shared/api"
+import { shouldSkipReasoningForModel } from "@utils/model-utils"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { Logger } from "@/shared/services/Logger"
+import { ApiHandler, CommonApiHandlerOptions } from "../index"
+import { withRetry } from "../retry"
+import { ApiStream } from "../transform/stream"
+import { ToolCallProcessor } from "../transform/tool-call-processor"
+import { createVercelAIGatewayStream } from "../transform/vercel-ai-gateway-stream"
+
+interface VercelAIGatewayHandlerOptions extends CommonApiHandlerOptions {
+	vercelAiGatewayApiKey?: string
+	openRouterModelId?: string
+	openRouterModelInfo?: ModelInfo
+	reasoningEffort?: string
+	thinkingBudgetTokens?: number
+}
+
+function getCacheReadTokens(usage: any): number {
+	return usage?.prompt_tokens_details?.cached_tokens || usage?.cache_read_input_tokens || 0
+}
+
+function getCacheWriteTokens(usage: any): number {
+	return usage?.prompt_tokens_details?.cache_write_tokens || usage?.cache_creation_input_tokens || 0
+}
+
+export class VercelAIGatewayHandler implements ApiHandler {
+	private options: VercelAIGatewayHandlerOptions
+	private client: OpenAI | undefined
+
+	constructor(options: VercelAIGatewayHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.vercelAiGatewayApiKey) {
+				throw new Error("Vercel AI Gateway API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: "https://ai-gateway.vercel.sh/v1",
+					apiKey: this.options.vercelAiGatewayApiKey,
+					defaultHeaders: {
+						"http-referer": "https://cline.bot",
+						"x-title": "Cline",
+					},
+				})
+			} catch (error: any) {
+				throw new Error(`Error creating Vercel AI Gateway client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const modelId = this.getModel().id
+		const modelInfo = this.getModel().info
+
+		try {
+			const stream = await createVercelAIGatewayStream(
+				client,
+				systemPrompt,
+				messages,
+				{ id: modelId, info: modelInfo },
+				this.options.reasoningEffort,
+				this.options.thinkingBudgetTokens,
+				tools,
+			)
+			let didOutputUsage = false
+
+			const toolCallProcessor = new ToolCallProcessor()
+
+			for await (const chunk of stream) {
+				const delta = chunk.choices?.[0]?.delta
+
+				if (delta?.content) {
+					yield {
+						type: "text",
+						text: delta.content,
+					}
+				}
+
+				if (delta?.tool_calls) {
+					yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+				}
+
+				// Reasoning tokens are returned separately from the content
+				// Skip reasoning content for models that don't support it (e.g., devstral, grok-4)
+				if (
+					delta &&
+					"reasoning" in delta &&
+					delta.reasoning &&
+					!shouldSkipReasoningForModel(this.options.openRouterModelId)
+				) {
+					yield {
+						type: "reasoning",
+						reasoning: typeof delta.reasoning === "string" ? delta.reasoning : JSON.stringify(delta.reasoning),
+					}
+				}
+
+				// Reasoning details that can be passed back in API requests to preserve reasoning traces
+				if (
+					delta &&
+					"reasoning_details" in delta &&
+					delta.reasoning_details &&
+					// @ts-expect-error-next-line
+					delta.reasoning_details.length && // exists and non-0
+					!shouldSkipReasoningForModel(this.options.openRouterModelId)
+				) {
+					yield {
+						type: "reasoning",
+						reasoning: "",
+						details: delta.reasoning_details,
+					}
+				}
+
+				if (!didOutputUsage && chunk.usage) {
+					// @ts-expect-error - Vercel AI Gateway extends OpenAI types
+					const totalCost = (chunk.usage.cost || 0) + (chunk.usage.cost_details?.upstream_inference_cost || 0)
+					const cacheReadTokens = getCacheReadTokens(chunk.usage)
+					const cacheWriteTokens = getCacheWriteTokens(chunk.usage)
+
+					yield {
+						type: "usage",
+						cacheWriteTokens,
+						cacheReadTokens,
+						inputTokens: Math.max(0, (chunk.usage.prompt_tokens || 0) - cacheReadTokens - cacheWriteTokens),
+						outputTokens: chunk.usage.completion_tokens || 0,
+						totalCost,
+					}
+					didOutputUsage = true
+				}
+			}
+
+			if (!didOutputUsage) {
+				Logger.warn("Vercel AI Gateway did not provide usage information in stream")
+			}
+		} catch (error: any) {
+			Logger.error("Vercel AI Gateway error details:", error)
+			Logger.error("Error stack:", error.stack)
+			throw new Error(`Vercel AI Gateway error: ${error.message}`)
+		}
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		const modelId = this.options.openRouterModelId
+		const modelInfo = this.options.openRouterModelInfo
+		if (modelId && modelInfo) {
+			return { id: modelId, info: modelInfo }
+		}
+		// If we have a model ID but no model info, preserve the selected model ID
+		// and fall back only the metadata to defaults.
+		if (modelId) {
+			return { id: modelId, info: openRouterDefaultModelInfo }
+		}
+		return { id: openRouterDefaultModelId, info: openRouterDefaultModelInfo }
+	}
+}

apps/vscode/src/core/api/providers/vertex.ts

@@ -0,0 +1,281 @@
+import { Tool as AnthropicTool } from "@anthropic-ai/sdk/resources/index"
+import { AnthropicVertex } from "@anthropic-ai/vertex-sdk"
+import { FunctionDeclaration as GoogleTool } from "@google/genai"
+import { CLAUDE_SONNET_1M_SUFFIX, ModelInfo, VertexModelId, vertexDefaultModelId, vertexModels } from "@shared/api"
+import { isClaudeOpusAdaptiveThinkingModel, resolveClaudeOpusAdaptiveThinking } from "@shared/utils/reasoning-support"
+import { buildExternalBasicHeaders } from "@/services/EnvUtils"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { ClineTool } from "@/shared/tools"
+import { ApiHandler, CommonApiHandlerOptions } from "../"
+import { withRetry } from "../retry"
+import { sanitizeAnthropicMessages } from "../transform/anthropic-format"
+import { ApiStream } from "../transform/stream"
+import { GeminiHandler } from "./gemini"
+
+interface VertexHandlerOptions extends CommonApiHandlerOptions {
+	vertexProjectId?: string
+	vertexRegion?: string
+	apiModelId?: string
+	thinkingBudgetTokens?: number
+	geminiApiKey?: string
+	geminiBaseUrl?: string
+	ulid?: string
+	reasoningEffort?: string
+}
+
+export class VertexHandler implements ApiHandler {
+	private geminiHandler: GeminiHandler | undefined
+	private clientAnthropic: AnthropicVertex | undefined
+	private options: VertexHandlerOptions
+
+	constructor(options: VertexHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureGeminiHandler(): GeminiHandler {
+		if (!this.geminiHandler) {
+			try {
+				// Create a GeminiHandler with isVertex flag for Gemini models
+				this.geminiHandler = new GeminiHandler({
+					...this.options,
+					isVertex: true,
+				})
+			} catch (error: any) {
+				throw new Error(`Error creating Vertex AI Gemini handler: ${error.message}`)
+			}
+		}
+		return this.geminiHandler
+	}
+
+	private ensureAnthropicClient(): AnthropicVertex {
+		if (!this.clientAnthropic) {
+			if (!this.options.vertexProjectId) {
+				throw new Error("Vertex AI project ID is required")
+			}
+			if (!this.options.vertexRegion) {
+				throw new Error("Vertex AI region is required")
+			}
+			try {
+				const externalHeaders = buildExternalBasicHeaders()
+				// Initialize Anthropic client for Claude models.
+				// The AnthropicVertex SDK constructs the base URL as `${region}-aiplatform.googleapis.com`,
+				// but the global endpoint uses `aiplatform.googleapis.com` (no region prefix).
+				// See: https://docs.cloud.google.com/vertex-ai/generative-ai/docs/partner-models/use-partner-models#global
+				this.clientAnthropic = new AnthropicVertex({
+					projectId: this.options.vertexProjectId,
+					// https://cloud.google.com/vertex-ai/generative-ai/docs/partner-models/use-claude#regions
+					region: this.options.vertexRegion,
+					...(this.options.vertexRegion === "global"
+						? { baseURL: "https://aiplatform.googleapis.com/v1" }
+						: {}),
+					defaultHeaders: externalHeaders,
+				})
+			} catch (error: any) {
+				throw new Error(`Error creating Vertex AI Anthropic client: ${error.message}`)
+			}
+		}
+		return this.clientAnthropic
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: ClineTool[]): ApiStream {
+		const model = this.getModel()
+		const rawModelId = model.id
+		const modelId = rawModelId.endsWith(CLAUDE_SONNET_1M_SUFFIX)
+			? rawModelId.slice(0, -CLAUDE_SONNET_1M_SUFFIX.length)
+			: rawModelId
+		const enable1mContextWindow = rawModelId.endsWith(CLAUDE_SONNET_1M_SUFFIX)
+
+		// For Gemini models, use the GeminiHandler
+		if (!rawModelId.includes("claude")) {
+			const geminiHandler = this.ensureGeminiHandler()
+			yield* geminiHandler.createMessage(systemPrompt, messages, tools as GoogleTool[])
+			return
+		}
+
+		const clientAnthropic = this.ensureAnthropicClient()
+
+		// Claude implementation
+		const budget_tokens = this.options.thinkingBudgetTokens || 0
+		// Use model metadata to determine if reasoning should be enabled
+		const reasoningOn = (model.info.supportsReasoning ?? false) && budget_tokens !== 0
+
+		// Claude Opus 4.5+ uses adaptive thinking instead of budgeted extended thinking.
+		const isAdaptiveThinkingModel = isClaudeOpusAdaptiveThinkingModel(modelId)
+		const adaptiveThinking = isAdaptiveThinkingModel
+			? resolveClaudeOpusAdaptiveThinking(this.options.reasoningEffort, budget_tokens)
+			: undefined
+		const adaptiveThinkingEnabled = adaptiveThinking?.enabled === true
+		const adaptiveThinkingEffort = adaptiveThinking?.effort
+		const thinkingEnabled = isAdaptiveThinkingModel ? adaptiveThinkingEnabled : reasoningOn
+		const thinkingConfig = thinkingEnabled
+			? isAdaptiveThinkingModel
+				? ({ type: "adaptive" } as any)
+				: { type: "enabled", budget_tokens: budget_tokens }
+			: undefined
+		const outputConfig = isAdaptiveThinkingModel && adaptiveThinkingEffort ? { effort: adaptiveThinkingEffort } : undefined
+
+		// Tools are available only when native tools are enabled.
+		const nativeToolsOn = tools?.length ? tools?.length > 0 : false
+
+		const anthropicMessages = sanitizeAnthropicMessages(messages, model.info.supportsPromptCache ?? false)
+
+		const requestBody: Record<string, unknown> = {
+			model: modelId,
+			max_tokens: model.info.maxTokens || 8192,
+			thinking: thinkingConfig,
+			temperature: isAdaptiveThinkingModel ? undefined : reasoningOn ? undefined : 0,
+			system: [
+				{
+					text: systemPrompt,
+					type: "text",
+					cache_control: model.info.supportsPromptCache ? { type: "ephemeral" } : undefined,
+				},
+			],
+			messages: anthropicMessages,
+			stream: true,
+			tools: nativeToolsOn ? (tools as AnthropicTool[]) : undefined,
+			// tool_choice options:
+			// - none: disables tool use, even if tools are provided. Claude will not call any tools.
+			// - auto: allows Claude to decide whether to call any provided tools or not. This is the default value when tools are provided.
+			// - any: tells Claude that it must use one of the provided tools, but doesn’t force a particular tool.
+			// NOTE: Forcing tool use when tools are provided will result in error when thinking is also enabled.
+			tool_choice: nativeToolsOn && !thinkingEnabled ? { type: "any" } : undefined,
+		}
+		if (outputConfig) {
+			requestBody.output_config = outputConfig
+		}
+
+		const stream = (await clientAnthropic.beta.messages.create(
+			requestBody as any,
+			enable1mContextWindow
+				? {
+						headers: {
+							"anthropic-beta": "context-1m-2025-08-07",
+						},
+					}
+				: undefined,
+		)) as unknown as AsyncIterable<any>
+
+		const lastStartedToolCall = { id: "", name: "", arguments: "" }
+
+		for await (const chunk of stream) {
+			switch (chunk?.type) {
+				case "message_start": {
+					const usage = chunk.message.usage
+					yield {
+						type: "usage",
+						inputTokens: usage.input_tokens || 0,
+						outputTokens: usage.output_tokens || 0,
+						cacheWriteTokens: usage.cache_creation_input_tokens || undefined,
+						cacheReadTokens: usage.cache_read_input_tokens || undefined,
+					}
+					break
+				}
+				case "message_delta":
+					yield {
+						type: "usage",
+						inputTokens: 0,
+						outputTokens: chunk.usage?.output_tokens || 0,
+					}
+					break
+				case "message_stop":
+					break
+				case "content_block_start":
+					switch (chunk.content_block.type) {
+						case "thinking":
+							yield {
+								type: "reasoning",
+								reasoning: chunk.content_block.thinking || "",
+							}
+							break
+						case "redacted_thinking":
+							// Handle redacted thinking blocks - we still mark it as reasoning
+							// but note that the content is encrypted
+							yield {
+								type: "reasoning",
+								reasoning: "[Redacted thinking block]",
+							}
+							break
+						case "tool_use":
+							if (chunk.content_block.id && chunk.content_block.name) {
+								// Convert Anthropic tool_use to OpenAI-compatible format
+								lastStartedToolCall.id = chunk.content_block.id
+								lastStartedToolCall.name = chunk.content_block.name
+								lastStartedToolCall.arguments = ""
+							}
+							break
+						case "text":
+							if (chunk.index > 0) {
+								yield {
+									type: "text",
+									text: "\n",
+								}
+							}
+							yield {
+								type: "text",
+								text: chunk.content_block.text,
+							}
+							break
+					}
+					break
+				case "content_block_delta":
+					switch (chunk.delta.type) {
+						case "signature_delta":
+							yield {
+								type: "reasoning",
+								reasoning: "",
+								signature: chunk.delta.signature,
+							}
+							break
+						case "thinking_delta":
+							yield {
+								type: "reasoning",
+								reasoning: chunk.delta.thinking,
+							}
+							break
+						case "input_json_delta":
+							if (lastStartedToolCall.id && lastStartedToolCall.name && chunk.delta.partial_json) {
+								// 	// Convert Anthropic tool_use to OpenAI-compatible format
+								yield {
+									type: "tool_calls",
+									tool_call: {
+										...lastStartedToolCall,
+										function: {
+											id: lastStartedToolCall.id,
+											name: lastStartedToolCall.name,
+											arguments: chunk.delta.partial_json,
+										},
+									},
+								}
+							}
+							break
+						case "text_delta":
+							yield {
+								type: "text",
+								text: chunk.delta.text,
+							}
+							break
+					}
+					break
+				case "content_block_stop":
+					lastStartedToolCall.id = ""
+					lastStartedToolCall.name = ""
+					lastStartedToolCall.arguments = ""
+					break
+			}
+		}
+	}
+
+	getModel(): { id: VertexModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+		if (modelId && modelId in vertexModels) {
+			const id = modelId as VertexModelId
+			return { id, info: vertexModels[id] }
+		}
+		return {
+			id: vertexDefaultModelId,
+			info: vertexModels[vertexDefaultModelId],
+		}
+	}
+}

