temp - use the ubuntu latest runner and remove dependency on tests

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

.changeset/breezy-shrimps-begin.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+fix showing the ai core exisiting models when resource group field is empty (using the default resource group)

.changeset/cuddly-dingos-tap.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Fix issue on Account view where balance is fetched twice that cause janky UI

.changeset/cyan-bottles-battle.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Fixes an issue where thinking text from litellm was not being passed through to Cline thinking UI

.changeset/fluffy-toys-beam.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Fix Ollama connection issue to default endpoint at port 11434

.changeset/funny-ants-travel.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Optimized Cline for GPT-5 model family with an aligned system prompt

.changeset/healthy-waves-eat.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+REfactoring Tool Executor

.changeset/heavy-dolls-cry.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Add search functionality to API provider dropdown

.changeset/hip-chairs-wash.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Adding telemetry for terminal command issue

.changeset/hot-turtles-flash.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Remove disabled approve / reject buttons from UI.

.changeset/khaki-donuts-sneeze.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": minor
+---
+
+Sync taskHistory across Cline instances so that it will update for all Cline instances when one instance creates, adds to, or deletes a task.

.changeset/late-fans-grow.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Adding a grpc recording class with its builder

.changeset/late-wasps-agree.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Fixed minor bug where entering slash commands would result in the following word being removed from the chat text area

.changeset/lucky-keys-visit.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Fix issue where editor panel gets reset to chat view on each chat input focus event

.changeset/nasty-plums-run.md

@@ -1,5 +0,0 @@
----
-"claude-dev": patch
----
-
-Fix Global Rules directory documentation for Linux/WSL systems

.changeset/nine-adults-play.md

@@ -1,5 +0,0 @@
----
-"claude-dev": patch
----
-
-DeepSeek R1 0528 support under Hugging Face

.changeset/olive-frogs-applaud.md

@@ -1,5 +0,0 @@
----
-"claude-dev": patch
----
-
-remove duplicate tool registration for claude4-experimental

.changeset/orange-bees-itch.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Exclude dependency folders for deep-planning prompt

.changeset/pretty-apricots-help.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Sambanova provider: add DeepSeek-V3.1

.changeset/rare-fans-clean.md

@@ -1,5 +0,0 @@
----
-"claude-dev": minor
----
-
-Add Huawei Cloud MaaS Provider

.changeset/rare-oranges-flash.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Add "Use custom prompt" option to Ollama provider

.changeset/real-ads-glow.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+add free grok-coder-free model to cline provider

.changeset/rich-berries-applaud.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Adding a new standalone core api server script for local and integration testing purposes

.changeset/slimy-kids-sip.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Fix AutoApproveModal overflowing issue

.changeset/small-lions-train.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Setup Raw Structure for implementing multi-workspace support with WorkspaceRoot

.changeset/smart-flowers-deny.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Dify.ai api integration

.changeset/smart-jobs-impress.md

@@ -1,5 +0,0 @@
----
-"claude-dev": patch
----
-
-Fixed token counting when using VSCode LM API provider

.changeset/stale-balloons-scream.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": minor
+---
+
+support orchestration mode for sap provider

.changeset/sweet-days-whisper.md

@@ -1,5 +0,0 @@
----
-"claude-dev": patch
----
-
-fix: only focus on editor panel that is visible and active to stop input field stealing issue

.changeset/swift-ravens-lay.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Improve Gemini Rate Limit handling

.changeset/thirty-spies-deliver.md

@@ -1,5 +0,0 @@
----
-"claude-dev": minor
----
-
-Add Cerebras Qwen 3 235B instruct

.changeset/tiny-pants-train.md

@@ -1,5 +0,0 @@
----
-"claude-dev": patch
----
-
-trim input value for URL fields

.changeset/weak-years-sin.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+fix: Support Anthropic Caching when using LiteLLM

.changeset/yellow-cycles-join.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Prompt changes for deep-planning in windows/powershell

.clinerules/workflows/git-branch-analysis.md

@@ -6,28 +6,14 @@ Analyze the current branch's changes against main to provide informed insights a
 ## Step 1: Gather Git Information
 <important>Do not return any text or conversation other than what is necessary to run these commands</important>
 
-**First, check the expected output size:**
-   ```shell
-   (git branch --show-current && echo "=== STATUS ===" && git status --porcelain | cat && echo "=== COMMIT MESSAGES ===" && git log main..HEAD --oneline | cat && echo "=== CHANGED FILES ===" && git diff main --name-only | cat && echo "=== FULL DIFF ===" && git diff main | cat) | wc -l
-   ```
+**Run the following command to get the latest changes (bash):**
+```bash
+B=$(for c in main master origin/main origin/master; do git rev-parse --verify -q "$c" >/dev/null && echo "$c" && break; done); B=${B:-HEAD}; r(){ git branch --show-current; printf "=== STATUS ===\n"; git status --porcelain | cat; printf "=== COMMIT MESSAGES ===\n"; git log "$B"..HEAD --oneline | cat; printf "=== CHANGED FILES ===\n"; git diff "$B" --name-only | cat; printf "=== FULL DIFF ===\n"; git diff "$B" | cat; }; L=$(r | wc -l); if [ "$L" -gt 500 ]; then r > cline-git-analysis.temp && echo "::OUTPUT_FILE=cline-git-analysis.temp"; else r; fi
+```
 
-**If the expected line count is greater than 500 lines, use the file-based approach:**
-   ```shell
-   git branch --show-current > cline-git-analysis.temp && echo "=== STATUS ===" >> cline-git-analysis.temp && git status --porcelain >> cline-git-analysis.temp && echo "=== COMMIT MESSAGES ===" >> cline-git-analysis.temp && git log main..HEAD --oneline >> cline-git-analysis.temp && echo "=== CHANGED FILES ===" >> cline-git-analysis.temp && git diff main --name-only >> cline-git-analysis.temp && echo "=== FULL DIFF ===" >> cline-git-analysis.temp && git diff main >> cline-git-analysis.temp
-   ```
-
-   Then, read the file using the read_file tool. After you have read the file but before you proceed with subsequent steps, delete it:
-   ```shell
-   rm cline-git-analysis.temp
-   ```
-
-**If the expected line count is 500 lines or fewer, use the direct approach:**
-   ```shell
-   git branch --show-current && echo "=== STATUS ===" && git status --porcelain | cat && echo "=== COMMIT MESSAGES ===" && git log main..HEAD --oneline | cat && echo "=== CHANGED FILES ===" && git diff main --name-only | cat && echo "=== FULL DIFF ===" && git diff main | cat
-   ```
-
-<important>If using the direct approach, pipe outputs through `cat` to avoid interactive terminals. If the user's shell is not bash/zsh, adjust the command and chaining
- syntax accordingly.</important>
+```powershell
+$B=$null;foreach($c in 'main','master','origin/main','origin/master'){git rev-parse --verify -q $c *> $null;if($LASTEXITCODE -eq 0){$B=$c;break}};if(-not $B){$B='HEAD'};function r([string]$b){git rev-parse --abbrev-ref HEAD; '=== STATUS ==='; git status --porcelain | cat; '=== COMMIT MESSAGES ==='; git log "$b"..HEAD --oneline | cat; '=== CHANGED FILES ==='; git diff "$b" --name-only | cat; '=== FULL DIFF ==='; git diff "$b" | cat};$out=r $B|Out-String;$lines=($out -split "`r?`n").Count;if($lines -gt 500){$out|Set-Content -NoNewline cline-git-analysis.temp; '::OUTPUT_FILE=cline-git-analysis.temp'}else{$out}
+```
 
 ## Step 2: Silent, Structured Analysis Phase
 - Analyze all git output without providing commentary or narration

.clinerules/workflows/pr-review.md

@@ -219,6 +219,9 @@ EOF
 
 ## Basic PR Commands
 ```bash
+# Get current PR number
+gh pr view --json number -q .number
+
 # List open PRs
 gh pr list
 

.codespellrc

@@ -1,6 +0,0 @@
-[codespell]
-# Ref: https://github.com/codespell-project/codespell#using-a-config-file
-skip = .git*,*.svg,package-lock.json,*.css,.codespellrc,locales
-check-hidden = true
-ignore-regex = (\b(optIn|isTaller)\b|https://\S+)
-# ignore-words-list =

.eslintrc.json

@@ -1,33 +0,0 @@
-{
-	"root": true,
-	"parser": "@typescript-eslint/parser",
-	"parserOptions": {
-		"ecmaVersion": 6,
-		"sourceType": "module"
-	},
-	"plugins": ["@typescript-eslint", "eslint-rules"],
-	"rules": {
-		"@typescript-eslint/naming-convention": [
-			"warn",
-			{
-				"selector": "import",
-				"format": ["camelCase", "PascalCase"]
-			}
-		],
-		"@typescript-eslint/semi": "off",
-		"curly": "warn",
-		"eqeqeq": "warn",
-		"no-throw-literal": "warn",
-		"semi": "off",
-		"react-hooks/exhaustive-deps": "off",
-		"eslint-rules/no-direct-vscode-api": "warn",
-		"no-restricted-syntax": [
-			"error",
-			{
-				"selector": "VariableDeclarator[id.type=\"ObjectPattern\"][init.object.name=\"process\"][init.property.name=\"env\"]",
-				"message": "Use process.env.VARIABLE_NAME directly instead of destructuring"
-			}
-		]
-	},
-	"ignorePatterns": ["out", "dist", "**/*.d.ts"]
-}

.github/CODEOWNERS

@@ -1 +1,3 @@
-*       @saoudrizwan @ocasta181 @NightTrek @pashpashpash @dcbartlett @saito-sv @Garoth
+/docs/
+/.github/   @saoudrizwan @dcbartlett
+/README.md  @saoudrizwan @nickbaumann98

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -1,64 +1,69 @@
 name: 🐛 Bug Report
 description: File a bug report
-labels: ["bug"]
+labels: ['bug']
 body:
-    - type: markdown
-      attributes:
-          value: |
-              **Important:** All bug reports must be reproducible using Claude 4 Sonnet. Cline uses complex prompts so less capable models may not work as expected.
-    - type: textarea
-      id: what-happened
-      attributes:
-          label: What happened?
-          description: Also tell us, what did you expect to happen?
-          placeholder: Tell us what you see!
-      validations:
-          required: true
-    - type: textarea
-      id: steps
-      attributes:
-          label: Steps to reproduce
-          description: How do you trigger this bug? Please walk us through it step by step.
-          value: |
-              1.
-              2.
-              3.
-      validations:
-          required: false
-    - type: textarea
-      id: logs
-      attributes:
-          label: Relevant API REQUEST output
-          description: Please copy and paste any relevant output. This will be automatically formatted into code, so no need for backticks.
-          render: shell
-    - type: input
-      id: provider-model
-      attributes:
-          label: Provider/Model
-          description: What provider and model were you using when the issue occurred?
-          placeholder: "e.g., cline:anthropic/claude-3.7-sonnet, gemini:gemini-2.5-pro-exp-03-25"
-      validations:
-          required: true
-    - type: textarea
-      id: system-info
-      attributes:
-          label: System Information
-          description: What operating system and hardware are you using?
-          placeholder: |
-              Operating System: Windows 11, macOS Sonoma, Ubuntu 22.04, etc.
-              Hardware: CPU, GPU, RAM specifications if relevant
-              e.g.,
-              OS: Windows 11
-              CPU: Intel Core i7-11700K
-              GPU: NVIDIA GeForce RTX 3070
-              RAM: 32GB DDR4
-      validations:
-          required: true
-    - type: input
-      id: cline-version
-      attributes:
-          label: Cline Version
-          description: What version of Cline are you using? (You can find this at the bottom of the Settings view)
-          placeholder: "e.g., 1.2.3"
-      validations:
-          required: true
+  - type: markdown
+    attributes:
+      value: |
+        **Important:** All bug reports must be reproducible using Claude Sonnet 4. Cline uses complex prompts so less capable models may not work as expected.
+  - type: dropdown
+    id: plugin-type
+    attributes:
+      label: Plugin Type
+      description: Which plugin are you reporting a bug for?
+      options:
+        - VSCode Extension
+        - JetBrains Plugin
+      default: 0
+    validations:
+      required: true
+  - type: input
+    id: cline-version
+    attributes:
+      label: Cline Version
+      description: What version of Cline are you using? (You can find this at the bottom of the Settings view)
+      placeholder: 'e.g., 1.2.3'
+    validations:
+      required: true
+  - type: textarea
+    id: what-happened
+    attributes:
+      label: What happened?
+      description: Also tell us, what did you expect to happen?
+      placeholder: Tell us what you see!
+    validations:
+      required: true
+  - type: textarea
+    id: steps
+    attributes:
+      label: Steps to reproduce
+      description: How do you trigger this bug? Please walk us through it step by step.
+      value: |
+        1.
+        2.
+        3.
+    validations:
+      required: false
+  - type: input
+    id: provider-model
+    attributes:
+      label: Provider/Model
+      description: What provider and model were you using when the issue occurred?
+      placeholder: 'e.g., cline:anthropic/claude-sonnet-4, gemini:gemini-2.5-pro-exp-03-25'
+    validations:
+      required: false
+  - type: textarea
+    id: system-info
+    attributes:
+      label: System Information
+      description: What operating system and hardware are you using?
+      placeholder: |
+        Operating System: Windows 11, macOS Sonoma, Ubuntu 22.04, etc.
+        Hardware: CPU, GPU, RAM specifications if relevant
+        e.g.,
+        OS: Windows 11
+        CPU: Intel Core i7-11700K
+        GPU: NVIDIA GeForce RTX 3070
+        RAM: 32GB DDR4
+    validations:
+      required: false

.github/workflows/build.yml

@@ -0,0 +1,73 @@
+name: "Build Extension"
+
+on:
+    push:
+        branches:
+            - main
+            - build-package-per-commit
+
+permissions:
+    contents: write
+    packages: write
+    checks: write
+    pull-requests: write
+
+jobs:
+    build:
+        name: Build Extension
+        runs-on: 'ubuntu-latest'
+        # runs-on: ${{ github.repository == 'cline/cline' && 'high-capacity-runner' || 'ubuntu-latest' }}
+
+        steps:
+            - uses: actions/checkout@v4
+
+            - name: Setup Node.js
+              uses: actions/setup-node@v4
+              with:
+                  node-version: "lts/*"
+
+            - name: Cache root dependencies
+              uses: actions/cache@v4
+              id: root-cache
+              with:
+                  path: node_modules
+                  key: ${{ runner.os }}-npm-${{ hashFiles('package-lock.json') }}
+
+            # Cache webview-ui dependencies - only reuse if package-lock.json exactly matches
+            - name: Cache webview-ui dependencies
+              uses: actions/cache@v4
+              id: webview-cache
+              with:
+                  path: webview-ui/node_modules
+                  key: ${{ runner.os }}-npm-webview-${{ hashFiles('webview-ui/package-lock.json') }}
+
+            - name: Install root dependencies
+              if: steps.root-cache.outputs.cache-hit != 'true'
+              run: npm ci --include=optional
+
+            - name: Install webview-ui dependencies
+              if: steps.webview-cache.outputs.cache-hit != 'true'
+              run: cd webview-ui && npm ci --include=optional
+
+            - name: Install Publishing Tools
+              run: npm install -g @vscode/vsce ovsx
+
+            - name: Set Variables
+              id: set_variables
+              run: |
+                  VERSION=$(node -p "require('./package.json').version")
+                  echo "version=$VERSION" >> $GITHUB_OUTPUT
+                  SHORT_SHA=$(git rev-parse --short HEAD)
+                  echo "short_sha=$SHORT_SHA" >> $GITHUB_OUTPUT
+                  echo "file_path=cline-${VERSION}-${SHORT_SHA}.vsix" >> $GITHUB_OUTPUT
+
+            - name: Package Extension
+              run: |
+                  echo "Packaging extension with version ${{ steps.set_variables.outputs.version }}"
+                  vsce package  --allow-package-secrets sendgrid --out "${{ steps.set_variables.outputs.file_path }}"
+
+            - name: Upload Artifact
+              uses: actions/upload-artifact@v4
+              with:
+                  name: vsix-package
+                  path: ${{ steps.set_variables.outputs.file_path }}

.github/workflows/codespell.yml

@@ -1,28 +0,0 @@
-# Codespell configuration is within .codespellrc
----
-name: Codespell
-
-on:
-    push:
-        branches: [main]
-    pull_request:
-        branches: [main]
-
-permissions:
-    contents: read
-
-jobs:
-    codespell:
-        if: false
-        name: Check for spelling errors
-        runs-on: ubuntu-latest
-
-        steps:
-            - name: Checkout
-              uses: actions/checkout@v4
-            - name: Annotate locations with typos
-              uses: codespell-project/codespell-problem-matcher@v1
-            - name: Codespell
-              uses: codespell-project/actions-codespell@v2
-              with:
-                  only_warn: 1

.github/workflows/publish.yml

@@ -23,11 +23,7 @@ permissions:
     pull-requests: write
 
 jobs:
-    test:
-        uses: ./.github/workflows/test.yml
-
     publish:
-        needs: test
         name: Publish Extension
         runs-on: ubuntu-latest
         environment: publish
@@ -42,30 +38,6 @@ jobs:
               with:
                   node-version: "lts/*"
 
-            # Cache root dependencies - only reuse if package-lock.json exactly matches
-            - name: Cache root dependencies
-              uses: actions/cache@v4
-              id: root-cache
-              with:
-                  path: node_modules
-                  key: ${{ runner.os }}-npm-${{ hashFiles('package-lock.json') }}
-
-            # Cache webview-ui dependencies - only reuse if package-lock.json exactly matches
-            - name: Cache webview-ui dependencies
-              uses: actions/cache@v4
-              id: webview-cache
-              with:
-                  path: webview-ui/node_modules
-                  key: ${{ runner.os }}-npm-webview-${{ hashFiles('webview-ui/package-lock.json') }}
-
-            - name: Install root dependencies
-              if: steps.root-cache.outputs.cache-hit != 'true'
-              run: npm ci --include=optional
-
-            - name: Install webview-ui dependencies
-              if: steps.webview-cache.outputs.cache-hit != 'true'
-              run: cd webview-ui && npm ci --include=optional
-
             - name: Install Publishing Tools
               run: npm install -g @vscode/vsce ovsx
 
@@ -90,15 +62,48 @@ jobs:
 
                   echo "Tag '$TAG' validated successfully"
 
+            - name: Download Artifact
+              id: download_artifact
+              timeout-minutes: 30
+              env:
+                  GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+              run: |
+                  TAG="${{ github.event.inputs.tag }}"
+                  COMMIT_SHA=$(git rev-parse "$TAG")
+
+                  while true; do
+                    RUN_ID=$(gh run list \
+                    --workflow="Build Extension" \
+                    --commit="${COMMIT_SHA}" \
+                    --status=success \
+                    --limit=1 \
+                    --json databaseId \
+                    --jq '.[0].databaseId')
+                    
+                    if [ -n "$RUN_ID" ]; then
+                      echo "Build completed! Downloading artifact..."
+
+                      gh run download $RUN_ID --name vsix-package
+                      exit 0
+                    fi
+                    
+                    echo "Waiting for build..."
+                    sleep 10
+                  done
+
+            - name: Set VSIX Path
+              id: set_vsix_path
+              run: |
+                  FILE_PATH=$(ls *.vsix | head -n 1)
+                  echo "VSIX_PATH=$FILE_PATH" >> $GITHUB_OUTPUT
+