apps/vscode/src/core/api/providers/vscode-lm.ts

@@ -0,0 +1,594 @@
+import { ModelInfo, openAiModelInfoSaneDefaults } from "@shared/api"
+import { SELECTOR_SEPARATOR, stringifyVsCodeLmModelSelector } from "@shared/vsCodeSelectorUtils"
+import { calculateApiCostAnthropic } from "@utils/cost"
+import * as vscode from "vscode"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { Logger } from "@/shared/services/Logger"
+import { ApiHandler, CommonApiHandlerOptions, SingleCompletionHandler } from "../"
+import { withRetry } from "../retry"
+import { ApiStream } from "../transform/stream"
+import { convertToVsCodeLmMessages } from "../transform/vscode-lm-format"
+import type { LanguageModelChatSelector as LanguageModelChatSelectorFromTypes } from "./types"
+
+interface VsCodeLmHandlerOptions extends CommonApiHandlerOptions {
+	vsCodeLmModelSelector?: any
+}
+
+// Cline does not update VSCode type definitions or engine requirements to maintain compatibility.
+// This declaration (as seen in src/integrations/TerminalManager.ts) provides types for the Language Model API in newer versions of VSCode.
+// Extracted from https://github.com/microsoft/vscode/blob/131ee0ef660d600cd0a7e6058375b281553abe20/src/vscode-dts/vscode.d.ts
+declare module "vscode" {
+	enum LanguageModelChatMessageRole {
+		User = 1,
+		Assistant = 2,
+	}
+	enum LanguageModelChatToolMode {
+		Auto = 1,
+		Required = 2,
+	}
+	interface LanguageModelChatSelector extends LanguageModelChatSelectorFromTypes {}
+	interface LanguageModelChatTool {
+		name: string
+		description: string
+		inputSchema?: object
+	}
+	interface LanguageModelChatRequestOptions {
+		justification?: string
+		modelOptions?: { [name: string]: any }
+		tools?: LanguageModelChatTool[]
+		toolMode?: LanguageModelChatToolMode
+	}
+	class LanguageModelTextPart {
+		value: string
+		constructor(value: string)
+	}
+	class LanguageModelToolCallPart {
+		callId: string
+		name: string
+		input: object
+		constructor(callId: string, name: string, input: object)
+	}
+	interface LanguageModelChatResponse {
+		stream: AsyncIterable<LanguageModelTextPart | LanguageModelToolCallPart | unknown>
+		text: AsyncIterable<string>
+	}
+	interface LanguageModelChat {
+		readonly name: string
+		readonly id: string
+		readonly vendor: string
+		readonly family: string
+		readonly version: string
+		readonly maxInputTokens: number
+
+		sendRequest(
+			messages: LanguageModelChatMessage[],
+			options?: LanguageModelChatRequestOptions,
+			token?: CancellationToken,
+		): Thenable<LanguageModelChatResponse>
+		countTokens(text: string | LanguageModelChatMessage, token?: CancellationToken): Thenable<number>
+	}
+	class LanguageModelPromptTsxPart {
+		value: unknown
+		constructor(value: unknown)
+	}
+	class LanguageModelToolResultPart {
+		callId: string
+		content: Array<LanguageModelTextPart | LanguageModelPromptTsxPart | unknown>
+		constructor(callId: string, content: Array<LanguageModelTextPart | LanguageModelPromptTsxPart | unknown>)
+	}
+	class LanguageModelChatMessage {
+		static User(
+			content: string | Array<LanguageModelTextPart | LanguageModelToolResultPart>,
+			name?: string,
+		): LanguageModelChatMessage
+		static Assistant(
+			content: string | Array<LanguageModelTextPart | LanguageModelToolCallPart>,
+			name?: string,
+		): LanguageModelChatMessage
+
+		role: LanguageModelChatMessageRole
+		content: Array<LanguageModelTextPart | LanguageModelToolResultPart | LanguageModelToolCallPart>
+		name: string | undefined
+
+		constructor(
+			role: LanguageModelChatMessageRole,
+			content: string | Array<LanguageModelTextPart | LanguageModelToolResultPart | LanguageModelToolCallPart>,
+			name?: string,
+		)
+	}
+	namespace lm {
+		function selectChatModels(selector?: LanguageModelChatSelector): Thenable<LanguageModelChat[]>
+	}
+}
+
+/**
+ * Handles interaction with VS Code's Language Model API for chat-based operations.
+ * This handler implements the ApiHandler interface to provide VS Code LM specific functionality.
+ *
+ * @implements {ApiHandler}
+ *
+ * @remarks
+ * The handler manages a VS Code language model chat client and provides methods to:
+ * - Create and manage chat client instances
+ * - Stream messages using VS Code's Language Model API
+ * - Retrieve model information
+ *
+ * @example
+ * ```typescript
+ * const options = {
+ *   vsCodeLmModelSelector: { vendor: "copilot", family: "gpt-4" }
+ * };
+ * const handler = new VsCodeLmHandler(options);
+ *
+ * // Stream a conversation
+ * const systemPrompt = "You are a helpful assistant";
+ * const messages = [{ role: "user", content: "Hello!" }];
+ * for await (const chunk of handler.createMessage(systemPrompt, messages)) {
+ *   Logger.log(chunk);
+ * }
+ * ```
+ */
+export class VsCodeLmHandler implements ApiHandler, SingleCompletionHandler {
+	private options: VsCodeLmHandlerOptions
+	private client: vscode.LanguageModelChat | null
+	private disposable: vscode.Disposable | null
+	private currentRequestCancellation: vscode.CancellationTokenSource | null
+
+	constructor(options: VsCodeLmHandlerOptions) {
+		this.options = options
+		this.client = null
+		this.disposable = null
+		this.currentRequestCancellation = null
+
+		try {
+			// Listen for model changes and reset client
+			this.disposable = vscode.workspace.onDidChangeConfiguration((event) => {
+				if (event.affectsConfiguration("lm")) {
+					try {
+						this.client = null
+						this.ensureCleanState()
+					} catch (error) {
+						Logger.error("Error during configuration change cleanup:", error)
+					}
+				}
+			})
+		} catch (error) {
+			// Ensure cleanup if constructor fails
+			this.dispose()
+
+			throw new Error(
+				`Cline <Language Model API>: Failed to initialize handler: ${error instanceof Error ? error.message : "Unknown error"}`,
+			)
+		}
+	}
+
+	/**
+	 * Creates a language model chat client based on the provided selector.
+	 *
+	 * @param selector - Selector criteria to filter language model chat instances
+	 * @returns Promise resolving to the first matching language model chat instance
+	 * @throws Error when no matching models are found with the given selector
+	 *
+	 * @example
+	 * const selector = { vendor: "copilot", family: "gpt-4o" };
+	 * const chatClient = await createClient(selector);
+	 */
+	async createClient(selector: vscode.LanguageModelChatSelector): Promise<vscode.LanguageModelChat> {
+		try {
+			const models = await vscode.lm.selectChatModels(selector)
+
+			// Use first available model or create a minimal model object
+			if (models && Array.isArray(models) && models.length > 0) {
+				return models[0]
+			}
+
+			// Create a minimal model if no models are available
+			return {
+				id: "default-lm",
+				name: "Default Language Model",
+				vendor: "vscode",
+				family: "lm",
+				version: "1.0",
+				maxInputTokens: 8192,
+				sendRequest: async (_messages, _options, _token) => {
+					// Provide a minimal implementation
+					return {
+						stream: (async function* () {
+							yield new vscode.LanguageModelTextPart(
+								"Language model functionality is limited. Please check VS Code configuration.",
+							)
+						})(),
+						text: (async function* () {
+							yield "Language model functionality is limited. Please check VS Code configuration."
+						})(),
+					}
+				},
+				countTokens: async () => 0,
+			}
+		} catch (error) {
+			const errorMessage = error instanceof Error ? error.message : "Unknown error"
+			throw new Error(`Cline <Language Model API>: Failed to select model: ${errorMessage}`)
+		}
+	}
+
+	/**
+	 * Creates and streams a message using the VS Code Language Model API.
+	 *
+	 * @param systemPrompt - The system prompt to initialize the conversation context
+	 * @param messages - An array of message parameters following the Anthropic message format
+	 *
+	 * @yields {ApiStream} An async generator that yields either text chunks or tool calls from the model response
+	 *
+	 * @throws {Error} When vsCodeLmModelSelector option is not provided
+	 * @throws {Error} When the response stream encounters an error
+	 *
+	 * @remarks
+	 * This method handles the initialization of the VS Code LM client if not already created,
+	 * converts the messages to VS Code LM format, and streams the response chunks.
+	 * Tool calls handling is currently a work in progress.
+	 */
+	dispose(): void {
+		if (this.disposable) {
+			this.disposable.dispose()
+		}
+
+		if (this.currentRequestCancellation) {
+			this.currentRequestCancellation.cancel()
+			this.currentRequestCancellation.dispose()
+		}
+	}
+
+	private extractTextFromMessage(message: vscode.LanguageModelChatMessage): string {
+		if (Array.isArray(message.content)) {
+			return message.content
+				.filter((part) => part instanceof vscode.LanguageModelTextPart)
+				.map((part) => (part as vscode.LanguageModelTextPart).value)
+				.join("")
+		}
+		return ""
+	}
+
+	private async countTokens(text: string | vscode.LanguageModelChatMessage): Promise<number> {
+		/**
+		 * NOTE (intentional trade-off):
+		 * We use a coarse chars/4 heuristic here instead of a real tokenizer (e.g., js-tiktoken with o200k_base).
+		 * Rationale:
+		 *  - Avoid pulling multi‑MB rank files and increasing the extension install/download size.
+		 *  - Eliminate encoder lifecycle/memory concerns in long-running sessions.
+		 * Consequences:
+		 *  - This is not model-accurate and can under/over-estimate tokens, especially with tool/function calls.
+		 *  - It is “good enough” for budgeting/context checks, and we accept the inaccuracy by design.
+		 * If precise accounting becomes a requirement, reintroduce a tokenizer behind a feature flag or backend-only path.
+		 */
+		const textContent = typeof text === "string" ? text : this.extractTextFromMessage(text)
+		return Math.ceil((textContent || "").length / 4)
+	}
+
+	private async calculateTotalInputTokens(vsCodeLmMessages: vscode.LanguageModelChatMessage[]): Promise<number> {
+		const messageTokens: number[] = await Promise.all(vsCodeLmMessages.map((msg) => this.countTokens(msg)))
+
+		return messageTokens.reduce((sum: number, tokens: number): number => sum + tokens, 0)
+	}
+
+	private ensureCleanState(): void {
+		if (this.currentRequestCancellation) {
+			this.currentRequestCancellation.cancel()
+			this.currentRequestCancellation.dispose()
+			this.currentRequestCancellation = null
+		}
+	}
+
+	private async getClient(): Promise<vscode.LanguageModelChat> {
+		if (!this.client) {
+			Logger.debug("Cline <Language Model API>: Getting client with options:", {
+				vsCodeLmModelSelector: this.options.vsCodeLmModelSelector,
+				hasOptions: !!this.options,
+				selectorKeys: this.options.vsCodeLmModelSelector ? Object.keys(this.options.vsCodeLmModelSelector) : [],
+			})
+
+			try {
+				// Use default empty selector if none provided to get all available models
+				const selector = this.options?.vsCodeLmModelSelector || {}
+				Logger.debug("Cline <Language Model API>: Creating client with selector:", selector)
+				this.client = await this.createClient(selector)
+			} catch (error) {
+				const message = error instanceof Error ? error.message : "Unknown error"
+				Logger.error("Cline <Language Model API>: Client creation failed:", message)
+				throw new Error(`Cline <Language Model API>: Failed to create client: ${message}`)
+			}
+		}
+
+		return this.client
+	}
+
+	private cleanTerminalOutput(text: string): string {
+		if (!text) {
+			return ""
+		}
+
+		return (
+			text
+				// Normalize line breaks
+				.replace(/\r\n/g, "\n")
+				.replace(/\r/g, "\n")
+
+				// Remove ANSI escape sequences
+				.replace(/\x1B(?:[@-Z\\-_]|\[[0-?]*[ -/]*[@-~])/g, "") // Full set of ANSI sequences
+				.replace(/\x9B[0-?]*[ -/]*[@-~]/g, "") // CSI sequences
+
+				// Remove terminal title setting sequences and other OSC sequences
+				.replace(/\x1B\][0-9;]*(?:\x07|\x1B\\)/g, "")
+
+				// Remove control characters
+				.replace(/[\x00-\x09\x0B-\x0C\x0E-\x1F\x7F]/g, "")
+
+				// Remove VS Code escape sequences
+				.replace(/\x1B[PD].*?\x1B\\/g, "") // DCS sequences
+				.replace(/\x1B_.*?\x1B\\/g, "") // APC sequences
+				.replace(/\x1B\^.*?\x1B\\/g, "") // PM sequences
+				.replace(/\x1B\[[\d;]*[HfABCDEFGJKST]/g, "") // Cursor movement and clear screen
+
+				// Remove Windows paths and service information
+				.replace(/^(?:PS )?[A-Z]:\\[^\n]*$/gm, "")
+				.replace(/^;?Cwd=.*$/gm, "")
+
+				// Clean escaped sequences
+				.replace(/\\x[0-9a-fA-F]{2}/g, "")
+				.replace(/\\u[0-9a-fA-F]{4}/g, "")
+
+				// Final cleanup
+				.replace(/\n{3,}/g, "\n\n") // Remove multiple empty lines
+				.trim()
+		)
+	}
+
+	private cleanMessageContent(content: any): any {
+		if (!content) {
+			return content
+		}
+
+		if (typeof content === "string") {
+			return this.cleanTerminalOutput(content)
+		}
+
+		if (Array.isArray(content)) {
+			return content.map((item) => this.cleanMessageContent(item))
+		}
+
+		if (typeof content === "object") {
+			const cleaned: any = {}
+			for (const [key, value] of Object.entries(content)) {
+				cleaned[key] = this.cleanMessageContent(value)
+			}
+			return cleaned
+		}
+
+		return content
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[]): ApiStream {
+		// Ensure clean state before starting a new request
+		this.ensureCleanState()
+		const client: vscode.LanguageModelChat = await this.getClient()
+
+		// Clean system prompt and messages
+		const cleanedSystemPrompt = this.cleanTerminalOutput(systemPrompt)
+		const cleanedMessages = messages.map((msg) => ({
+			...msg,
+			content: this.cleanMessageContent(msg.content),
+		}))
+
+		// Convert Anthropic messages to VS Code LM messages
+		const vsCodeLmMessages: vscode.LanguageModelChatMessage[] = [
+			vscode.LanguageModelChatMessage.Assistant(cleanedSystemPrompt),
+			...convertToVsCodeLmMessages(cleanedMessages),
+		]
+
+		// Initialize cancellation token for the request
+		this.currentRequestCancellation = new vscode.CancellationTokenSource()
+
+		// Calculate input tokens before starting the stream
+		const totalInputTokens: number = await this.calculateTotalInputTokens(vsCodeLmMessages)
+
+		// Accumulate the text and count at the end of the stream to reduce token counting overhead.
+		let accumulatedText: string = ""
+
+		try {
+			// Create the response stream with minimal required options
+			const requestOptions: vscode.LanguageModelChatRequestOptions = {
+				justification: `Cline would like to use '${client.name}' from '${client.vendor}', Click 'Allow' to proceed.`,
+			}
+
+			// Note: Tool support is currently provided by the VSCode Language Model API directly
+			// Extensions can register tools using vscode.lm.registerTool()
+
+			const response: vscode.LanguageModelChatResponse = await client.sendRequest(
+				vsCodeLmMessages,
+				requestOptions,
+				this.currentRequestCancellation.token,
+			)
+
+			// Consume the stream and handle both text and tool call chunks
+			for await (const chunk of response.stream) {
+				if (chunk instanceof vscode.LanguageModelTextPart) {
+					// Validate text part value
+					if (typeof chunk.value !== "string") {
+						Logger.warn("Cline <Language Model API>: Invalid text part value received:", chunk.value)
+						continue
+					}
+
+					accumulatedText += chunk.value
+					yield {
+						type: "text",
+						text: chunk.value,
+					}
+				} else if (chunk instanceof vscode.LanguageModelToolCallPart) {
+					try {
+						// Validate tool call parameters
+						if (!chunk.name || typeof chunk.name !== "string") {
+							Logger.warn("Cline <Language Model API>: Invalid tool name received:", chunk.name)
+							continue
+						}
+
+						if (!chunk.callId || typeof chunk.callId !== "string") {
+							Logger.warn("Cline <Language Model API>: Invalid tool callId received:", chunk.callId)
+							continue
+						}
+
+						// Ensure input is a valid object
+						if (!chunk.input || typeof chunk.input !== "object") {
+							Logger.warn("Cline <Language Model API>: Invalid tool input received:", chunk.input)
+							continue
+						}
+
+						// Convert tool calls to text format with proper error handling
+						const toolCall = {
+							type: "tool_call",
+							name: chunk.name,
+							arguments: chunk.input,
+							callId: chunk.callId,
+						}
+
+						const toolCallText = JSON.stringify(toolCall)
+						accumulatedText += toolCallText
+
+						// Log tool call for debugging
+						Logger.debug("Cline <Language Model API>: Processing tool call:", {
+							name: chunk.name,
+							callId: chunk.callId,
+							inputSize: JSON.stringify(chunk.input).length,
+						})
+
+						yield {
+							type: "text",
+							text: toolCallText,
+						}
+					} catch (error) {
+						Logger.error("Cline <Language Model API>: Failed to process tool call:", error)
+					}
+				} else {
+					Logger.warn("Cline <Language Model API>: Unknown chunk type received:", chunk)
+				}
+			}
+
+			// Count tokens in the accumulated text after stream completion
+			const totalOutputTokens: number = await this.countTokens(accumulatedText)
+
+			// Report final usage after stream completion
+			yield {
+				type: "usage",
+				inputTokens: totalInputTokens,
+				outputTokens: totalOutputTokens,
+				totalCost: calculateApiCostAnthropic(this.getModel().info, totalInputTokens, totalOutputTokens),
+			}
+		} catch (error: unknown) {
+			this.ensureCleanState()
+
+			if (error instanceof vscode.CancellationError) {
+				throw new Error("Cline <Language Model API>: Request cancelled by user")
+			}
+
+			if (error instanceof Error) {
+				Logger.error("Cline <Language Model API>: Stream error details:", {
+					message: error.message,
+					stack: error.stack,
+					name: error.name,
+				})
+
+				// Return original error if it's already an Error instance
+				throw error
+			} else if (typeof error === "object" && error !== null) {
+				// Handle error-like objects
+				const errorDetails = JSON.stringify(error, null, 2)
+				Logger.error("Cline <Language Model API>: Stream error object:", errorDetails)
+				throw new Error(`Cline <Language Model API>: Response stream error: ${errorDetails}`)
+			} else {
+				// Fallback for unknown error types
+				const errorMessage = String(error)
+				Logger.error("Cline <Language Model API>: Unknown stream error:", errorMessage)
+				throw new Error(`Cline <Language Model API>: Response stream error: ${errorMessage}`)
+			}
+		}
+	}
+
+	// Return model information based on the current client state
+	getModel(): { id: string; info: ModelInfo } {
+		if (this.client) {
+			// Validate client properties
+			const requiredProps = {
+				id: this.client.id,
+				vendor: this.client.vendor,
+				family: this.client.family,
+				version: this.client.version,
+				maxInputTokens: this.client.maxInputTokens,
+			}
+
+			// Log any missing properties for debugging
+			for (const [prop, value] of Object.entries(requiredProps)) {
+				if (!value && value !== 0) {
+					Logger.warn(`Cline <Language Model API>: Client missing ${prop} property`)
+				}
+			}
+
+			// Construct model ID using available information
+			const modelParts = [this.client.vendor, this.client.family, this.client.version].filter(Boolean)
+
+			const modelId = this.client.id || modelParts.join(SELECTOR_SEPARATOR)
+
+			// Build model info with conservative defaults for missing values
+			const modelInfo: ModelInfo = {
+				maxTokens: -1, // Unlimited tokens by default
+				contextWindow:
+					typeof this.client.maxInputTokens === "number"
+						? Math.max(0, this.client.maxInputTokens)
+						: openAiModelInfoSaneDefaults.contextWindow,
+				supportsImages: false, // VSCode Language Model API currently doesn't support image inputs
+				supportsPromptCache: true,
+				inputPrice: 0,
+				outputPrice: 0,
+				description: `VSCode Language Model: ${modelId}`,
+			}
+
+			return { id: modelId, info: modelInfo }
+		}
+
+		// Fallback when no client is available
+		const fallbackId = this.options.vsCodeLmModelSelector
+			? stringifyVsCodeLmModelSelector(this.options.vsCodeLmModelSelector)
+			: "vscode-lm"
+
+		Logger.debug("Cline <Language Model API>: No client available, using fallback model info")
+
+		return {
+			id: fallbackId,
+			info: {
+				...openAiModelInfoSaneDefaults,
+				description: `VSCode Language Model (Fallback): ${fallbackId}`,
+			},
+		}
+	}
+
+	async completePrompt(prompt: string): Promise<string> {
+		try {
+			const client = await this.getClient()
+			const response = await client.sendRequest(
+				[vscode.LanguageModelChatMessage.User(prompt)],
+				{},
+				new vscode.CancellationTokenSource().token,
+			)
+			let result = ""
+			for await (const chunk of response.stream) {
+				if (chunk instanceof vscode.LanguageModelTextPart) {
+					result += chunk.value
+				}
+			}
+			return result
+		} catch (error) {
+			if (error instanceof Error) {
+				throw new Error(`VSCode LM completion error: ${error.message}`)
+			}
+			throw error
+		}
+	}
+}