             - name: Package and Publish Extension
               env:
                   VSCE_PAT: ${{ secrets.VSCE_PAT }}
                   OVSX_PAT: ${{ secrets.OVSX_PAT }}
                   CLINE_ENVIRONMENT: production
+                  VSIX_PATH: ${{ steps.set_vsix_path.outputs.VSIX_PATH }}
               run: |
-                  # Required to generate the .vsix
-                  vsce package --out "cline-${{ steps.get_version.outputs.version }}.vsix"
-
                   if [ "${{ github.event.inputs.release-type }}" = "pre-release" ]; then
                     npm run publish:marketplace:prerelease
                     echo "Successfully published pre-release version ${{ steps.get_version.outputs.version }} to VS Code Marketplace and Open VSX Registry"

.github/workflows/test.yml

@@ -71,13 +71,7 @@ jobs:
             - name: Install xvfb on Linux
               if: runner.os == 'Linux'
               run: sudo apt-get update && sudo apt-get install -y xvfb
-
-            - name: Install local modules on windows
-              if: runner.os == 'Windows' && steps.root-cache.outputs.cache-hit == 'true'
-              run: |
-                  npm install eslint-plugin-eslint-rules 
-                  cd webview-ui/ && npm install eslint-plugin-eslint-rules
-
+              
             - name: Set up NPM on Windows
               if: runner.os == 'Windows'
               run: |
@@ -86,26 +80,25 @@ jobs:
             - name: Type Check
               run: npm run check-types
 
-            - name: ESLint Check
+            - name: Lint Check
               run: npm run lint
 
-            - name: Prettier / Format Check
+            - name: Format Check
               run: npm run format
 
             # Build the extension before running tests
             - name: Build Tests and Extension
               run: npm run pretest
 
-            # Unit Tests disabled due to module system conflicts between backend and webview-ui
-            # - name: Unit Tests
-            #   run: npm run test:unit
+            - name: Unit Tests
+              run: npm run test:unit
 
             # Run extension tests with coverage
-            - name: Extension Tests with Coverage
+            - name: Extension Integration Tests with Coverage
               id: extension_coverage
               continue-on-error: true
               run: |
-                  node ./scripts/test-ci.js > extension_coverage.txt 2>&1
+                  node ./scripts/test-ci.js 2>&1 | tee extension_coverage.txt
                   # Default the encoding to UTF-8 - It's not the default on Windows
                   PYTHONUTF8=1 PYTHONPATH=.github/scripts python -m coverage_check extract-coverage extension_coverage.txt --type=extension --github-output --verbose
 
@@ -117,7 +110,7 @@ jobs:
                   cd webview-ui
                   # Ensure coverage dependency is installed
                   npm install --no-save @vitest/coverage-v8
-                  npm run test:coverage > webview_coverage.txt 2>&1
+                  npm run test:coverage 2>&1 | tee webview_coverage.txt
                   cd ..
                   # Default the encoding to UTF-8 - It's not the default on Windows
                   PYTHONUTF8=1 PYTHONPATH=.github/scripts python -m coverage_check extract-coverage webview-ui/webview_coverage.txt --type=webview --github-output --verbose
@@ -132,21 +125,19 @@ jobs:
                   path: |
                       extension_coverage.txt
                       webview-ui/webview_coverage.txt
-                  retention-period: workflow # Artifacts are automatically deleted when the workflow completes
 
             # Set the check as failed if any of the tests failed
-            - name: Print test results and check for failures
+            - name: Check for test failures
               run: |
-                  echo "Extension Tests Result: ${{ steps.extension_coverage.outcome }}"
-                  cat extension_coverage.txt
-
-                  echo "Webview Tests Result: ${{ steps.webview_coverage.outcome }}"
-                  cat webview-ui/webview_coverage.txt
-
                   # Check if any of the test steps failed
                   # https://docs.github.com/en/actions/writing-workflows/choosing-what-your-workflow-does/accessing-contextual-information-about-workflow-runs#steps-context
+                  if [ "${{ steps.extension_coverage.outcome }}" != "success" ]; then
+                      echo "Extension Integration Tests failed, see previous step for test output."
+                  fi
+                  if [ "${{ steps.webview_coverage.outcome }}" != "success" ]; then
+                      echo "Webview Tests failed, see previous step for test output."
+                  fi
                   if [ "${{ steps.extension_coverage.outcome }}" != "success" ] || [ "${{ steps.webview_coverage.outcome }}" != "success" ]; then
-                      echo "Tests failed."
                       exit 1
                   fi
 

.gitignore

@@ -32,3 +32,6 @@ webview-ui/src/services/grpc-client.ts
 
 # E2E Tests
 test-results
+
+## CLI pre-release  ##
+/cli 

.husky/pre-commit

@@ -1,17 +1 @@
-echo "Running pre-commit checks..."
-
-# Run ESLint
-echo "Running ESLint..."
-npm run lint || {
-    echo "❌ ESLint check failed. Please fix the errors and try committing again."
-    exit 1
-}
-
-# Run Prettier
-echo "Running Prettier..."
-npx lint-staged --verbose || {
-     echo "❌ Prettier failed. Please fix the errors and try committing again."
-     exit 1
- }
-
-echo "✅ All checks passed!"
+lint-staged --no-stash

.mocharc.json

@@ -1,6 +1,15 @@
 {
-	"extension": ["ts"],
-	"spec": ["src/**/__tests__/*.ts", "eslint-rules/__tests__/**/*.test.ts"],
-	"require": ["ts-node/register", "source-map-support/register", "./src/test/requires.ts"],
-	"recursive": true
+	"extension": [
+		"ts"
+	],
+	"spec": [
+		"src/**/__tests__/*.ts"
+	],
+	"require": [
+		"ts-node/register",
+		"source-map-support/register",
+		"./src/test/requires.ts"
+	],
+	"recursive": true,
+	"exit": true
 }

.prettierignore

@@ -1,10 +0,0 @@
-dist/
-node_modules
-webview-ui/build/
-*.md
-package-lock.json
-src/core/prompts/system.ts
-src/core/prompts/model_prompts/claude4.ts
-evals/
-docs/
-out/

.prettierrc.json

@@ -1,8 +0,0 @@
-{
-	"tabWidth": 4,
-	"useTabs": true,
-	"printWidth": 130,
-	"semi": false,
-	"bracketSameLine": true,
-	"endOfLine": "lf"
-}

.vscode/extensions.json

@@ -2,9 +2,9 @@
 	// See http://go.microsoft.com/fwlink/?LinkId=827846
 	// for the documentation about the extensions.json format
 	"recommendations": [
-		"dbaeumer.vscode-eslint",
 		"connor4312.esbuild-problem-matchers",
 		"ms-vscode.extension-test-runner",
-		"bradlc.vscode-tailwindcss"
+		"bradlc.vscode-tailwindcss",
+		"biomejs.biome"
 	]
 }

.vscode/launch.json

@@ -9,8 +9,14 @@
 			"name": "Run Extension (production)",
 			"type": "extensionHost",
 			"request": "launch",
-			"args": ["--extensionDevelopmentPath=${workspaceFolder}", "--disable-workspace-trust", "${workspaceFolder}"],
-			"outFiles": ["${workspaceFolder}/dist/**/*.js"],
+			"args": [
+				"--extensionDevelopmentPath=${workspaceFolder}",
+				"--disable-workspace-trust",
+				"${workspaceFolder}"
+			],
+			"outFiles": [
+				"${workspaceFolder}/dist/**/*.js"
+			],
 			"preLaunchTask": "${defaultBuildTask}",
 			"env": {
 				"IS_DEV": "true",
@@ -22,8 +28,14 @@
 			"name": "Run Extension (staging)",
 			"type": "extensionHost",
 			"request": "launch",
-			"args": ["--extensionDevelopmentPath=${workspaceFolder}", "--disable-workspace-trust", "${workspaceFolder}"],
-			"outFiles": ["${workspaceFolder}/dist/**/*.js"],
+			"args": [
+				"--extensionDevelopmentPath=${workspaceFolder}",
+				"--disable-workspace-trust",
+				"${workspaceFolder}"
+			],
+			"outFiles": [
+				"${workspaceFolder}/dist/**/*.js"
+			],
 			"preLaunchTask": "${defaultBuildTask}",
 			"env": {
 				"IS_DEV": "true",
@@ -35,8 +47,14 @@
 			"name": "Run Extension (local)",
 			"type": "extensionHost",
 			"request": "launch",
-			"args": ["--extensionDevelopmentPath=${workspaceFolder}", "--disable-workspace-trust", "${workspaceFolder}"],
-			"outFiles": ["${workspaceFolder}/dist/**/*.js"],
+			"args": [
+				"--extensionDevelopmentPath=${workspaceFolder}",
+				"--disable-workspace-trust",
+				"${workspaceFolder}"
+			],
+			"outFiles": [
+				"${workspaceFolder}/dist/**/*.js"
+			],
 			"preLaunchTask": "${defaultBuildTask}",
 			"env": {
 				"IS_DEV": "true",
@@ -57,7 +75,9 @@
 				"--extensionDevelopmentPath=${workspaceFolder}",
 				"${workspaceFolder}"
 			],
-			"outFiles": ["${workspaceFolder}/dist/**/*.js"],
+			"outFiles": [
+				"${workspaceFolder}/dist/**/*.js"
+			],
 			"preLaunchTask": "clean-tmp-user",
 			"internalConsoleOptions": "openOnSessionStart",
 			"postDebugTask": "stop",
@@ -71,20 +91,71 @@
 		{
 			"type": "node",
 			"request": "launch",
-			"name": "Run cline-core service",
-			"skipFiles": ["<node_internals>/**"],
+			"name": "Debug Test Standalone Core Api Server (test:sca-server)",
+			"skipFiles": [
+				"<node_internals>/**"
+			],
 			"sourceMaps": true,
-			"resolveSourceMapLocations": ["${workspaceFolder}/**", "!**/node_modules/**"],
-			"cwd": "${workspaceFolder}/dist-standalone",
-			"outFiles": ["${workspaceFolder}/dist-standalone/**/*.js"],
+			"resolveSourceMapLocations": [
+				"${workspaceFolder}/**",
+				"!**/node_modules/**"
+			],
+			"cwd": "${workspaceFolder}",
+			"outFiles": [
+				"${workspaceFolder}/dist/**/*.js",
+				"${workspaceFolder}/dist-standalone/**/*.js"
+			],
 			"preLaunchTask": "compile-standalone",
+			"runtimeExecutable": "npx",
+			"runtimeArgs": [
+				"tsx"
+			],
+			"program": "scripts/test-standalone-core-api-server.ts",
 			"env": {
-				// Turns on grpc debug log.
-				//"GRPC_TRACE": "all",
-				//"GRPC_VERBOSITY": "DEBUG",
-				"NODE_PATH": "${workspaceFolder}/dist-standalone/node_modules"
+				"PROTOBUS_PORT": "26040",
+				"HOSTBRIDGE_PORT": "26041",
+				"WORKSPACE_DIR": "${workspaceFolder}",
+				"E2E_TEST": "true",
+				"CLINE_ENVIRONMENT": "local"
 			},
-			"program": "cline-core.js"
+			"console": "integratedTerminal",
+			"internalConsoleOptions": "neverOpen"
+		},
+		{
+			"type": "node",
+			"request": "launch",
+			"name": "Debug Current Test File",
+			"skipFiles": [
+				"<node_internals>/**"
+			],
+			"sourceMaps": true,
+			"resolveSourceMapLocations": [
+				"${workspaceFolder}/**",
+				"!**/node_modules/**"
+			],
+			"cwd": "${workspaceFolder}",
+			"runtimeExecutable": "npx",
+			"runtimeArgs": [
+				"mocha"
+			],
+			"args": [
+				"--require",
+				"ts-node/register",
+				"--require",
+				"source-map-support/register",
+				"--require",
+				"./src/test/requires.ts",
+				"--exit",
+				"${file}"
+			],
+			"env": {
+				"TS_NODE_PROJECT": "./tsconfig.unit-test.json",
+				"NODE_ENV": "test",
+				"IS_DEV": "true",
+				"CLINE_ENVIRONMENT": "local"
+			},
+			"console": "integratedTerminal",
+			"internalConsoleOptions": "openOnSessionStart"
 		}
 	]
 }

.vscode/settings.json

@@ -6,12 +6,31 @@
 	},
 	"search.exclude": {
 		"out": true, // set this to false to include "out" folder in search results
-		"dist": true // set this to false to include "dist" folder in search results
+		"dist": true, // set this to false to include "dist" folder in search results,
+		"node_modules": true,
+		"dist-standalone": true
 	},
 	// Turn off tsc task auto detection since we have the necessary tasks as npm scripts
 	"typescript.tsc.autoDetect": "off",
+	"typescript.preferences.quoteStyle": "double",
 	// Protobuf settings
 	"protoc": {
-		"options": ["--proto_path=proto"]
+		"options": [
+			"--proto_path=proto"
+		]
+	},
+	// Enable Lint and format using Biome
+	"biome.enabled": true,
+	"[typescript]": {
+		"editor.defaultFormatter": "biomejs.biome"
+	},
+	"[typescriptreact]": {
+		"editor.defaultFormatter": "biomejs.biome"
+	},
+	"[jsonc]": {
+		"editor.defaultFormatter": "biomejs.biome"
+	},
+	"[json]": {
+		"editor.defaultFormatter": "biomejs.biome"
 	}
 }

.vscode/tasks.json

@@ -30,7 +30,13 @@
 		},
 		{
 			"label": "watch",
-			"dependsOn": ["npm: protos", "npm: build:webview", "npm: dev:webview", "npm: watch:tsc", "npm: watch:esbuild"],
+			"dependsOn": [
+				"npm: protos",
+				"npm: build:webview",
+				"npm: dev:webview",
+				"npm: watch:tsc",
+				"npm: watch:esbuild"
+			],
 			"presentation": {
 				"reveal": "always"
 			},
@@ -60,7 +66,9 @@
 			"problemMatcher": [],
 			"isBackground": true,
 			"label": "npm: build:webview",