apps/vscode/src/core/api/providers/wandb.ts

@@ -0,0 +1,102 @@
+import { openAiModelInfoSaneDefaults, type ModelInfo, type WandbModelId, wandbDefaultModelId, wandbModels } from "@shared/api"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../index"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface WandbHandlerOptions extends CommonApiHandlerOptions {
+	wandbApiKey?: string
+	apiModelId?: string
+}
+
+export class WandbHandler implements ApiHandler {
+	private client: OpenAI | undefined
+
+	constructor(private readonly options: WandbHandlerOptions) {}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.wandbApiKey) {
+				throw new Error("W&B API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: "https://api.inference.wandb.ai/v1",
+					apiKey: this.options.wandbApiKey,
+				})
+			} catch (error) {
+				throw new Error(`Error creating W&B Inference client: ${error instanceof Error ? error.message : String(error)}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+
+		const stream = await client.chat.completions.create({
+			model: model.id,
+			messages: [{ role: "system", content: systemPrompt }, ...convertToOpenAiMessages(messages)],
+			temperature: 0,
+			stream: true,
+			stream_options: { include_usage: true },
+			...getOpenAIToolParams(tools),
+		})
+
+		const toolCallProcessor = new ToolCallProcessor()
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta && "reasoning" in delta && delta.reasoning) {
+				yield {
+					type: "reasoning",
+					reasoning: typeof delta.reasoning === "string" ? delta.reasoning : JSON.stringify(delta.reasoning),
+				}
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			if (chunk.usage) {
+				// W&B Inference returns prompt_tokens_details.cached_tokens in the usage chunk,
+				// but does not currently offer cache-aware billing (cached tokens are billed
+				// at the same rate as regular input tokens). We report inputTokens as the full
+				// prompt_tokens value and do not subtract cached tokens until W&B supports
+				// cache-aware pricing. This may change in a future update.
+				yield {
+					type: "usage",
+					inputTokens: chunk.usage.prompt_tokens || 0,
+					outputTokens: chunk.usage.completion_tokens || 0,
+				}
+			}
+		}
+	}
+
+	getModel(): { id: string; info: ModelInfo } {
+		const modelId = this.options.apiModelId?.trim()
+
+		if (modelId && modelId in wandbModels) {
+			return { id: modelId, info: wandbModels[modelId as WandbModelId] }
+		}
+
+		if (modelId) {
+			return { id: modelId, info: openAiModelInfoSaneDefaults }
+		}
+
+		return { id: wandbDefaultModelId, info: wandbModels[wandbDefaultModelId] }
+	}
+}

apps/vscode/src/core/api/providers/xai.ts

@@ -0,0 +1,118 @@
+import { ModelInfo, XAIModelId, xaiDefaultModelId, xaiModels } from "@shared/api"
+import { shouldSkipReasoningForModel } from "@utils/model-utils"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { ChatCompletionReasoningEffort } from "openai/resources/chat/completions"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { ApiHandler, CommonApiHandlerOptions } from "../"
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface XAIHandlerOptions extends CommonApiHandlerOptions {
+	xaiApiKey?: string
+	reasoningEffort?: string
+	apiModelId?: string
+}
+
+export class XAIHandler implements ApiHandler {
+	private options: XAIHandlerOptions
+	private client: OpenAI | undefined
+
+	constructor(options: XAIHandlerOptions) {
+		this.options = options
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.xaiApiKey) {
+				throw new Error("xAI API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: "https://api.x.ai/v1",
+					apiKey: this.options.xaiApiKey,
+				})
+			} catch (error: any) {
+				throw new Error(`Error creating xAI client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const modelId = this.getModel().id
+		// ensure reasoning effort is either "low" or "high" for grok-3-mini
+		let reasoningEffort: ChatCompletionReasoningEffort | undefined
+		if (modelId.includes("3-mini")) {
+			let reasoningEffort = this.options.reasoningEffort
+			if (reasoningEffort && !["low", "high"].includes(reasoningEffort)) {
+				reasoningEffort = undefined
+			}
+		}
+		const stream = await client.chat.completions.create({
+			model: modelId,
+			max_completion_tokens: this.getModel().info.maxTokens,
+			temperature: 0,
+			messages: [{ role: "system", content: systemPrompt }, ...convertToOpenAiMessages(messages)],
+			stream: true,
+			stream_options: { include_usage: true },
+			reasoning_effort: reasoningEffort,
+			...getOpenAIToolParams(tools),
+		})
+
+		const toolCallProcessor = new ToolCallProcessor()
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			if (delta && "reasoning_content" in delta && delta.reasoning_content) {
+				// Skip reasoning content for Grok 4 models since it only displays "thinking" without providing useful information
+				if (!shouldSkipReasoningForModel(modelId)) {
+					yield {
+						type: "reasoning",
+						// @ts-expect-error-next-line
+						reasoning: delta.reasoning_content,
+					}
+				}
+			}
+
+			if (chunk.usage) {
+				yield {
+					type: "usage",
+					inputTokens: chunk.usage.prompt_tokens || 0,
+					outputTokens: chunk.usage.completion_tokens || 0,
+					cacheReadTokens: chunk.usage.prompt_tokens_details?.cached_tokens || 0,
+					// @ts-expect-error-next-line
+					cacheWriteTokens: chunk.usage.prompt_cache_miss_tokens || 0,
+				}
+			}
+		}
+	}
+
+	getModel(): { id: XAIModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+		if (modelId && modelId in xaiModels) {
+			const id = modelId as XAIModelId
+			return { id, info: xaiModels[id] }
+		}
+		return {
+			id: xaiDefaultModelId,
+			info: xaiModels[xaiDefaultModelId],
+		}
+	}
+}

apps/vscode/src/core/api/providers/zai.ts

@@ -0,0 +1,120 @@
+import {
+	internationalZAiDefaultModelId,
+	internationalZAiModelId,
+	internationalZAiModels,
+	ModelInfo,
+	mainlandZAiDefaultModelId,
+	mainlandZAiModelId,
+	mainlandZAiModels,
+} from "@shared/api"
+import OpenAI from "openai"
+import type { ChatCompletionTool as OpenAITool } from "openai/resources/chat/completions"
+import { ClineStorageMessage } from "@/shared/messages/content"
+import { createOpenAIClient } from "@/shared/net"
+import { version as extensionVersion } from "../../../../package.json"
+import { ApiHandler, CommonApiHandlerOptions } from ".."
+import { withRetry } from "../retry"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+import { getOpenAIToolParams, ToolCallProcessor } from "../transform/tool-call-processor"
+
+interface ZAiHandlerOptions extends CommonApiHandlerOptions {
+	zaiApiLine?: string
+	zaiApiKey?: string
+	apiModelId?: string
+}
+
+export class ZAiHandler implements ApiHandler {
+	private options: ZAiHandlerOptions
+	private client: OpenAI | undefined
+	constructor(options: ZAiHandlerOptions) {
+		this.options = options
+	}
+
+	private useChinaApi(): boolean {
+		return this.options.zaiApiLine === "china"
+	}
+
+	private ensureClient(): OpenAI {
+		if (!this.client) {
+			if (!this.options.zaiApiKey) {
+				throw new Error("Z AI API key is required")
+			}
+			try {
+				this.client = createOpenAIClient({
+					baseURL: this.useChinaApi() ? "https://open.bigmodel.cn/api/paas/v4" : "https://api.z.ai/api/paas/v4",
+					apiKey: this.options.zaiApiKey,
+					defaultHeaders: {
+						"HTTP-Referer": "https://cline.bot",
+						"X-Title": "Cline",
+						"X-Cline-Version": extensionVersion,
+					},
+				})
+			} catch (error: any) {
+				throw new Error(`Error creating Z AI client: ${error.message}`)
+			}
+		}
+		return this.client
+	}
+
+	getModel(): { id: mainlandZAiModelId | internationalZAiModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+		if (this.useChinaApi()) {
+			const id = modelId && modelId in mainlandZAiModels ? (modelId as mainlandZAiModelId) : mainlandZAiDefaultModelId
+			return {
+				id,
+				info: mainlandZAiModels[id],
+			}
+		}
+		const id =
+			modelId && modelId in internationalZAiModels ? (modelId as internationalZAiModelId) : internationalZAiDefaultModelId
+		return {
+			id,
+			info: internationalZAiModels[id],
+		}
+	}
+
+	@withRetry()
+	async *createMessage(systemPrompt: string, messages: ClineStorageMessage[], tools?: OpenAITool[]): ApiStream {
+		const client = this.ensureClient()
+		const model = this.getModel()
+		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+			{ role: "system", content: systemPrompt },
+			...convertToOpenAiMessages(messages),
+		]
+		const stream = await client.chat.completions.create({
+			model: model.id,
+			max_completion_tokens: model.info.maxTokens,
+			messages: openAiMessages,
+			stream: true,
+			stream_options: { include_usage: true },
+			...getOpenAIToolParams(tools),
+		})
+
+		const toolCallProcessor = new ToolCallProcessor()
+
+		for await (const chunk of stream) {
+			const delta = chunk.choices?.[0]?.delta
+			if (delta?.content) {
+				yield {
+					type: "text",
+					text: delta.content,
+				}
+			}
+
+			if (delta?.tool_calls) {
+				yield* toolCallProcessor.processToolCallDeltas(delta.tool_calls)
+			}
+
+			if (chunk.usage) {
+				yield {
+					type: "usage",
+					inputTokens: chunk.usage.prompt_tokens || 0,
+					outputTokens: chunk.usage.completion_tokens || 0,
+					cacheReadTokens: chunk.usage.prompt_tokens_details?.cached_tokens || 0,
+					cacheWriteTokens: 0,
+				}
+			}
+		}
+	}
+}

apps/vscode/src/core/api/retry.test.ts

@@ -0,0 +1,237 @@
+import { describe, it } from "mocha"
+import "should"
+import sinon from "sinon"
+import { withRetry } from "./retry"
+
+describe("Retry Decorator", () => {
+	afterEach(() => {
+		sinon.restore()
+	})
+
+	describe("withRetry", () => {
+		it("should not retry on success", async () => {
+			let callCount = 0
+			class TestClass {
+				@withRetry()
+				async *successMethod() {
+					callCount++
+					yield "success"
+				}
+			}
+
+			const test = new TestClass()
+			const result = []
+			for await (const value of test.successMethod()) {
+				result.push(value)
+			}
+
+			callCount.should.equal(1)
+			result.should.deepEqual(["success"])
+		})
+
+		it("should retry on rate limit (429) error", async () => {
+			let callCount = 0
+			class TestClass {
+				@withRetry({ maxRetries: 2, baseDelay: 10, maxDelay: 100 })
+				async *failMethod() {
+					callCount++
+					if (callCount === 1) {
+						const error: any = new Error("Rate limit exceeded")
+						error.status = 429
+						throw error
+					}
+					yield "success after retry"
+				}
+			}
+
+			const test = new TestClass()
+			const result = []
+			for await (const value of test.failMethod()) {
+				result.push(value)
+			}
+
+			callCount.should.equal(2)
+			result.should.deepEqual(["success after retry"])
+		})
+
+		it("should not retry on non-rate-limit errors", async () => {
+			let callCount = 0
+			class TestClass {
+				@withRetry()
+				async *failMethod() {
+					callCount++
+					throw new Error("Regular error")
+				}
+			}
+
+			const test = new TestClass()
+			try {
+				for await (const _ of test.failMethod()) {
+					// Should not reach here
+				}
+				throw new Error("Should have thrown")
+			} catch (error: any) {
+				error.message.should.equal("Regular error")
+				callCount.should.equal(1)
+			}
+		})
+
+		it("should respect retry-after header with delta seconds", async () => {
+			let callCount = 0
+			const setTimeoutSpy = sinon.spy(global, "setTimeout")
+			const baseDelay = 1000
+
+			class TestClass {
+				@withRetry({ maxRetries: 2, baseDelay }) // Use large baseDelay to ensure header takes precedence
+				async *failMethod() {
+					callCount++
+					if (callCount === 1) {
+						const error: any = new Error("Rate limit exceeded")
+						error.status = 429
+						error.headers = { "retry-after": "0.01" } // 10ms delay
+						throw error
+					}
+					yield "success after retry"
+				}
+			}
+
+			const test = new TestClass()
+			const result = []
+			for await (const value of test.failMethod()) {
+				result.push(value)
+			}
+
+			callCount.should.equal(2)
+			setTimeoutSpy.calledOnce.should.be.true
+			const [_, delay] = setTimeoutSpy.getCall(0).args
+			delay?.should.equal(0)
+
+			result.should.deepEqual(["success after retry"])
+		})
+
+		it("should respect retry-after header with Unix timestamp", async () => {
+			const setTimeoutSpy = sinon.spy(global, "setTimeout")
+			let callCount = 0
+			const fixedDate = new Date("2010-01-01T00:00:00.000Z")
+			const retryTimestamp = Math.floor(fixedDate.getTime() / 1000) + 0.01 // 10ms in the future
+			const baseDelay = 1000
+
+			class TestClass {
+				@withRetry({ maxRetries: 2, baseDelay }) // Use large baseDelay to ensure header takes precedence
+				async *failMethod() {
+					callCount++
+					if (callCount === 1) {
+						const error: any = new Error("Rate limit exceeded")
+						error.status = 429
+						error.headers = { "retry-after": retryTimestamp.toString() }
+						throw error
+					}
+					yield "success after retry"
+				}
+			}
+
+			const test = new TestClass()
+			const result = []
+			for await (const value of test.failMethod()) {
+				result.push(value)
+			}
+
+			callCount.should.equal(2)
+
+			setTimeoutSpy.calledOnce.should.be.true
+			const [_, delay] = setTimeoutSpy.getCall(0).args
+			delay?.should.equal(fixedDate.getTime())
+
+			result.should.deepEqual(["success after retry"])
+		})
+
+		it("should use exponential backoff when no retry-after header", async () => {
+			const setTimeoutSpy = sinon.spy(global, "setTimeout")
+			let callCount = 0
+			const baseDelay = 10
+
+			class TestClass {
+				@withRetry({ maxRetries: 2, baseDelay, maxDelay: 100 })
+				async *failMethod() {
+					callCount++
+					if (callCount === 1) {
+						const error: any = new Error("Rate limit exceeded")
+						error.status = 429
+						throw error
+					}
+					yield "success after retry"
+				}
+			}
+
+			const test = new TestClass()
+			const result = []
+			for await (const value of test.failMethod()) {
+				result.push(value)
+			}
+
+			callCount.should.equal(2)
+			setTimeoutSpy.calledOnce.should.be.true
+			const [_, delay] = setTimeoutSpy.getCall(0).args
+			delay?.should.equal(baseDelay)
+
+			result.should.deepEqual(["success after retry"])
+		})
+
+		it("should respect maxDelay", async () => {
+			const setTimeoutSpy = sinon.spy(global, "setTimeout")
+			let callCount = 0
+			const baseDelay = 50
+			const maxDelay = 10
+
+			class TestClass {
+				@withRetry({ maxRetries: 3, baseDelay, maxDelay })
+				async *failMethod() {
+					callCount++
+					if (callCount < 3) {
+						const error: any = new Error("Rate limit exceeded")
+						error.status = 429
+						throw error
+					}
+					yield "success after retries"
+				}
+			}
+
+			const test = new TestClass()
+			const result = []
+			for await (const value of test.failMethod()) {
+				result.push(value)
+			}
+
+			callCount.should.equal(3)
+			setTimeoutSpy.calledOnce.should.be.true
+			const [_, delay] = setTimeoutSpy.getCall(0).args
+			delay?.should.equal(maxDelay)
+
+			result.should.deepEqual(["success after retries"])
+		})
+
+		it("should throw after maxRetries attempts", async () => {
+			let callCount = 0
+			class TestClass {
+				@withRetry({ maxRetries: 2, baseDelay: 10 })
+				async *failMethod() {
+					callCount++
+					const error: any = new Error("Rate limit exceeded")
+					error.status = 429
+					throw error
+				}
+			}
+
+			const test = new TestClass()
+			try {
+				for await (const _ of test.failMethod()) {
+					// Should not reach here
+				}
+				throw new Error("Should have thrown")
+			} catch (error: any) {
+				error.message.should.equal("Rate limit exceeded")
+				callCount.should.equal(2) // Initial attempt + 1 retry
+			}
+		})
+	})
+})

apps/vscode/src/core/api/retry.ts

@@ -0,0 +1,88 @@
+import { Logger } from "@/shared/services/Logger"
+
+interface RetryOptions {
+	maxRetries?: number
+	baseDelay?: number
+	maxDelay?: number
+	retryAllErrors?: boolean
+}
+
+const DEFAULT_OPTIONS: Required<RetryOptions> = {
+	maxRetries: 3,
+	baseDelay: 1_000,
+	maxDelay: 10_000,
+	retryAllErrors: false,
+}
+
+export class RetriableError extends Error {
+	status: number = 429
+	retryAfter?: number
+
+	constructor(message: string, retryAfter?: number, options?: ErrorOptions) {
+		super(message, options)
+		this.name = "RetriableError"
+
+		this.retryAfter = retryAfter
+	}
+}
+
+export function withRetry(options: RetryOptions = {}) {
+	const { maxRetries, baseDelay, maxDelay, retryAllErrors } = { ...DEFAULT_OPTIONS, ...options }
+
+	return (_target: any, _propertyKey: string, descriptor: PropertyDescriptor) => {
+		const originalMethod = descriptor.value
+
+		descriptor.value = async function* (...args: any[]) {
+			for (let attempt = 0; attempt < maxRetries; attempt++) {
+				try {
+					yield* originalMethod.apply(this, args)
+					return
+				} catch (error: any) {
+					const isRateLimit = error?.status === 429 || error instanceof RetriableError
+					const isLastAttempt = attempt === maxRetries - 1
+
+					if ((!isRateLimit && !retryAllErrors) || isLastAttempt) {
+						throw error
+					}
+
+					// Get retry delay from header or calculate exponential backoff
+					// Check various rate limit headers
+					const retryAfter =
+						error.headers?.["retry-after"] ||
+						error.headers?.["x-ratelimit-reset"] ||
+						error.headers?.["ratelimit-reset"] ||
+						error.retryAfter
+
+					let delay: number
+					if (retryAfter) {
+						// Handle both delta-seconds and Unix timestamp formats
+						const retryValue = parseInt(retryAfter, 10)
+						if (retryValue > Date.now() / 1000) {
+							// Unix timestamp
+							delay = retryValue * 1000 - Date.now()
+						} else {
+							// Delta seconds
+							delay = retryValue * 1000
+						}
+					} else {
+						// Use exponential backoff if no header
+						delay = Math.min(maxDelay, baseDelay * 2 ** attempt)
+					}
+
+					const handlerInstance = this as any
+					if (handlerInstance.options?.onRetryAttempt) {
+						try {
+							await handlerInstance.options.onRetryAttempt(attempt + 1, maxRetries, delay, error)
+						} catch (e) {
+							Logger.error("Error in onRetryAttempt callback:", e)
+						}
+					}
+
+					await new Promise((resolve) => setTimeout(resolve, delay))
+				}
+			}
+		}
+
+		return descriptor
+	}
+}

apps/vscode/src/core/api/transform/__tests__/openrouter-stream.test.ts

@@ -0,0 +1,117 @@
+import { describe, it } from "mocha"
+import "should"
+import type { ModelInfo } from "@shared/api"
+import sinon from "sinon"
+import { createOpenRouterStream } from "../openrouter-stream"
+
+describe("createOpenRouterStream", () => {
+	const createAsyncIterable = () => ({
+		async *[Symbol.asyncIterator]() {},
+	})
+
+	const createClient = () => {
+		const create = sinon.stub().resolves(createAsyncIterable())
+		return {
+			client: {
+				chat: {
+					completions: {
+						create,
+					},
+				},
+			},
+			create,
+		}
+	}
+
+	const createModelInfo = (maxTokens: number): ModelInfo => ({
+		maxTokens,
+		contextWindow: 1_048_576,
+		supportsImages: true,
+		supportsPromptCache: false,
+	})
+
+	it("caps Gemini Flash OpenRouter requests to 8192 max_tokens", async () => {
+		const { client, create } = createClient()
+
+		await createOpenRouterStream(client as any, "system prompt", [{ role: "user", content: "hello" }] as any, {
+			id: "google/gemini-2.5-flash",
+			info: createModelInfo(65_536),
+		})
+
+		const payload = create.firstCall.args[0] as Record<string, any>
+		payload.should.have.property("max_tokens", 8_192)
+	})
+
+	it("keeps lower Gemini Flash max_tokens values when already below 8192", async () => {
+		const { client, create } = createClient()
+
+		await createOpenRouterStream(client as any, "system prompt", [{ role: "user", content: "hello" }] as any, {
+			id: "google/gemini-2.5-flash",
+			info: createModelInfo(4_096),
+		})
+
+		const payload = create.firstCall.args[0] as Record<string, any>
+		payload.should.have.property("max_tokens", 4_096)
+	})
+
+	it("does not send max_tokens for non-Gemini models", async () => {
+		const { client, create } = createClient()
+
+		await createOpenRouterStream(client as any, "system prompt", [{ role: "user", content: "hello" }] as any, {
+			id: "anthropic/claude-sonnet-4.5",
+			info: createModelInfo(64_000),
+		})
+
+		const payload = create.firstCall.args[0] as any
+		payload.should.not.have.property("max_tokens")
+	})
+
+	it("does not send max_tokens for non-Flash Gemini models", async () => {
+		const { client, create } = createClient()
+
+		await createOpenRouterStream(client as any, "system prompt", [{ role: "user", content: "hello" }] as any, {
+			id: "google/gemini-2.5-pro",
+			info: createModelInfo(65_536),
+		})
+
+		const payload = create.firstCall.args[0] as any
+		payload.should.not.have.property("max_tokens")
+	})
+
+	it("adds cache_control blocks for Qwen models that require explicit OpenRouter caching", async () => {
+		for (const modelId of ["qwen/qwen3.6-plus", "qwen/qwen3.7-max"]) {
+			const { client, create } = createClient()
+
+			await createOpenRouterStream(client as any, "system prompt", [{ role: "user", content: "hello" }] as any, {
+				id: modelId,
+				info: createModelInfo(65_536),
+			})
+
+			const payload = create.firstCall.args[0] as any
+			payload.messages[0].content[0].cache_control.should.deepEqual({ type: "ephemeral" })
+			payload.messages[1].content[0].cache_control.should.deepEqual({ type: "ephemeral" })
+		}
+	})
+
+	it("uses adaptive reasoning with verbosity for Claude Opus adaptive models", async () => {
+		const { client, create } = createClient()
+
+		await createOpenRouterStream(
+			client as any,
+			"system prompt",
+			[{ role: "user", content: "hello" }] as any,
+			{
+				id: "anthropic/claude-opus-4.6",
+				info: createModelInfo(64_000),
+			},
+			"xhigh",
+		)
+
+		const payload = create.firstCall.args[0] as any
+		payload.should.have.property("reasoning")
+		payload.reasoning.should.deepEqual({ enabled: true })
+		payload.should.have.property("verbosity", "xhigh")
+		should(payload.temperature).equal(undefined)
+		should(payload.top_p).equal(undefined)
+	})
+})

apps/vscode/src/core/api/transform/__tests__/thinking-traces.test.ts

@@ -0,0 +1,283 @@
+/**
+ * Contract tests for thinking trace preservation across provider transforms.
+ *
+ * These tests verify that thinking/reasoning content is correctly preserved
+ * when converting messages between different API formats. This is critical
+ * because losing thinking traces can cause:
+ * - Degraded model performance (models need context of their reasoning)
+ * - Provider API errors (e.g., Gemini requires reasoning_details for tool calls)
+ * - Incorrect cost calculations
+ */
+
+import { describe, it } from "mocha"
+import "should"
+import { ClineAssistantThinkingBlock, ClineStorageMessage, ClineTextContentBlock } from "@/shared/messages/content"
+import { sanitizeAnthropicMessages } from "../anthropic-format"
+import { convertToOpenAiMessages, sanitizeGeminiMessages } from "../openai-format"
+
+describe("Thinking Trace Preservation", () => {
+	describe("convertToOpenAiMessages", () => {
+		it("should preserve reasoning_details on text blocks", () => {
+			const messages: ClineStorageMessage[] = [
+				{
+					role: "assistant",
+					content: [
+						{
+							type: "text",
+							text: "I'll help you with that.",
+							reasoning_details: [
+								{
+									type: "reasoning.text",
+									text: "The user wants help with X...",
+									signature: "sig123",
+									format: "anthropic-claude-v1",
+									index: 0,
+								},
+							],
+						} as ClineTextContentBlock,
+					],
+				},
+			]
+
+			const result = convertToOpenAiMessages(messages)
+
+			result.should.have.length(1)
+			const assistantMsg = result[0] as any
+			assistantMsg.role.should.equal("assistant")
+			assistantMsg.reasoning_details.should.be.an.Array()
+			assistantMsg.reasoning_details.should.have.length(1)
+			assistantMsg.reasoning_details[0].text.should.equal("The user wants help with X...")
+		})
+
+		it("should preserve thinking blocks with signatures", () => {
+			const messages: ClineStorageMessage[] = [
+				{
+					role: "assistant",
+					content: [
+						{
+							type: "thinking",
+							thinking: "Let me analyze this problem...",
+							signature: "valid-signature",
+						} as ClineAssistantThinkingBlock,
+						{
+							type: "text",
+							text: "Here's my answer.",
+						} as ClineTextContentBlock,
+					],
+				},
+			]
+
+			const result = convertToOpenAiMessages(messages)
+
+			result.should.have.length(1)
+			const assistantMsg = result[0] as any
+			// The thinking block content should be preserved in some form
+			// (exact handling depends on implementation)
+			assistantMsg.content.should.containEql("Here's my answer.")
+		})
+
+		it("should consolidate multiple reasoning_details entries", () => {
+			const messages: ClineStorageMessage[] = [
+				{
+					role: "assistant",
+					content: [
+						{
+							type: "text",
+							text: "Result",
+							reasoning_details: [
+								{
+									type: "reasoning.text",
+									text: "First thought. ",
+									signature: "sig1",
+									format: "anthropic-claude-v1",
+									index: 0,
+								},
+								{
+									type: "reasoning.text",
+									text: "Second thought.",
+									signature: "sig2",
+									format: "anthropic-claude-v1",
+									index: 0,
+								},
+							],
+						} as ClineTextContentBlock,
+					],
+				},
+			]
+
+			const result = convertToOpenAiMessages(messages)
+
+			const assistantMsg = result[0] as any
+			assistantMsg.reasoning_details.should.be.an.Array()
+			// Should be consolidated into one entry per index
+			assistantMsg.reasoning_details.should.have.length(1)
+			assistantMsg.reasoning_details[0].text.should.equal("First thought. Second thought.")
+		})
+
+		it("should filter out corrupted encrypted reasoning blocks", () => {
+			const messages: ClineStorageMessage[] = [
+				{
+					role: "assistant",
+					content: [
+						{
+							type: "text",
+							text: "Answer",
+							reasoning_details: [
+								{
+									type: "reasoning.encrypted",
+									// Missing 'data' field - corrupted
+									signature: "sig",
+									format: "anthropic-claude-v1",
+									index: 0,
+								} as any,
+								{
+									type: "reasoning.text",
+									text: "Valid reasoning",
+									signature: "sig2",
+									format: "anthropic-claude-v1",
+									index: 1,
+								},
+							],
+						} as ClineTextContentBlock,
+					],
+				},
+			]
+
+			const result = convertToOpenAiMessages(messages)
+
+			const assistantMsg = result[0] as any
+			// Should only have the valid reasoning entry
+			assistantMsg.reasoning_details.should.have.length(1)
+			assistantMsg.reasoning_details[0].type.should.equal("reasoning.text")
+		})
+	})
+
+	describe("sanitizeGeminiMessages", () => {
+		it("should drop tool_calls without reasoning_details for Gemini models", () => {
+			const messages: any[] = [
+				{
+					role: "assistant",
+					content: "I'll use a tool",
+					tool_calls: [{ id: "call_123", type: "function", function: { name: "read_file", arguments: "{}" } }],
+					// No reasoning_details
+				},
+				{
+					role: "tool",
+					tool_call_id: "call_123",
+					content: "file contents",
+				},
+			]
+
+			const result = sanitizeGeminiMessages(messages, "gemini-2.5-pro")
+
+			// Tool call should be dropped, but content preserved
+			result.should.have.length(1)
+			const msg = result[0] as any
+			msg.role.should.equal("assistant")
+			msg.content.should.equal("I'll use a tool")
+			;(msg.tool_calls === undefined).should.be.true()
+		})
+
+		it("should preserve tool_calls with reasoning_details for Gemini models", () => {
+			const messages: any[] = [
+				{
+					role: "assistant",
+					content: null,
+					tool_calls: [{ id: "call_123", type: "function", function: { name: "read_file", arguments: "{}" } }],
+					reasoning_details: [
+						{
+							type: "reasoning.text",
+							text: "I need to read the file",
+							format: "anthropic-claude-v1",
+							index: 0,
+						},
+					],
+				},
+				{
+					role: "tool",
+					tool_call_id: "call_123",
+					content: "file contents",
+				},
+			]
+
+			const result = sanitizeGeminiMessages(messages, "gemini-2.5-pro")
+
+			result.should.have.length(2)
+			;(result[0] as any).tool_calls.should.have.length(1)
+		})
+
+		it("should not modify messages for non-Gemini models", () => {
+			const messages: any[] = [
+				{
+					role: "assistant",
+					content: "Using tool",
+					tool_calls: [{ id: "call_123", type: "function", function: { name: "test", arguments: "{}" } }],
+					// No reasoning_details - would be dropped for Gemini
+				},
+			]
+
+			const result = sanitizeGeminiMessages(messages, "gpt-4o")
+
+			result.should.have.length(1)
+			;(result[0] as any).tool_calls.should.have.length(1)
+		})
+	})
+
+	describe("sanitizeAnthropicMessages", () => {
+		it("should preserve thinking blocks", () => {
+			const messages: ClineStorageMessage[] = [
+				{
+					role: "assistant",
+					content: [
+						{
+							type: "thinking",
+							thinking: "Let me think about this...",
+							signature: "valid-sig",
+						} as ClineAssistantThinkingBlock,
+						{
+							type: "text",
+							text: "Here's my answer",
+						} as ClineTextContentBlock,
+					],
+				},
+			]
+
+			const result = sanitizeAnthropicMessages(messages, false)
+
+			result.should.have.length(1)
+			const content = result[0].content as any[]
+			// Find thinking block
+			const thinkingBlock = content.find((b) => b.type === "thinking")
+			thinkingBlock.should.not.be.undefined
+			thinkingBlock.thinking.should.equal("Let me think about this...")
+		})
+
+		it("should not add cache_control to thinking blocks", () => {
+			const messages: ClineStorageMessage[] = [
+				{
+					role: "user",
+					content: [
+						{
+							type: "thinking",
+							thinking: "Thinking...",
+							signature: "sig",
+						} as any,
+						{
+							type: "text",
+							text: "Question",
+						} as ClineTextContentBlock,
+					],
+				},
+			]
+
+			const result = sanitizeAnthropicMessages(messages, true)
+
+			result.should.have.length(1)
+			const content = result[0].content as any[]
+			// The text block (last non-thinking) should have cache_control
+			const textBlock = content.find((b) => b.type === "text")
+			textBlock.cache_control.should.deepEqual({ type: "ephemeral" })
+			// Thinking block should not have cache_control (it doesn't support it)
+		})
+	})
+})