-			"dependsOn": ["npm: protos"],
+			"dependsOn": [
+				"npm: protos"
+			],
 			"presentation": {
 				"group": "watch",
 				"reveal": "always"
@@ -78,7 +86,9 @@
 			"problemMatcher": [],
 			"isBackground": true,
 			"label": "npm: build:webview:test",
-			"dependsOn": ["npm: protos"],
+			"dependsOn": [
+				"npm: protos"
+			],
 			"presentation": {
 				"group": "watch",
 				"reveal": "always"
@@ -113,7 +123,9 @@
 			],
 			"isBackground": true,
 			"label": "npm: dev:webview",
-			"dependsOn": ["npm: protos"],
+			"dependsOn": [
+				"npm: protos"
+			],
 			"presentation": {
 				"group": "watch",
 				"reveal": "always"
@@ -149,7 +161,9 @@
 			},
 			"isBackground": true,
 			"label": "npm: watch:esbuild",
-			"dependsOn": ["npm: protos"],
+			"dependsOn": [
+				"npm: protos"
+			],
 			"presentation": {
 				"group": "watch",
 				"reveal": "always"
@@ -185,7 +199,9 @@
 			},
 			"isBackground": true,
 			"label": "npm: watch:esbuild:test",
-			"dependsOn": ["npm: protos"],
+			"dependsOn": [
+				"npm: protos"
+			],
 			"presentation": {
 				"group": "watch",
 				"reveal": "always"
@@ -204,7 +220,9 @@
 			"problemMatcher": "$tsc-watch",
 			"isBackground": true,
 			"label": "npm: watch:tsc",
-			"dependsOn": ["npm: protos"],
+			"dependsOn": [
+				"npm: protos"
+			],
 			"presentation": {
 				"group": "watch",
 				"reveal": "always"
@@ -215,7 +233,9 @@
 			"script": "watch-tests",
 			"problemMatcher": "$tsc-watch",
 			"isBackground": true,
-			"dependsOn": ["npm: protos"],
+			"dependsOn": [
+				"npm: protos"
+			],
 			"presentation": {
 				"reveal": "always",
 				"group": "watchers"
@@ -224,7 +244,11 @@
 		},
 		{
 			"label": "tasks: watch-tests",
-			"dependsOn": ["npm: protos", "npm: watch", "npm: watch-tests"],
+			"dependsOn": [
+				"npm: protos",
+				"npm: watch",
+				"npm: watch-tests"
+			],
 			"problemMatcher": []
 		},
 		{
@@ -235,7 +259,9 @@
 		{
 			"label": "clean-tmp-user",
 			"type": "shell",
-			"dependsOn": ["watch"],
+			"dependsOn": [
+				"watch"
+			],
 			"command": "rm -rf ${workspaceFolder}/dist/tmp/user && mkdir -p ${workspaceFolder}/dist/tmp/user"
 		}
 	],

.vscodeignore

@@ -61,6 +61,6 @@ old_docs/**
 !assets/icons/**
 
 # Ignore E2E build files
-e2e-build.js
+e2e-build.mjs
 e2e.vsix
 test-results/

CHANGELOG.md

@@ -1,5 +1,181 @@
 # Changelog
 
+## [3.27.1]
+
+- Add new Kimi K2 model to groq and moonshot providers
+
+## [3.27.0]
+
+- Fix `grok-code-fast-1` model information
+- Add call to action for trying free `grok-code-fast-1` in Announcement banner
+
+## [3.26.7]
+
+- Add 200k context window variant for Claude Sonnet 4 to OpenRouter and Cline providers
+
+## [3.26.6]
+
+- Add free Grok Coder model to Cline provider for users looking for a fast, free coding model option
+- Fix GPT-5 models not respecting auto-compact setting when enabled, improving context window management
+- Fix provider retry attempts not showing proper user feedback during rate limiting scenarios
+- Improve markdown and code block styling to automatically adapt when switching VS Code themes
+
+## [3.26.5]
+
+- fix (provider/vercel-ai-gateway): reduce model list load frequency in settings view
+- Fix OVSX publish command to resolve deployment failure
+
+## [3.26.4]
+
+- Update nebius ai studio models
+- Update sap provider - support reasoning effort for open ai models
+- Fix Claude 4 image input in SAP AI Core Provider
+
+## [3.26.3]
+
+- Add compact system prompt option for LM Studio and Ollama models, optimized for smaller context windows (8k or less)
+- Add token usage tracking for LM Studio models to better monitor API consumption
+- Add "Use compact prompt" checkbox in LM Studio provider settings
+- Fix "Unexpected API Response" bug with gpt-5
+
+## [3.26.2]
+
+- Improve OpenRouter model parsing to show reasoning budget sliders for all models that support thinking, not just Claude models
+- Fix OpenRouter context window error handling to properly extract error codes from error messages, resolving "Unexpected API Response" errors with GPT-5 on Cline provider
+- Fix GPT-5 context window configuration for OpenAI/OpenRouter/Cline providers to use correct 272K limit
+- Remove max tokens configuration from Sonic Alpha model
+- Add Go language support to deep-planning feature (Thanks @yuvalman!)
+- Fix typo in Focus Chain settings page (Thanks @joyceerhl!)
+
+## [3.26.1]
+
+- Add Vercel AI Gateway as a new API provider option (Thanks @joshualipman123!)
+- Improve SAP AI Core provider to show deployed and undeployed models in the UI (Thanks @yuvalman!)
+- Fix Fireworks provider configuration and functionality (Thanks @ershang-fireworks!)
+- Add telemetry tracking for MCP tool usage to help improve the extension
+- Improve telemetry tracking for rules and workflow usage analytics
+- Set Plan mode to use strict mode by default for better planning results
+
+## [3.26.0]
+
+- Add Z AI as a new API provider with GLM-4.5 and GLM-4.5 Air models, offering competitive performance with cost-effective pricing especially for Chinese language tasks (Thanks @jues!)
+- Add Cline Sonic Alpha model - experimental advanced model with 262K context window for complex coding tasks
+- Add support for LM Studio local models from v0 API endpoint with configurable max tokens
+- Fix Ollama context window configuration not being used in requests
+
+## [3.25.3]
+
+- Fix bug where 'Enable checkpoints' and 'Disable MCP Marketplace' settings would be reset to default on reload
+- Move the position of the focus chain edit button when a scrollbar is present. Make the pencil icon bigger and better centered.
+
+## [3.25.2]
+
+- Fix attempt_completion showing twice in chat due to partial logic not being handled correctly
+- Fix OpenRouter showing cline credits error after 402 response
+
+## [3.25.1]
+
+- Fix attempt_completion command showing twice in chat view when updating progress checklist
+- Fix bug where announcement banner could not be dismissed
+- Add GPT-OSS models to AWS Bedrock
+
+## [3.25.0]
+
+- **Focus Chain:** Automatically creates and maintains todo lists as you work with Cline, breaking down complex tasks into manageable steps with real-time progress tracking