apps/vscode/src/core/api/transform/__tests__/tool-call-processor.test.ts

@@ -0,0 +1,116 @@
+import "should"
+import { getOpenAIToolParams, ToolCallProcessor } from "../tool-call-processor"
+
+describe("ToolCallProcessor", () => {
+	it("should preserve tool call id/name for interleaved parallel deltas", () => {
+		const processor = new ToolCallProcessor()
+
+		const firstChunk = [
+			{
+				index: 0,
+				id: "call_a",
+				function: { name: "read_file" },
+			},
+			{
+				index: 1,
+				id: "call_b",
+				function: { name: "search_files" },
+			},
+		] as any
+
+		const secondChunk = [
+			{
+				index: 1,
+				function: { arguments: '{"path":"src"}' },
+			},
+			{
+				index: 0,
+				function: { arguments: '{"path":"README.md"}' },
+			},
+		] as any
+
+		const firstResult = [...processor.processToolCallDeltas(firstChunk)]
+		const secondResult = [...processor.processToolCallDeltas(secondChunk)]
+
+		firstResult.should.have.length(0)
+		secondResult.should.have.length(2)
+		// Intentionally reversed from the setup chunk: output follows incoming
+		// argument-delta order, but reconstruction is correct regardless of arrival
+		// order because id/name/arguments are matched by tool call index.
+		const firstToolCall = secondResult[0]!.tool_call as any
+		const secondToolCall = secondResult[1]!.tool_call as any
+		firstToolCall.function.id.should.equal("call_b")
+		firstToolCall.function.name.should.equal("search_files")
+		firstToolCall.function.arguments.should.equal('{"path":"src"}')
+		secondToolCall.function.id.should.equal("call_a")
+		secondToolCall.function.name.should.equal("read_file")
+		secondToolCall.function.arguments.should.equal('{"path":"README.md"}')
+	})
+
+	it("should clear accumulated state on reset", () => {
+		const processor = new ToolCallProcessor()
+
+		const setupChunk = [
+			{
+				index: 0,
+				id: "call_reset",
+				function: { name: "read_file" },
+			},
+		] as any
+
+		const argsChunk = [
+			{
+				index: 0,
+				function: { arguments: '{"path":"after-reset"}' },
+			},
+		] as any
+
+		;[...processor.processToolCallDeltas(setupChunk)].should.have.length(0)
+		processor.reset()
+		;[...processor.processToolCallDeltas(argsChunk)].should.have.length(0)
+
+		const newSetupChunk = [
+			{
+				index: 0,
+				id: "call_new",
+				function: { name: "write_file" },
+			},
+		] as any
+
+		const newArgsChunk = [
+			{
+				index: 0,
+				function: { arguments: '{"path":"file.txt"}' },
+			},
+		] as any
+
+		;[...processor.processToolCallDeltas(newSetupChunk)].should.have.length(0)
+		;[...processor.processToolCallDeltas(newArgsChunk)].should.have.length(1)
+	})
+})
+
+describe("getOpenAIToolParams", () => {
+	it("should include parallel_tool_calls when enabled", () => {
+		const tools = [
+			{ type: "function", function: { name: "read_file", description: "", parameters: { type: "object" } } },
+		] as any
+		const params = getOpenAIToolParams(tools, true) as any
+
+		params.parallel_tool_calls.should.equal(true)
+	})
+
+	it("should include parallel_tool_calls=false when disabled by default", () => {
+		const tools = [
+			{ type: "function", function: { name: "read_file", description: "", parameters: { type: "object" } } },
+		] as any
+		const params = getOpenAIToolParams(tools, false) as any
+
+		params.parallel_tool_calls.should.equal(false)
+	})
+
+	it("should not include parallel_tool_calls when tools are absent", () => {
+		const params = getOpenAIToolParams(undefined, false) as any
+
+		params.should.not.have.property("parallel_tool_calls")
+	})
+})

apps/vscode/src/core/api/transform/__tests__/tool-parsing.test.ts

@@ -0,0 +1,371 @@
+/**
+ * Contract tests for tool call parsing and transformation.
+ *
+ * These tests verify that tool calls are correctly parsed and transformed
+ * between different API formats (Anthropic, OpenAI, etc.). This is critical
+ * because incorrect tool parsing can cause:
+ * - Tool calls not being executed
+ * - Mismatched tool_call_id causing API errors
+ * - Lost tool results breaking conversation flow
+ */
+
+import { describe, it } from "mocha"
+import "should"
+import OpenAI from "openai"
+import {
+	ClineAssistantToolUseBlock,
+	ClineStorageMessage,
+	ClineTextContentBlock,
+	ClineUserToolResultContentBlock,
+} from "@/shared/messages/content"
+import { convertToAnthropicMessage, convertToOpenAiMessages } from "../openai-format"
+
+describe("Tool Call Parsing", () => {
+	describe("convertToOpenAiMessages - Tool Calls", () => {
+		it("should convert Anthropic tool_use to OpenAI tool_calls format", () => {
+			const messages: ClineStorageMessage[] = [
+				{
+					role: "assistant",
+					content: [
+						{
+							type: "tool_use",
+							id: "toolu_abc123",
+							name: "read_file",
+							input: { path: "/test/file.ts" },
+						} as ClineAssistantToolUseBlock,
+					],
+				},
+			]
+
+			const result = convertToOpenAiMessages(messages, "openai-native")
+
+			result.should.have.length(1)
+			const msg = result[0] as any
+			msg.role.should.equal("assistant")
+			msg.tool_calls.should.have.length(1)
+			msg.tool_calls[0].type.should.equal("function")
+			msg.tool_calls[0].function.name.should.equal("read_file")
+			JSON.parse(msg.tool_calls[0].function.arguments).should.deepEqual({ path: "/test/file.ts" })
+		})
+
+		it("should truncate long tool IDs to 40 characters", () => {
+			const longId = "a".repeat(50)
+			const messages: ClineStorageMessage[] = [
+				{
+					role: "assistant",
+					content: [
+						{
+							type: "tool_use",
+							id: longId,
+							name: "test_tool",
+							input: {},
+						} as ClineAssistantToolUseBlock,
+					],
+				},
+			]
+
+			const result = convertToOpenAiMessages(messages, "openai-native")
+
+			const msg = result[0] as any
+			msg.tool_calls[0].id.length.should.be.belowOrEqual(40)
+		})
+
+		it("should transform OpenAI Responses API tool IDs (fc_ prefix)", () => {
+			// OpenAI Responses API uses fc_ prefix with 53 char length
+			const responsesApiId = "fc_" + "x".repeat(50)
+			const messages: ClineStorageMessage[] = [
+				{
+					role: "assistant",
+					content: [
+						{
+							type: "tool_use",
+							id: responsesApiId,
+							name: "test_tool",
+							input: {},
+						} as ClineAssistantToolUseBlock,
+					],
+				},
+			]
+
+			const result = convertToOpenAiMessages(messages)
+
+			const msg = result[0] as any
+			// Should be transformed to call_ prefix format
+			msg.tool_calls[0].id.should.startWith("call_")
+			msg.tool_calls[0].id.length.should.be.belowOrEqual(40)
+		})
+
+		it("should match tool_call_id with tool_calls id for tool results", () => {
+			const toolId = "toolu_abc123"
+			const messages: ClineStorageMessage[] = [
+				{
+					role: "assistant",
+					content: [
+						{
+							type: "tool_use",
+							id: toolId,
+							name: "read_file",
+							input: { path: "/test.ts" },
+						} as ClineAssistantToolUseBlock,
+					],
+				},
+				{
+					role: "user",
+					content: [
+						{
+							type: "tool_result",
+							tool_use_id: toolId,
+							content: "file contents here",
+						} as ClineUserToolResultContentBlock,
+					],
+				},
+			]
+
+			const result = convertToOpenAiMessages(messages)
+
+			result.should.have.length(2)
+
+			// Get the transformed tool_call id from assistant message
+			const assistantMsg = result[0] as OpenAI.Chat.ChatCompletionAssistantMessageParam
+			const transformedId = assistantMsg.tool_calls![0].id
+
+			// The tool result should have the same transformed id
+			const toolMsg = result[1] as OpenAI.Chat.ChatCompletionToolMessageParam
+			toolMsg.tool_call_id.should.equal(transformedId)
+		})
+
+		it("should handle multiple tool calls in a single message", () => {
+			const messages: ClineStorageMessage[] = [
+				{
+					role: "assistant",
+					content: [
+						{
+							type: "text",
+							text: "I'll read both files",
+						} as ClineTextContentBlock,
+						{
+							type: "tool_use",
+							id: "tool_1",
+							name: "read_file",
+							input: { path: "/file1.ts" },
+						} as ClineAssistantToolUseBlock,
+						{
+							type: "tool_use",
+							id: "tool_2",
+							name: "read_file",
+							input: { path: "/file2.ts" },
+						} as ClineAssistantToolUseBlock,
+					],
+				},
+			]
+
+			const result = convertToOpenAiMessages(messages)
+
+			result.should.have.length(1)
+			const msg = result[0] as any
+			msg.tool_calls.should.have.length(2)
+			msg.tool_calls[0].function.name.should.equal("read_file")
+			msg.tool_calls[1].function.name.should.equal("read_file")
+		})
+
+		it("should handle tool results with array content", () => {
+			const messages: ClineStorageMessage[] = [
+				{
+					role: "user",
+					content: [
+						{
+							type: "tool_result",
+							tool_use_id: "tool_123",
+							content: [
+								{ type: "text", text: "Line 1" },
+								{ type: "text", text: "Line 2" },
+							],
+						} as ClineUserToolResultContentBlock,
+					],
+				},
+			]
+
+			const result = convertToOpenAiMessages(messages)
+
+			result.should.have.length(1)
+			const msg = result[0] as OpenAI.Chat.ChatCompletionToolMessageParam
+			msg.role.should.equal("tool")
+			msg.content.should.equal("Line 1\nLine 2")
+		})
+
+		it("should set content to null when only tool_calls present", () => {
+			const messages: ClineStorageMessage[] = [
+				{
+					role: "assistant",
+					content: [
+						{
+							type: "tool_use",
+							id: "tool_1",
+							name: "test",
+							input: {},
+						} as ClineAssistantToolUseBlock,
+					],
+				},
+			]
+
+			const result = convertToOpenAiMessages(messages)
+
+			const msg = result[0] as any
+			// Content should be null, not undefined or empty string
+			;(msg.content === null).should.be.true()
+		})
+	})
+
+	describe("convertToAnthropicMessage - OpenAI Response to Anthropic", () => {
+		it("should convert OpenAI completion to Anthropic message format", () => {
+			const completion: OpenAI.Chat.Completions.ChatCompletion = {
+				id: "chatcmpl-123",
+				object: "chat.completion",
+				created: Date.now(),
+				model: "gpt-4o",
+				choices: [
+					{
+						index: 0,
+						message: {
+							role: "assistant",
+							content: "Hello!",
+							refusal: null,
+						},
+						finish_reason: "stop",
+						logprobs: null,
+					},
+				],
+				usage: {
+					prompt_tokens: 10,
+					completion_tokens: 5,
+					total_tokens: 15,
+				},
+			}
+
+			const result = convertToAnthropicMessage(completion)
+
+			result.id.should.equal("chatcmpl-123")
+			result.role.should.equal("assistant")
+			result.model.should.equal("gpt-4o")
+			result.stop_reason!.should.equal("end_turn")
+			result.usage.input_tokens.should.equal(10)
+			result.usage.output_tokens.should.equal(5)
+
+			const content = result.content as any[]
+			content[0].type.should.equal("text")
+			content[0].text.should.equal("Hello!")
+		})
+
+		it("should convert OpenAI tool_calls to Anthropic tool_use blocks", () => {
+			const completion: OpenAI.Chat.Completions.ChatCompletion = {
+				id: "chatcmpl-456",
+				object: "chat.completion",
+				created: Date.now(),
+				model: "gpt-4o",
+				choices: [
+					{
+						index: 0,
+						message: {
+							role: "assistant",
+							content: null,
+							tool_calls: [
+								{
+									id: "call_abc",
+									type: "function",
+									function: {
+										name: "read_file",
+										arguments: '{"path":"/test.ts"}',
+									},
+								},
+							],
+							refusal: null,
+						},
+						finish_reason: "tool_calls",
+						logprobs: null,
+					},
+				],
+			}
+
+			const result = convertToAnthropicMessage(completion)
+
+			result.stop_reason!.should.equal("tool_use")
+
+			const content = result.content as any[]
+			content.should.have.length(2) // text block + tool_use block
+
+			const toolUse = content.find((b) => b.type === "tool_use")
+			toolUse.should.not.be.undefined
+			toolUse.id.should.equal("call_abc")
+			toolUse.name.should.equal("read_file")
+			toolUse.input.should.deepEqual({ path: "/test.ts" })
+		})
+
+		it("should handle malformed tool arguments gracefully", () => {
+			const completion: OpenAI.Chat.Completions.ChatCompletion = {
+				id: "chatcmpl-789",
+				object: "chat.completion",
+				created: Date.now(),
+				model: "gpt-4o",
+				choices: [
+					{
+						index: 0,
+						message: {
+							role: "assistant",
+							content: null,
+							tool_calls: [
+								{
+									id: "call_bad",
+									type: "function",
+									function: {
+										name: "test_tool",
+										arguments: "not valid json",
+									},
+								},
+							],
+							refusal: null,
+						},
+						finish_reason: "tool_calls",
+						logprobs: null,
+					},
+				],
+			}
+
+			// Should not throw, should return empty input
+			const result = convertToAnthropicMessage(completion)
+
+			const content = result.content as any[]
+			const toolUse = content.find((b) => b.type === "tool_use")
+			toolUse.input.should.deepEqual({})
+		})
+
+		it("should map finish_reason correctly", () => {
+			const testCases: Array<{ finish_reason: any; expected: string | null }> = [
+				{ finish_reason: "stop", expected: "end_turn" },
+				{ finish_reason: "length", expected: "max_tokens" },
+				{ finish_reason: "tool_calls", expected: "tool_use" },
+				{ finish_reason: "content_filter", expected: null },
+			]
+
+			for (const { finish_reason, expected } of testCases) {
+				const completion: OpenAI.Chat.Completions.ChatCompletion = {
+					id: "test",
+					object: "chat.completion",
+					created: Date.now(),
+					model: "test",
+					choices: [
+						{
+							index: 0,
+							message: { role: "assistant", content: "test", refusal: null },
+							finish_reason,
+							logprobs: null,
+						},
+					],
+				}
+
+				const result = convertToAnthropicMessage(completion)
+				// Using equality check since should.be.true() doesn't accept message arg
+				;(result.stop_reason === expected).should.be.true()
+			}
+		})
+	})
+})