+- **Auto Compact:** Intelligently manages conversation context to prevent token limit errors by automatically compacting older messages while preserving important context
+- **Deep Planning:** New `/deep-planning` slash command for structured 4-step implementation planning that integrates with Focus Chain for automatic progress tracking
+- Add support for 200k context window for Claude Sonnet 4 in OpenRouter and Cline providers
+- Add option to configure custom base URL for Requesty provider
+
+## [3.24.0]
+
+- Add OpenAI GPT-5 Chat(gpt-5-chat-latest)
+- Add custom browser arguments setting to allow passing flags to the Chrome executable for better headless compatibility.
+- Add 1m context window model support for claude sonnet 4
+- Fis the API Keys URL for Requesty
+- Set gpt5 max tokens to 8_192 to fix 'context window exceeded' error
+- Fix issue where fallback request to retrieve cost was not using correct auth token
+- Add OpenAI context window exceeded error handling
+- Calibrate input token counts when using anthropic models of sap ai core provider
+
+## [3.23.0]
+
+- Add caching support for Bedrock inferences using SAP AI Core and minor refactor
+- Improve visibility for mode switch background color on different themes
+- Fix terminal commands putting webview in blocked state
+
+## [3.22.0]
+
+- Implemented a retry strategy for Cerebras to handle rate limit issues due to its generation speed
+- Add support for GPT-5 models to SAP AI Core Provider
+- Support sending context to active webview when editor panels are opened.
+- Fix bug where running out of credits on Cline accounts would show '402 empty body' response instead of 'buy credits' component
+- Fix LiteLLM Proxy Provider Cost Tracking
+
+## [3.21.0]
+
+- Add support for GPT-5 model family including GPT-5, GPT-5 Mini, and GPT-5 Nano with prompt caching support and set GPT-5 as the new default model
+- Add "Take a Tour" button for new users to easily access the VSCode walkthrough and improve onboarding experience
+- Enhance plan mode response handling with better exploration parameter support
+
+## [3.20.13]
+
+- Fix prompt caching support for Opus 4.1 on OpenRouter/Cline
+
+## [3.20.12]
+
+- Add Claude Opus 4.1 model support to AWS Bedrock provider (Thanks @omercelik!)
+- Fix prompt caching and extended thinking support for Claude Opus 4.1 in Anthropic provider
+
+## [3.20.11]
+
+Add gpt-oss-120b as a Cerebras model
+Add Opus 4.1 through Claude Code
+
+## [3.20.10]
+
+- Add OpenAI's new open-source models (GPT-OSS-120B and GPT-OSS-20B) to Hugging Face and Groq providers
+
+## [3.20.9]
+
+- Add support for Claude Opus 4.1 model in Anthropic provider
+- Add Baseten as a new API provider with support for DeepSeek, Llama, and Kimi K2 models (Thanks @AlexKer!)
+- Fix error messages not clearing from UI when retrying failed tasks
+- Fix chat input box positioning issues
+
+## [3.20.8]
+
+- Add navbar tooltips on hover
+
+## [3.20.7]
+
+- Fix circular dependency that affect the github workflow Tests / test (pull_request)
+
+## [3.20.6]
+
+- Fix login check on extension restart
+
+## [3.20.5]
+
+- Fix authentication persistence issues that could cause users to be logged out unexpectedly
+
+## [3.20.4]
+
+- Add new Cerebras models
+- Update rate limits for existing Cerebras models
+- Fix for delete task dialog
+
+## [3.20.3]
+
+- Add Huawei Cloud MaaS Provider (Thanks @ddling!)
+- Add Cerebras Qwen 3 235B instruct model (Thanks @kevint-cerebras!)
+- Add DeepSeek R1 0528 support under Hugging Face (Thanks @0ne0rZer0!)
+- Fix Global Rules directory documentation for Linux/WSL systems
+- Fix token counting when using VSCode LM API provider
+- Fix input field stealing focus issue by only focusing on visible and active editor panels
+- Fix duplicate tool registration for claude4-experimental
+- Trim input value for URL fields
+
 ## [3.20.2]
 
 - Fixed issue with sap ai core client credentials storage

CONTRIBUTING.md

@@ -147,7 +147,7 @@ Anyone can contribute code to Cline, but we ask that you follow these guidelines
     - Run `npm run lint` to check code style
     - Run `npm run format` to automatically format code
     - All PRs must pass CI checks which include both linting and formatting
-    - Address any ESLint warnings or errors before submitting
+    - Address any warnings or errors from linter before submitting
     - Follow TypeScript best practices and maintain type safety
 
 3. **Testing**

README.md

@@ -30,9 +30,9 @@ English | <a href="https://github.com/cline/cline/blob/main/locales/es/README.md
 </table>
 </div>
 
-Meet Cline (pronounced /klaɪn/, like "Klein"), an AI assistant that can use your **CLI** a**N**d **E**ditor.
+Meet Cline, an AI assistant that can use your **CLI** a**N**d **E**ditor.
 
-Thanks to [Claude 3.7 Sonnet's agentic coding capabilities](https://www.anthropic.com/claude/sonnet), Cline can handle complex software development tasks step-by-step. With tools that let him create & edit files, explore large projects, use the browser, and execute terminal commands (after you grant permission), he can assist you in ways that go beyond code completion or tech support. Cline can even use the Model Context Protocol (MCP) to create new tools and extend his own capabilities. While autonomous AI scripts traditionally run in sandboxed environments, this extension provides a human-in-the-loop GUI to approve every file change and terminal command, providing a safe and accessible way to explore the potential of agentic AI.
+Thanks to [Claude Sonnet's agentic coding capabilities](https://www.anthropic.com/claude/sonnet), Cline can handle complex software development tasks step-by-step. With tools that let him create & edit files, explore large projects, use the browser, and execute terminal commands (after you grant permission), he can assist you in ways that go beyond code completion or tech support. Cline can even use the Model Context Protocol (MCP) to create new tools and extend his own capabilities. While autonomous AI scripts traditionally run in sandboxed environments, this extension provides a human-in-the-loop GUI to approve every file change and terminal command, providing a safe and accessible way to explore the potential of agentic AI.
 
 1. Enter your task and add images to convert mockups into functional apps or fix bugs with screenshots.