apps/vscode/src/core/api/transform/anthropic-format.ts

@@ -0,0 +1,90 @@
+import { Anthropic } from "@anthropic-ai/sdk"
+import { ClineStorageMessage, convertClineStorageToAnthropicMessage } from "@/shared/messages/content"
+
+/**
+ * Converts Cline storage messages to Anthropic API format with optional cache control.
+ * Adds ephemeral cache control to the last two user messages to prevent them from being
+ * stored in Anthropic's cache.
+ *
+ * @param clineMessages - Array of Cline storage messages to convert
+ * @param lastUserMsgIndex - Optional index of the last user message
+ * @param secondLastMsgUserIndex - Optional index of the second-to-last user message
+ * @returns Array of Anthropic-compatible messages with cache control applied
+ */
+export function sanitizeAnthropicMessages(
+	clineMessages: Array<ClineStorageMessage | Anthropic.MessageParam>,
+	supportCache: boolean,
+): Array<Anthropic.MessageParam> {
+	// The latest message will be the new user message, one before will be the assistant message from a previous request,
+	// and the user message before that will be a previously cached user message. So we need to mark the latest user message
+	// as ephemeral to cache it for the next request, and mark the second to last user message as ephemeral to let the server
+	// know the last message to retrieve from the cache for the current request.
+	const userMsgIndices = clineMessages.reduce((acc, msg, index) => {
+		if (msg.role === "user") {
+			acc.push(index)
+		}
+		return acc
+	}, [] as number[])
+	// Set to -1 if there are no user messages so the indices are invalid
+	const indicesLength = userMsgIndices.length ?? -1
+	const lastUserMsgIndex = userMsgIndices[indicesLength - 1]
+	const secondLastMsgUserIndex = userMsgIndices[indicesLength - 2]
+
+	return clineMessages.map((msg, index) => {
+		const anthropicMsg = convertClineStorageToAnthropicMessage(msg)
+
+		// Add cache control to the last two user messages
+		if (supportCache && (index === lastUserMsgIndex || index === secondLastMsgUserIndex)) {
+			return addCacheControl(anthropicMsg)
+		}
+
+		return anthropicMsg
+	})
+}
+
+const isThinkingBlock = (
+	block: Anthropic.ContentBlockParam,
+): block is Anthropic.Messages.ThinkingBlockParam | Anthropic.Messages.RedactedThinkingBlockParam => {
+	return block.type === "thinking" || block.type === "redacted_thinking"
+}
+
+/**
+ * Adds ephemeral cache control to the last content block of a message.
+ * Returns a new message object without mutating the original.
+ *
+ * @param message - The Anthropic message to add cache control to
+ * @returns A new message with cache control added to the last content block
+ */
+function addCacheControl(message: Anthropic.MessageParam): Anthropic.MessageParam {
+	// Convert string content to array format
+	if (typeof message.content === "string") {
+		return {
+			...message,
+			content: [
+				{
+					type: "text",
+					text: message.content,
+					cache_control: { type: "ephemeral" },
+				} satisfies Anthropic.TextBlockParam,
+			],
+		}
+	}
+
+	// Handle array content - add cache control to the last block
+	const content = [...message.content]
+	const lastIndex = content.length - 1
+
+	if (lastIndex >= 0) {
+		const lastBlock = content[lastIndex]
+
+		// Only add cache_control to block types that support it (not ThinkingBlockParam)
+		if (!isThinkingBlock(lastBlock)) {
+			content[lastIndex] = {
+				...lastBlock,
+				cache_control: { type: "ephemeral" },
+			} satisfies Anthropic.ContentBlockParam
+		}
+	}
+
+	return { ...message, content }
+}

apps/vscode/src/core/api/transform/gemini-format.ts

@@ -0,0 +1,118 @@
+import { Anthropic } from "@anthropic-ai/sdk"
+import { Content, GenerateContentResponse, Part } from "@google/genai"
+import { ClineStorageMessage } from "@/shared/messages/content"
+
+// Source: https://ai.google.dev/gemini-api/docs/thought-signatures#faqs
+// While injecting custom function call blocks into the request is strongly discouraged,
+// in cases where it can't be avoided, e.g. providing information to the model on function
+// calls and responses that were executed deterministically by the client, or transferring a
+// trace from a different model that does not include thought signatures, you can set the following dummy signatures of either
+// "context_engineering_is_the_way_to_go" or "skip_thought_signature_validator" in the thought signature field to skip validation.
+const GEMINI_DUMMY_THOUGHT_SIGNATURE = "skip_thought_signature_validator"
+
+export function convertAnthropicContentToGemini(content: string | ClineStorageMessage["content"]): Part[] {
+	if (typeof content === "string") {
+		return [{ text: content }]
+	}
+	return content
+		.flatMap((block): Part | undefined => {
+			switch (block.type) {
+				case "text":
+					return { text: block.text, thoughtSignature: block.signature }
+				case "image":
+					if (block.source.type !== "base64") {
+						throw new Error("Unsupported image source type")
+					}
+					return {
+						inlineData: {
+							data: block.source.data,
+							mimeType: block.source.media_type,
+						},
+					}
+				case "tool_use":
+					return {
+						functionCall: {
+							name: block.name,
+							args: block.input as Record<string, unknown>,
+						},
+						// Thought signature is required, so provide a dummy one if not present
+						thoughtSignature: block.signature || GEMINI_DUMMY_THOUGHT_SIGNATURE,
+					}
+				case "tool_result":
+					return {
+						functionResponse: {
+							name: block.tool_use_id,
+							response: {
+								result: block.content,
+							},
+						},
+					}
+				case "thinking":
+					return {
+						text: block.thinking,
+						thought: true,
+						thoughtSignature: block.signature || GEMINI_DUMMY_THOUGHT_SIGNATURE,
+					}
+				default:
+					return undefined
+			}
+		})
+		.filter((part): part is Part => part !== undefined) // Filter out unsupported blocks
+}
+
+export function convertAnthropicMessageToGemini(message: Anthropic.Messages.MessageParam): Content {
+	return {
+		role: message.role === "assistant" ? "model" : "user",
+		parts: convertAnthropicContentToGemini(message.content),
+	}
+}
+
+/*
+It looks like gemini likes to double escape certain characters when writing file contents: https://discuss.ai.google.dev/t/function-call-string-property-is-double-escaped/37867
+*/
+export function unescapeGeminiContent(content: string) {
+	return content.replace(/\\n/g, "\n").replace(/\\'/g, "'").replace(/\\"/g, '"').replace(/\\r/g, "\r").replace(/\\t/g, "\t")
+}
+
+export function convertGeminiResponseToAnthropic(response: GenerateContentResponse): Anthropic.Messages.Message {
+	const content: Anthropic.Messages.ContentBlock[] = []
+
+	const text = response.text
+	if (text) {
+		content.push({ type: "text", text, citations: null })
+	}
+
+	let stop_reason: Anthropic.Messages.Message["stop_reason"] = null
+	const finishReason = response.candidates?.[0]?.finishReason
+	if (finishReason) {
+		switch (finishReason) {
+			case "STOP":
+				stop_reason = "end_turn"
+				break
+			case "MAX_TOKENS":
+				stop_reason = "max_tokens"
+				break
+			case "SAFETY":
+			case "RECITATION":
+			case "OTHER":
+				stop_reason = "stop_sequence"
+				break
+		}
+	}
+
+	return {
+		id: `msg_${Date.now()}`,
+		type: "message",
+		role: "assistant",
+		content,
+		model: "",
+		stop_reason,
+		stop_sequence: null, // Gemini doesn't provide this information
+		usage: {
+			input_tokens: response.usageMetadata?.promptTokenCount ?? 0,
+			output_tokens: response.usageMetadata?.candidatesTokenCount ?? 0,
+			cache_creation_input_tokens: null,
+			cache_read_input_tokens: null,
+		},
+	}
+}

apps/vscode/src/core/api/transform/mistral-format.ts

@@ -0,0 +1,61 @@
+import { Anthropic } from "@anthropic-ai/sdk"
+import { AssistantMessage } from "@mistralai/mistralai/models/components/assistantmessage"
+import { SystemMessage } from "@mistralai/mistralai/models/components/systemmessage"
+import { ToolMessage } from "@mistralai/mistralai/models/components/toolmessage"
+import { UserMessage } from "@mistralai/mistralai/models/components/usermessage"
+
+export type MistralMessage =
+	| (SystemMessage & { role: "system" })
+	| (UserMessage & { role: "user" })
+	| (AssistantMessage & { role: "assistant" })
+	| (ToolMessage & { role: "tool" })
+
+export function convertToMistralMessages(anthropicMessages: Anthropic.Messages.MessageParam[]): MistralMessage[] {
+	const mistralMessages: MistralMessage[] = []
+	for (const anthropicMessage of anthropicMessages) {
+		if (typeof anthropicMessage.content === "string") {
+			mistralMessages.push({
+				role: anthropicMessage.role,
+				content: anthropicMessage.content,
+			})
+		} else {
+			if (anthropicMessage.role === "user") {
+				// Filter to only include text and image blocks
+				const textAndImageBlocks = anthropicMessage.content.filter(
+					(part) => part.type === "text" || part.type === "image",
+				)
+
+				if (textAndImageBlocks.length > 0) {
+					mistralMessages.push({
+						role: "user",
+						content: textAndImageBlocks.map((part) => {
+							if (part.type === "image") {
+								return {
+									type: "image_url",
+									imageUrl: {
+										url: `data:${part.source.media_type};base64,${part.source.data}`,
+									},
+								}
+							}
+							return { type: "text", text: part.text }
+						}),
+					})
+				}
+			} else if (anthropicMessage.role === "assistant") {
+				// Only process text blocks - assistant cannot send images or other content types in Mistral's API format
+				const textBlocks = anthropicMessage.content.filter((part) => part.type === "text")
+
+				if (textBlocks.length > 0) {
+					const content = textBlocks.map((part) => part.text).join("\n")
+
+					mistralMessages.push({
+						role: "assistant",
+						content,
+					})
+				}
+			}
+		}
+	}
+
+	return mistralMessages
+}

apps/vscode/src/core/api/transform/o1-format.ts

@@ -0,0 +1,436 @@
+import { Anthropic } from "@anthropic-ai/sdk"
+import OpenAI from "openai"
+import { Logger } from "@/shared/services/Logger"
+
+const o1SystemPrompt = (systemPrompt: string) => `
+# System Prompt
+
+${systemPrompt}
+
+# Instructions for Formulating Your Response
+
+You must respond to the user's request by using at least one tool call. When formulating your response, follow these guidelines:
+
+1. Begin your response with normal text, explaining your thoughts, analysis, or plan of action.
+2. If you need to use any tools, place ALL tool calls at the END of your message, after your normal text explanation.
+3. You can use multiple tool calls if needed, but they should all be grouped together at the end of your message.
+4. After placing the tool calls, do not add any additional normal text. The tool calls should be the final content in your message.
+
+Here's the general structure your responses should follow:
+
+\`\`\`
+[Your normal text response explaining your thoughts and actions]
+
+[Tool Call 1]
+[Tool Call 2 if needed]
+[Tool Call 3 if needed]
+...
+\`\`\`
+
+Remember:
+- Choose the most appropriate tool(s) based on the task and the tool descriptions provided.
+- Formulate your tool calls using the XML format specified for each tool.
+- Provide clear explanations in your normal text about what actions you're taking and why you're using particular tools.
+- Act as if the tool calls will be executed immediately after your message, and your next response will have access to their results.
+
+# Tool Descriptions and XML Formats
+
+1. execute_command:
+<execute_command>
+<command>Your command here</command>
+</execute_command>
+Description: Execute a CLI command on the system. Use this when you need to perform system operations or run specific commands to accomplish any step in the user's task. You must tailor your command to the user's system and provide a clear explanation of what the command does. Prefer to execute complex CLI commands over creating executable scripts, as they are more flexible and easier to run. Commands will be executed in the current working directory.
+
+2. list_files:
+<list_files>
+<path>Directory path here</path>
+<recursive>true or false (optional)</recursive>
+</list_files>
+Description: List files and directories within the specified directory. If recursive is true, it will list all files and directories recursively. If recursive is false or not provided, it will only list the top-level contents.
+
+3. list_code_definition_names:
+<list_code_definition_names>
+<path>Directory path here</path>
+</list_code_definition_names>
+Description: Lists definition names (classes, functions, methods, etc.) used in source code files at the top level of the specified directory. This tool provides insights into the codebase structure and important constructs, encapsulating high-level concepts and relationships that are crucial for understanding the overall architecture.
+
+4. search_files:
+<search_files>
+<path>Directory path here</path>
+<regex>Your regex pattern here</regex>
+<filePattern>Optional file pattern here</filePattern>
+</search_files>
+Description: Perform a regex search across files in a specified directory, providing context-rich results. This tool searches for patterns or specific content across multiple files, displaying each match with encapsulating context.
+
+5. read_file:
+<read_file>
+<path>File path here</path>
+</read_file>
+Description: Read the contents of a file at the specified path. Use this when you need to examine the contents of an existing file, for example to analyze code, review text files, or extract information from configuration files. Automatically extracts raw text from PDF and DOCX files. May not be suitable for other types of binary files, as it returns the raw content as a string.
+
+6. write_to_file:
+<write_to_file>
+<path>File path here</path>
+<content>
+Your file content here
+</content>
+</write_to_file>
+Description: Write content to a file at the specified path. If the file exists, it will be overwritten with the provided content. If the file doesn't exist, it will be created. Always provide the full intended content of the file, without any truncation. This tool will automatically create any directories needed to write the file.
+
+7. ask_followup_question:
+<ask_followup_question>
+<question>Your question here</question>
+</ask_followup_question>
+Description: Ask the user a question to gather additional information needed to complete the task. This tool should be used when you encounter ambiguities, need clarification, or require more details to proceed effectively. It allows for interactive problem-solving by enabling direct communication with the user. Use this tool judiciously to maintain a balance between gathering necessary information and avoiding excessive back-and-forth.
+
+8. attempt_completion:
+<attempt_completion>
+<command>Optional command to demonstrate result</command>
+<result>
+Your final result description here
+</result>
+</attempt_completion>
+Description: Once you've completed the task, use this tool to present the result to the user. They may respond with feedback if they are not satisfied with the result, which you can use to make improvements and try again.
+
+# Examples
+
+Here are some examples of how to structure your responses with tool calls:
+
+Example 1: Using a single tool
+
+Let's run the test suite for our project. This will help us ensure that all our components are functioning correctly.
+
+<execute_command>
+<command>npm test</command>
+</execute_command>
+
+Example 2: Using multiple tools
+
+Let's create two new configuration files for the web application: one for the frontend and one for the backend.
+
+<write_to_file>
+<path>./frontend-config.json</path>
+<content>
+{
+  "apiEndpoint": "https://api.example.com",
+  "theme": {
+    "primaryColor": "#007bff",
+    "secondaryColor": "#6c757d",
+    "fontFamily": "Arial, sans-serif"
+  },
+  "features": {
+    "darkMode": true,
+    "notifications": true,
+    "analytics": false
+  },
+  "version": "1.0.0"
+}
+</content>
+</write_to_file>
+
+<write_to_file>
+<path>./backend-config.yaml</path>
+<content>
+database:
+  host: localhost
+  port: 5432
+  name: myapp_db
+  user: admin
+
+server:
+  port: 3000
+  environment: development
+  logLevel: debug
+
+security:
+  jwtSecret: your-secret-key-here
+  passwordSaltRounds: 10
+
+caching:
+  enabled: true
+  provider: redis
+  ttl: 3600
+
+externalServices:
+  emailProvider: sendgrid
+  storageProvider: aws-s3
+</content>
+</write_to_file>
+
+Example 3: Asking a follow-up question
+
+I've analyzed the project structure, but I need more information to proceed. Let me ask the user for clarification.
+
+<ask_followup_question>
+<question>Which specific feature would you like me to implement in the example.py file?</question>
+</ask_followup_question>
+`
+
+export function convertToO1Messages(
+	openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[],
+	systemPrompt: string,
+): OpenAI.Chat.ChatCompletionMessageParam[] {
+	const toolsReplaced = openAiMessages.reduce((acc, message) => {
+		if (message.role === "tool") {
+			// Convert tool messages to user messages
+			acc.push({
+				role: "user",
+				content: message.content || "",
+			})
+		} else if (message.role === "assistant" && message.tool_calls) {
+			// Convert tool calls to content and remove tool_calls
+			let content = message.content || ""
+			message.tool_calls.forEach((toolCall) => {
+				if (toolCall.type === "function") {
+					content += `\nTool Call: ${toolCall.function.name}\nArguments: ${toolCall.function.arguments}`
+				}
+			})
+			acc.push({
+				role: "assistant",
+				content: content,
+				tool_calls: undefined,
+			})
+		} else {
+			// Keep other messages as they are
+			acc.push(message)
+		}
+		return acc
+	}, [] as OpenAI.Chat.ChatCompletionMessageParam[])
+
+	// Find the index of the last assistant message
+	// const lastAssistantIndex = findLastIndex(toolsReplaced, (message) => message.role === "assistant")
+
+	// Create a new array to hold the modified messages
+	const messagesWithSystemPrompt = [
+		{
+			role: "user",
+			content: o1SystemPrompt(systemPrompt),
+		} as OpenAI.Chat.ChatCompletionUserMessageParam,
+		...toolsReplaced,
+	]
+
+	// If there's an assistant message, insert the system prompt after it
+	// if (lastAssistantIndex !== -1) {
+	// 	const insertIndex = lastAssistantIndex + 1
+	// 	if (insertIndex < messagesWithSystemPrompt.length && messagesWithSystemPrompt[insertIndex].role === "user") {
+	// 		messagesWithSystemPrompt.splice(insertIndex, 0, {
+	// 			role: "user",
+	// 			content: o1SystemPrompt(systemPrompt),
+	// 		})
+	// 	}
+	// } else {
+	// 	// If there were no assistant messages, prepend the system prompt
+	// 	messagesWithSystemPrompt.unshift({
+	// 		role: "user",
+	// 		content: o1SystemPrompt(systemPrompt),
+	// 	})
+	// }
+
+	return messagesWithSystemPrompt
+}
+
+interface ToolCall {
+	tool: string
+	tool_input: Record<string, string>
+}
+
+const toolNames = [
+	"execute_command",
+	"list_files",
+	"list_code_definition_names",
+	"search_files",
+	"read_file",
+	"write_to_file",
+	"ask_followup_question",
+	"attempt_completion",
+]
+
+function parseAIResponse(response: string): {
+	normalText: string
+	toolCalls: ToolCall[]
+} {
+	// Create a regex pattern to match any tool call opening tag
+	const toolCallPattern = new RegExp(`<(${toolNames.join("|")})`, "i")
+	const match = response.match(toolCallPattern)
+
+	if (!match) {
+		// No tool calls found
+		return { normalText: response.trim(), toolCalls: [] }
+	}
+
+	const toolCallStart = match.index!
+	const normalText = response.slice(0, toolCallStart).trim()
+	const toolCallsText = response.slice(toolCallStart)
+
+	const toolCalls = parseToolCalls(toolCallsText)
+
+	return { normalText, toolCalls }
+}
+
+function parseToolCalls(toolCallsText: string): ToolCall[] {
+	const toolCalls: ToolCall[] = []
+
+	let remainingText = toolCallsText
+
+	while (remainingText.length > 0) {
+		const toolMatch = toolNames.find((tool) => new RegExp(`<${tool}`, "i").test(remainingText))
+
+		if (!toolMatch) {
+			break // No more tool calls found
+		}
+
+		const startTag = `<${toolMatch}`
+		const endTag = `</${toolMatch}>`
+		const startIndex = remainingText.indexOf(startTag)
+		const endIndex = remainingText.indexOf(endTag, startIndex)
+
+		if (endIndex === -1) {
+			break // Malformed XML, no closing tag found
+		}
+
+		const toolCallContent = remainingText.slice(startIndex, endIndex + endTag.length)
+		remainingText = remainingText.slice(endIndex + endTag.length).trim()
+
+		const toolCall = parseToolCall(toolMatch, toolCallContent)
+		if (toolCall) {
+			toolCalls.push(toolCall)
+		}
+	}
+
+	return toolCalls
+}
+
+function parseToolCall(toolName: string, content: string): ToolCall | null {
+	const tool_input: Record<string, string> = {}
+
+	// Remove the outer tool tags
+	const innerContent = content.replace(new RegExp(`^<${toolName}>|</${toolName}>$`, "g"), "").trim()
+
+	// Parse nested XML elements
+	const paramRegex = /<(\w+)>([\s\S]*?)<\/\1>/gs
+	let match: RegExpExecArray | null
+
+	while ((match = paramRegex.exec(innerContent)) !== null) {
+		const [, paramName, paramValue] = match
+		// Preserve newlines and trim only leading/trailing whitespace
+		tool_input[paramName] = paramValue.replace(/^\s+|\s+$/g, "")
+	}
+
+	// Validate required parameters
+	if (!validateToolInput(toolName, tool_input)) {
+		Logger.error(`Invalid tool call for ${toolName}:`, content)
+		return null
+	}
+
+	return { tool: toolName, tool_input }
+}
+
+function validateToolInput(toolName: string, tool_input: Record<string, string>): boolean {
+	switch (toolName) {
+		case "execute_command":
+			return "command" in tool_input
+		case "read_file":
+		case "list_code_definition_names":
+		case "list_files":
+			return "path" in tool_input
+		case "search_files":
+			return "path" in tool_input && "regex" in tool_input
+		case "write_to_file":
+			return "path" in tool_input && "content" in tool_input
+		case "ask_followup_question":
+			return "question" in tool_input
+		case "attempt_completion":
+			return "result" in tool_input
+		default:
+			return false
+	}
+}
+
+// Example usage:
+// const aiResponse = `Here's my analysis of the situation...
+
+// <execute_command>
+//   <command>ls -la</command>
+// </execute_command>
+
+// <write_to_file>
+//   <path>./example.txt</path>
+//   <content>Hello, World!</content>
+// </write_to_file>`;
+//
+// const { normalText, toolCalls } = parseAIResponse(aiResponse);
+// Logger.log(normalText);
+// Logger.log(toolCalls);
+
+// Convert OpenAI response to Anthropic format
+export function convertO1ResponseToAnthropicMessage(
+	completion: OpenAI.Chat.Completions.ChatCompletion,
+): Anthropic.Messages.Message {
+	const openAiMessage = completion.choices[0].message
+	const { normalText, toolCalls } = parseAIResponse(openAiMessage.content || "")
+
+	const anthropicMessage: Anthropic.Messages.Message = {
+		id: completion.id,
+		type: "message",
+		role: openAiMessage.role, // always "assistant"
+		content: [
+			{
+				type: "text",
+				text: normalText,
+				citations: null,
+			},
+		],
+		model: completion.model,
+		stop_reason: (() => {
+			switch (completion.choices[0].finish_reason) {
+				case "stop":
+					return "end_turn"
+				case "length":
+					return "max_tokens"
+				case "tool_calls":
+					return "tool_use"
+				case "content_filter": // Anthropic doesn't have an exact equivalent
+				default:
+					return null
+			}
+		})(),
+		stop_sequence: null, // which custom stop_sequence was generated, if any (not applicable if you don't use stop_sequence)
+		usage: {
+			input_tokens: completion.usage?.prompt_tokens || 0,
+			output_tokens: completion.usage?.completion_tokens || 0,
+			cache_creation_input_tokens: null,
+			cache_read_input_tokens: null,
+		},
+	}
+
+	if (toolCalls.length > 0) {
+		anthropicMessage.content.push(
+			...toolCalls.map((toolCall: ToolCall, index: number): Anthropic.ToolUseBlock => {
+				return {
+					type: "tool_use",
+					id: `call_${index}_${Date.now()}`, // Generate a unique ID for each tool call
+					name: toolCall.tool,
+					input: toolCall.tool_input,
+				}
+			}),
+		)
+	}
+
+	return anthropicMessage
+}
+
+// Example usage:
+// const openAICompletion = {
+//     id: "cmpl-123",
+//     choices: [{
+//         message: {
+//             role: "assistant",
+//             content: "Here's my analysis...\n\n<execute_command>\n  <command>ls -la</command>\n</execute_command>"
+//         },
+//         finish_reason: "stop"
+//     }],
+//     model: "gpt-3.5-turbo",
+//     usage: { prompt_tokens: 50, completion_tokens: 100 }
+// };
+// const anthropicMessage = convertO1ResponseToAnthropicMessage(openAICompletion);
+// Logger.log(anthropicMessage);

apps/vscode/src/core/api/transform/ollama-format.ts

@@ -0,0 +1,115 @@
+import { Message } from "ollama"
+import {
+	ClineAssistantToolUseBlock,
+	ClineImageContentBlock,
+	ClineStorageMessage,
+	ClineTextContentBlock,
+	ClineUserToolResultContentBlock,
+} from "@/shared/messages/content"
+
+export function convertToOllamaMessages(anthropicMessages: Omit<ClineStorageMessage, "modelInfo">[]): Message[] {
+	const ollamaMessages: Message[] = []
+
+	for (const anthropicMessage of anthropicMessages) {
+		if (typeof anthropicMessage.content === "string") {
+			ollamaMessages.push({
+				role: anthropicMessage.role,
+				content: anthropicMessage.content,
+			})
+		} else {
+			if (anthropicMessage.role === "user") {
+				const { nonToolMessages, toolMessages } = anthropicMessage.content.reduce<{
+					nonToolMessages: (ClineTextContentBlock | ClineImageContentBlock)[]
+					toolMessages: ClineUserToolResultContentBlock[]
+				}>(
+					(acc, part) => {
+						if (part.type === "tool_result") {
+							acc.toolMessages.push(part)
+						} else if (part.type === "text" || part.type === "image") {
+							acc.nonToolMessages.push(part)
+						}
+						return acc
+					},
+					{ nonToolMessages: [], toolMessages: [] },
+				)
+
+				// Process tool result messages FIRST since they must follow the tool use messages
+				const toolResultImages: string[] = []
+				toolMessages.forEach((toolMessage) => {
+					// The Anthropic SDK allows tool results to be a string or an array of text and image blocks, enabling rich and structured content. In contrast, the Ollama SDK only supports tool results as a single string, so we map the Anthropic tool result parts into one concatenated string to maintain compatibility.
+					let content: string
+
+					if (typeof toolMessage.content === "string") {
+						content = toolMessage.content
+					} else {
+						content =
+							toolMessage.content
+								?.map((part) => {
+									if (part.type === "image") {
+										toolResultImages.push(`data:${part.source.media_type};base64,${part.source.data}`)
+										return "(see following user message for image)"
+									}
+									return part.text
+								})
+								.join("\n") ?? ""
+					}
+					ollamaMessages.push({
+						role: "user",
+						images: toolResultImages.length > 0 ? toolResultImages : undefined,
+						content: content,
+					})
+				})
+
+				// Process non-tool messages
+				if (nonToolMessages.length > 0) {
+					ollamaMessages.push({
+						role: "user",
+						content: nonToolMessages
+							.map((part) => {
+								if (part.type === "image") {
+									return `data:${part.source.media_type};base64,${part.source.data}`
+								}
+								return part.text
+							})
+							.join("\n"),
+					})
+				}
+			} else if (anthropicMessage.role === "assistant") {
+				const { nonToolMessages, toolMessages } = anthropicMessage.content.reduce<{
+					nonToolMessages: (ClineTextContentBlock | ClineImageContentBlock)[]
+					toolMessages: ClineAssistantToolUseBlock[]
+				}>(
+					(acc, part) => {
+						if (part.type === "tool_use") {
+							acc.toolMessages.push(part)
+						} else if (part.type === "text" || part.type === "image") {
+							acc.nonToolMessages.push(part)
+						} // assistant cannot send tool_result messages
+						return acc
+					},
+					{ nonToolMessages: [], toolMessages: [] },
+				)
+
+				// Process non-tool messages
+				let content: string = ""
+				if (nonToolMessages.length > 0) {
+					content = nonToolMessages
+						.map((part) => {
+							if (part.type === "image") {
+								return "" // impossible as the assistant cannot send images
+							}
+							return part.text
+						})
+						.join("\n")
+				}
+
+				ollamaMessages.push({
+					role: "assistant",
+					content,
+				})
+			}
+		}
+	}
+
+	return ollamaMessages
+}

apps/vscode/src/core/api/transform/openai-format.ts

@@ -0,0 +1,512 @@
+import { Anthropic } from "@anthropic-ai/sdk"
+import OpenAI from "openai"
+import { ApiProvider } from "@/shared/api"
+import {
+	ClineAssistantRedactedThinkingBlock,
+	ClineAssistantThinkingBlock,
+	ClineAssistantToolUseBlock,
+	ClineImageContentBlock,
+	ClineStorageMessage,
+	ClineTextContentBlock,
+	ClineUserToolResultContentBlock,
+} from "@/shared/messages/content"
+import { Logger } from "@/shared/services/Logger"
+
+// OpenAI API has a maximum tool call ID length of 40 characters
+const MAX_TOOL_CALL_ID_LENGTH = 40
+
+/**
+ * Determines if a given tool ID follows the OpenAI Responses API format for tool calls.
+ * OpenAI tool call IDs start with "fc_" and are exactly 53 characters long.
+ *
+ * @param callId - The tool ID to check
+ * @returns True if the tool ID matches the OpenAI Responses API format, false otherwise
+ */
+function isOpenAIResponseToolId(callId: string): boolean {
+	return callId.startsWith("fc_") && callId.length === 53
+}
+
+/**
+ * Transforms a tool ID to a consistent format for OpenAI's Chat Completions API.
+ * NOTE: We do not want to transform tool IDs for non-OpenAI providers that may have different requirements.
+ * This function MUST be used for both tool_calls[].id (assistant) and tool_call_id (tool result)
+ * to ensure they match - otherwise OpenAI will reject the request with:
+ * "Invalid parameter: 'tool_call_id' of 'xxx' not found in 'tool_calls' of previous message."
+ *
+ * @param toolId - The original tool ID from Cline/Anthropic format
+ * @param provider - The API provider that the OpenAI formatted messages will be sent to
+ * @returns The transformed ID suitable for OpenAI API
+ */
+function transformToolCallIdForNativeApi(toolId: string, provider?: ApiProvider): string {
+	// OpenAI Responses API uses "fc_" prefix with 53 char length
+	// Convert these to "call_" prefix format for Chat Completions API
+	if (isOpenAIResponseToolId(toolId)) {
+		// Use the last 33 chars + "call_" (5 chars) to stay under the 40-char limit.
+		return `call_${toolId.slice(toolId.length - (MAX_TOOL_CALL_ID_LENGTH - 5))}`
+	}
+	if (provider !== "openai-native") {
+		return toolId
+	}
+	// Ensure ID doesn't exceed max length
+	if (toolId.length > MAX_TOOL_CALL_ID_LENGTH) {
+		return toolId.slice(0, MAX_TOOL_CALL_ID_LENGTH)
+	}
+	return toolId
+}
+
+/**
+ * Converts an array of ClineStorageMessage objects to OpenAI's Completions API format.
+ *
+ * Handles conversion of Cline-specific content types (tool uses, tool results, images, reasoning details)
+ * into OpenAI's expected message structure, including tool_calls and tool_call_id fields.
+ *
+ * @param anthropicMessages - Array of ClineStorageMessage objects to be converted
+ * @param provider - Optional parameter to indicate the API provider, which may affect ID transformation logic
+ * @returns Array of OpenAI.Chat.ChatCompletionMessageParam objects
+ */
+export function convertToOpenAiMessages(
+	anthropicMessages: Omit<ClineStorageMessage, "modelInfo">[],
+	provider?: ApiProvider,
+): OpenAI.Chat.ChatCompletionMessageParam[] {
+	const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = []
+
+	for (const anthropicMessage of anthropicMessages) {
+		if (typeof anthropicMessage.content === "string") {
+			openAiMessages.push({
+				role: anthropicMessage.role,
+				content: anthropicMessage.content,
+			})
+		} else {
+			// image_url.url is base64 encoded image data
+			// ensure it contains the content-type of the image: data:image/png;base64,
+			/*
+        { role: "user", content: "" | { type: "text", text: string } | { type: "image_url", image_url: { url: string } } },
+         // content required unless tool_calls is present
+        { role: "assistant", content?: "" | null, tool_calls?: [{ id: "", function: { name: "", arguments: "" }, type: "function" }] },
+        { role: "tool", tool_call_id: "", content: ""}
+         */
+			if (anthropicMessage.role === "user") {
+				const { nonToolMessages, toolMessages } = anthropicMessage.content.reduce<{
+					nonToolMessages: (ClineTextContentBlock | ClineImageContentBlock)[]
+					toolMessages: ClineUserToolResultContentBlock[]
+				}>(
+					(acc, part) => {
+						if (part.type === "tool_result") {
+							acc.toolMessages.push(part)
+						} else if (part.type === "text" || part.type === "image") {
+							acc.nonToolMessages.push(part)
+						} // user cannot send tool_use messages
+						return acc
+					},
+					{ nonToolMessages: [], toolMessages: [] },
+				)
+
+				// Process tool result messages FIRST since they must follow the tool use messages
+				const toolResultImages: ClineImageContentBlock[] = []
+				toolMessages.forEach((toolMessage) => {
+					// The Anthropic SDK allows tool results to be a string or an array of text and image blocks, enabling rich and structured content. In contrast, the OpenAI SDK only supports tool results as a single string, so we map the Anthropic tool result parts into one concatenated string to maintain compatibility.
+					let content: string
+
+					if (typeof toolMessage.content === "string") {
+						content = toolMessage.content
+					} else if (Array.isArray(toolMessage.content)) {
+						content =
+							toolMessage.content
+								?.map((part) => {
+									if (part.type === "image") {
+										toolResultImages.push(part)
+										return "(see following user message for image)"
+									}
+									return part.text
+								})
+								.join("\n") ?? ""
+					} else {
+						// Handle undefined content
+						content = ""
+					}
+					openAiMessages.push({
+						role: "tool",
+						// The tool_call_id must match the id used in the assistant's tool_calls array.
+						// Use the same transformation logic as tool_calls to ensure IDs match.
+						tool_call_id: transformToolCallIdForNativeApi(toolMessage.tool_use_id, provider),
+						content: content,
+					})
+				})
+
+				// If tool results contain images, send as a separate user message
+				// I ran into an issue where if I gave feedback for one of many tool uses, the request would fail.
+				// "Messages following `tool_use` blocks must begin with a matching number of `tool_result` blocks."
+				// Therefore we need to send these images after the tool result messages
+				// NOTE: it's actually okay to have multiple user messages in a row, the model will treat them as a continuation of the same input (this way works better than combining them into one message, since the tool result specifically mentions (see following user message for image)
+				// UPDATE v2.0: we don't use tools anymore, but if we did it's important to note that the openrouter prompt caching mechanism requires one user message at a time, so we would need to add these images to the user content array instead.
+				if (toolResultImages.length > 0) {
+					openAiMessages.push({
+						role: "user",
+						content: toolResultImages.map((part) => ({
+							type: "image_url",
+							image_url: { url: `data:${part.source.media_type};base64,${part.source.data}` },
+						})),
+					})
+				}
+
+				// Process non-tool messages
+				if (nonToolMessages.length > 0) {
+					openAiMessages.push({
+						role: "user",
+						content: nonToolMessages.map((part) => {
+							if (part.type === "image") {
+								return {
+									type: "image_url",
+									image_url: {
+										url: `data:${part.source.media_type};base64,${part.source.data}`,
+									},
+								}
+							}
+							return { type: "text", text: part.text }
+						}),
+					})
+				}
+			} else if (anthropicMessage.role === "assistant") {
+				const { nonToolMessages, toolMessages } = anthropicMessage.content.reduce<{
+					nonToolMessages: (
+						| ClineTextContentBlock
+						| ClineImageContentBlock
+						| ClineAssistantThinkingBlock
+						| ClineAssistantRedactedThinkingBlock
+					)[]
+					toolMessages: ClineAssistantToolUseBlock[]
+				}>(
+					(acc, part) => {
+						if (part.type === "tool_use") {
+							acc.toolMessages.push(part)
+						} else if (part.type === "text" || part.type === "image") {
+							acc.nonToolMessages.push(part)
+						} // assistant cannot send tool_result messages
+						return acc
+					},
+					{ nonToolMessages: [], toolMessages: [] },
+				)
+
+				// Process non-tool messages
+				let content: string | undefined
+				const reasoningDetails: any[] = []
+				const thinkingBlock = []
+				if (nonToolMessages.length > 0) {
+					nonToolMessages.forEach((part) => {
+						const anyPart = part as any
+						if (part.type === "text" && anyPart.reasoning_details) {
+							if (Array.isArray(anyPart.reasoning_details)) {
+								reasoningDetails.push(...anyPart.reasoning_details)
+							} else {
+								reasoningDetails.push(anyPart.reasoning_details)
+							}
+						}
+						if (part.type === "thinking" && part.thinking) {
+							// Reasoning details should have been moved to the text block
+							thinkingBlock.push(part)
+						}
+					})
+					content = nonToolMessages
+						.map((part) => {
+							if (part.type === "text" && part.text) {
+								return part.text
+							}
+							return ""
+						})
+						.join("\n")
+				}
+
+				// Process tool use messages
+				const tool_calls: OpenAI.Chat.ChatCompletionMessageToolCall[] = toolMessages.map((toolMessage) => {
+					const toolDetails = toolMessage.reasoning_details
+					const toolId = toolMessage.id
+					if (toolDetails) {
+						if (Array.isArray(toolDetails)) {
+							// For Gemini: reasoning details must be linkable back to the tool call.
+							// Sometimes OpenRouter/Gemini returns entries without `id`; those poison the next request.
+							// Keep only entries with an id matching the tool call id.
+							// See: https://github.com/cline/cline/issues/8214
+							const validDetails = toolDetails.filter((detail: any) => detail?.id === toolId)
+							if (validDetails.length > 0) {
+								reasoningDetails.push(...validDetails)
+							}
+						} else {
+							// Single reasoning detail - only include if it has matching id
+							const detail = toolDetails as any
+							if (detail?.id === toolId) {
+								reasoningDetails.push(toolDetails)
+							}
+						}
+					}
+
+					return {
+						// Use the same transformation as tool_call_id to ensure IDs match
+						id: transformToolCallIdForNativeApi(toolId, provider),
+						type: "function",
+						function: {
+							name: toolMessage.name,
+							// json string
+							arguments: JSON.stringify(toolMessage.input),
+						},
+					}
+				})
+
+				// Set content to blank when tool_calls are present but content has no text, per OpenAI API spec
+				const hasToolCalls = tool_calls.length > 0
+				const hasMeaningfulContent = content !== undefined && content.trim() !== ""
+				const finalContent = hasMeaningfulContent ? content : hasToolCalls ? null : undefined
+
+				const consolidatedReasoningDetails =
+					reasoningDetails.length > 0 ? consolidateReasoningDetails(reasoningDetails as any) : []
+
+				openAiMessages.push({
+					role: "assistant",
+					content: finalContent,
+					// Cannot be an empty array. API expects an array with minimum length 1, and will respond with an error if it's empty
+					tool_calls: tool_calls?.length > 0 ? tool_calls : undefined,
+					// Only include reasoning_details when non-empty; sending [] can trigger provider validation issues.
+					// @ts-expect-error
+					reasoning_details: consolidatedReasoningDetails.length > 0 ? consolidatedReasoningDetails : undefined,
+				})
+			}
+		}
+	}
+
+	return openAiMessages
+}
+
+// Type for OpenRouter's reasoning detail elements
+// https://openrouter.ai/docs/use-cases/reasoning-tokens#streaming-response
+type ReasoningDetail = {
+	// https://openrouter.ai/docs/use-cases/reasoning-tokens#reasoning-detail-types
+	type: string // "reasoning.summary" | "reasoning.encrypted" | "reasoning.text"
+	text?: string
+	data?: string // Encrypted reasoning data
+	signature?: string | null
+	id?: string | null // Unique identifier for the reasoning detail
+	/*
+	 The format of the reasoning detail, with possible values:
+	 	"unknown" - Format is not specified
+		"openai-responses-v1" - OpenAI responses format version 1
+		"anthropic-claude-v1" - Anthropic Claude format version 1 (default)
+	 */
+	format: string //"unknown" | "openai-responses-v1" | "anthropic-claude-v1" | "xai-responses-v1"
+	index?: number // Sequential index of the reasoning detail
+}
+
+// Helper function to convert reasoning_details array to the format OpenRouter API expects
+// Takes an array of reasoning detail objects and consolidates them by index
+function consolidateReasoningDetails(reasoningDetails: ReasoningDetail[]): ReasoningDetail[] {
+	if (!reasoningDetails || reasoningDetails.length === 0) {
+		return []
+	}
+
+	// Group by index
+	const groupedByIndex = new Map<number, ReasoningDetail[]>()
+
+	for (const detail of reasoningDetails) {
+		// Drop corrupted encrypted reasoning blocks that would otherwise trigger:
+		// "Invalid input: expected string, received undefined" for reasoning_details.*.data
+		// See: https://github.com/cline/cline/issues/8214
+		if (detail.type === "reasoning.encrypted" && !detail.data) continue
+
+		const index = detail.index ?? 0
+		if (!groupedByIndex.has(index)) {
+			groupedByIndex.set(index, [])
+		}
+		groupedByIndex.get(index)!.push(detail)
+	}
+
+	// Consolidate each group
+	const consolidated: ReasoningDetail[] = []
+
+	for (const [index, details] of groupedByIndex.entries()) {
+		// Concatenate all text parts
+		let concatenatedText = ""
+		let signature: string | undefined
+		let id: string | undefined
+		let format = "unknown"
+		let type = "reasoning.text"
+
+		for (const detail of details) {
+			if (detail.text) {
+				concatenatedText += detail.text
+			}
+			// Keep the signature from the last item that has one
+			if (detail.signature) {
+				signature = detail.signature
+			}
+			// Keep the id from the last item that has one
+			if (detail.id) {
+				id = detail.id
+			}
+			// Keep format and type from any item (they should all be the same)
+			if (detail.format) {
+				format = detail.format
+			}
+			if (detail.type) {
+				type = detail.type
+			}
+		}
+
+		// Create consolidated entry for text
+		if (concatenatedText) {
+			const consolidatedEntry: ReasoningDetail = {
+				type: type,
+				text: concatenatedText,
+				signature: signature,
+				id: id,
+				format: format,
+				index: index,
+			}
+			consolidated.push(consolidatedEntry)
+		}
+
+		// For encrypted chunks (data), only keep the last one
+		let lastDataEntry: ReasoningDetail | undefined
+		for (const detail of details) {
+			if (detail.data) {
+				lastDataEntry = {
+					type: detail.type,
+					data: detail.data,
+					signature: detail.signature,
+					id: detail.id,
+					format: detail.format,
+					index: index,
+				}
+			}
+		}
+		if (lastDataEntry) {
+			consolidated.push(lastDataEntry)
+		}
+	}
+
+	return consolidated
+}
+
+// Unique name to use to filter out tool call that cannot be parsed correctly
+const UNIQUE_ERROR_TOOL_NAME = "_cline_error_unknown_function_"
+
+// Convert OpenAI response to Anthropic format
+export function convertToAnthropicMessage(completion: OpenAI.Chat.Completions.ChatCompletion): Anthropic.Messages.Message {
+	const openAiMessage = completion.choices[0].message
+	const anthropicMessage: Anthropic.Messages.Message = {
+		id: completion.id,
+		type: "message",
+		role: openAiMessage.role, // always "assistant"
+		content: [
+			{
+				type: "text",
+				text: openAiMessage.content || "",
+				citations: null,
+			},
+		],
+		model: completion.model,
+		stop_reason: (() => {
+			switch (completion.choices[0].finish_reason) {
+				case "stop":
+					return "end_turn"
+				case "length":
+					return "max_tokens"
+				case "tool_calls":
+					return "tool_use"
+				case "content_filter": // Anthropic doesn't have an exact equivalent
+				default:
+					return null
+			}
+		})(),
+		stop_sequence: null, // which custom stop_sequence was generated, if any (not applicable if you don't use stop_sequence)
+		usage: {
+			input_tokens: completion.usage?.prompt_tokens || 0,
+			output_tokens: completion.usage?.completion_tokens || 0,
+			cache_creation_input_tokens: null,
+			cache_read_input_tokens: null,
+		},
+	}
+	try {
+		if (openAiMessage?.tool_calls?.length) {
+			const functionCalls = openAiMessage.tool_calls.filter((tc: any) => tc?.type === "function" && tc.function)
+			if (functionCalls.length > 0) {
+				anthropicMessage.content.push(
+					...functionCalls.map((toolCall: any): Anthropic.ToolUseBlock => {
+						let parsedInput = {}
+						try {
+							parsedInput = JSON.parse(toolCall.function?.arguments || "{}")
+						} catch (error) {
+							Logger.error("Failed to parse tool arguments:", error)
+						}
+						return {
+							type: "tool_use",
+							id: toolCall.id,
+							name: toolCall.function?.name || UNIQUE_ERROR_TOOL_NAME,
+							input: parsedInput,
+						}
+					}),
+				)
+			}
+
+			return anthropicMessage
+		}
+	} catch (error) {
+		Logger.error("Error converting OpenAI message to Anthropic format:", error)
+	}
+
+	return anthropicMessage
+}
+
+/**
+ * Sanitizes OpenAI messages for Gemini models by removing tool_calls that lack reasoning_details.
+ *
+ * Gemini models require thought signatures for tool calls. When switching providers mid-conversation,
+ * historical tool calls may not include Gemini reasoning details, which can poison the next request.
+ * This function drops tool_calls that lack reasoning_details and their paired tool messages.
+ *
+ * @param messages - Array of OpenAI chat completion messages
+ * @param modelId - The model ID to check if sanitization is needed
+ * @returns Sanitized array of messages (unchanged if not a Gemini model)
+ */
+export function sanitizeGeminiMessages(
+	messages: OpenAI.Chat.ChatCompletionMessageParam[],
+	modelId: string,
+): OpenAI.Chat.ChatCompletionMessageParam[] {
+	if (!modelId.includes("gemini")) {
+		return messages
+	}
+
+	const droppedToolCallIds = new Set<string>()
+	const sanitized: OpenAI.Chat.ChatCompletionMessageParam[] = []
+
+	for (const msg of messages) {
+		if (msg.role === "assistant") {
+			const anyMsg = msg as any
+			const toolCalls = anyMsg.tool_calls
+			if (Array.isArray(toolCalls) && toolCalls.length > 0) {
+				const reasoningDetails = anyMsg.reasoning_details
+				const hasReasoningDetails = Array.isArray(reasoningDetails) && reasoningDetails.length > 0
+				if (!hasReasoningDetails) {
+					for (const tc of toolCalls) {
+						if (tc?.id) {
+							droppedToolCallIds.add(tc.id)
+						}
+					}
+					// Keep any textual content, but drop the tool_calls themselves.
+					if (anyMsg.content) {
+						sanitized.push({ role: "assistant", content: anyMsg.content } as any)
+					}
+					continue
+				}
+			}
+		}
+
+		if (msg.role === "tool") {
+			const anyMsg = msg as any
+			if (anyMsg.tool_call_id && droppedToolCallIds.has(anyMsg.tool_call_id)) {
+				continue
+			}
+		}
+
+		sanitized.push(msg)
+	}
+
+	return sanitized
+}