 2. Cline starts by analyzing your file structure & source code ASTs, running regex searches, and reading relevant files to get up to speed in existing projects. By carefully managing what information is added to context, Cline can provide valuable assistance even for large, complex projects without overwhelming the context window.
@@ -87,7 +87,7 @@ All changes made by Cline are recorded in your file's Timeline, providing an eas
 
 ### Use the Browser
 
-With Claude 3.5 Sonnet's new [Computer Use](https://www.anthropic.com/news/3-5-models-and-computer-use) capability, Cline can launch a browser, click elements, type text, and scroll, capturing screenshots and console logs at each step. This allows for interactive debugging, end-to-end testing, and even general web use! This gives him autonomy to fixing visual bugs and runtime issues without you needing to handhold and copy-pasting error logs yourself.
+With Claude Sonnet's new [Computer Use](https://www.anthropic.com/news/3-5-models-and-computer-use) capability, Cline can launch a browser, click elements, type text, and scroll, capturing screenshots and console logs at each step. This allows for interactive debugging, end-to-end testing, and even general web use! This gives him autonomy to fixing visual bugs and runtime issues without you needing to handhold and copy-pasting error logs yourself.
 
 Try asking Cline to "test the app", and watch as he runs a command like `npm run dev`, launches your locally running dev server in a browser, and performs a series of tests to confirm that everything works. [See a demo here.](https://x.com/sdrzn/status/1850880547825823989)
 

biome.jsonc

@@ -0,0 +1,160 @@
+{
+	"$schema": "https://biomejs.dev/schemas/2.1.4/schema.json",
+	"vcs": {
+		"enabled": true,
+		"clientKind": "git",
+		"useIgnoreFile": true,
+		"defaultBranch": "main"
+	},
+	"assist": {
+		"enabled": true,
+		"actions": {
+			"source": {
+				"organizeImports": "on",
+				"useSortedAttributes": "on"
+			}
+		}
+	},
+	"linter": {
+		"enabled": true,
+		"domains": {
+			"react": "recommended"
+		},
+		// Ideally we would want to turn on all the rules that are currently off,
+		// keeping them off currently to make sure only changes on the migrations
+		// are included in the initial PR before we apply the format and lint changes.
+		// TODO: turn on all rules that are currently off if applicable.
+		// TODO: Remove --diagnostic-level=error from CI commands.
+		"rules": {
+			"recommended": true,
+			"correctness": {
+				"useExhaustiveDependencies": "off",
+				"noUndeclaredVariables": "off",
+				"noEmptyPattern": "off",
+				"useJsxKeyInIterable": "off",
+				"noInnerDeclarations": "off",
+				"useHookAtTopLevel": "off",
+				"useYield": "off",
+				"noConstructorReturn": "off",
+				"noInvalidPositionAtImportRule": "off",
+				"noSwitchDeclarations": "off",
+				"noUnusedImports": "error"
+			},
+			"a11y": "off",
+			"style": {
+				"useNodejsImportProtocol": "off",
+				"useImportType": "off",
+				"useBlockStatements": "warn",
+				"useNamingConvention": "off",
+				"useThrowOnlyError": "info",
+				"useConsistentArrayType": "off",
+				"noParameterAssign": "off",
+				"useAsConstAssertion": "off",
+				"useDefaultParameterLast": "off",
+				"noNonNullAssertion": "off",
+				"useEnumInitializers": "off",
+				"useSelfClosingElements": "off",
+				"useSingleVarDeclarator": "off",
+				"useNumberNamespace": "off",
+				"noInferrableTypes": "off",
+				"useTemplate": "off",
+				"noUselessElse": "off"
+			},
+			"suspicious": {
+				"noDoubleEquals": "warn",
+				"noImplicitAnyLet": "info",
+				"noThenProperty": "off",
+				"noAsyncPromiseExecutor": "off",
+				"noImportAssign": "off",
+				"noExplicitAny": "off",
+				"noControlCharactersInRegex": "off",
+				"noShadowRestrictedNames": "off",
+				"noArrayIndexKey": "info",
+				"noAssignInExpressions": "warn"
+			},
+			"complexity": {
+				"noUselessConstructor": "off",
+				"useOptionalChain": "off",
+				"noBannedTypes": "off",
+				"useLiteralKeys": "off",
+				"noUselessCatch": "off",
+				"noUselessSwitchCase": "off",
+				"noStaticOnlyClass": "off"
+			},
+			"security": {
+				"noDangerouslySetInnerHtml": "warn"
+			}
+		}
+	},
+	"formatter": {
+		"enabled": true,
+		"indentStyle": "tab",
+		"indentWidth": 4,
+		"lineWidth": 130,
+		"lineEnding": "lf",
+		"formatWithErrors": true
+	},
+	"javascript": {
+		"formatter": {
+			"semicolons": "asNeeded",
+			"arrowParentheses": "always",
+			"bracketSameLine": true,
+			"bracketSpacing": true,
+			"jsxQuoteStyle": "double",
+			"quoteProperties": "asNeeded",
+			"trailingCommas": "all"
+		}
+	},
+	"json": {
+		"formatter": {
+			"trailingCommas": "none",
+			"expand": "always"
+		}
+	},
+	"files": {
+		"includes": [
+			"**",
+			"!**/dist/**",
+			"!**/dist-*/**",
+			"!**/out/**",
+			"!**/evals/**",
+			"!**/playwright/**",
+			"!**/test-results/**",
+			"!**/node_modules/**",
+			"!**/webview-ui/build/**",
+			"!**/generated/**",
+			"!**/proto/**"
+		]
+	},
+	"plugins": [
+		"src/dev/grit/process-env.grit"
+	],
+	"overrides": [
+		{
+			"includes": [
+				"**",
+				"!**/hosts/vscode/**",
+				"!**/test/**",
+				"!src/extension.ts"
+			],
+			"plugins": [
+				"src/dev/grit/vscode-api.grit"
+			]
+		},
+		{
+			"includes": [
+				"**",
+				"!src/core/storage/state-migrations.ts",
+				"!src/core/storage/FileContextTracker.ts",
+				"!src/core/context/context-tracking/FileContextTracker.ts",
+				"!src/common.ts",
+				"!src/services/logging/distinctId.ts",
+				"!src/core/storage/utils/state-helpers.ts",
+				"!src/extension.ts"
+			],
+			"plugins": [
+				"src/dev/grit/use-cache-service.grit"
+			]
+		}
+	]
+}

docs/docs.json

@@ -57,18 +57,27 @@
 			{
 				"group": "Getting Started",
 				"pages": [
-					"getting-started/for-new-coders",
-					"getting-started/installing-cline",
-					"getting-started/installing-dev-essentials",
+					"getting-started/what-is-cline",
 					"getting-started/model-selection-guide",
+					"getting-started/installing-cline",
+					"getting-started/installing-cline-jetbrains",
 					"getting-started/task-management",
 					"getting-started/understanding-context-management",
-					"getting-started/what-is-cline"
+					{
+						"group": "For New Coders",
+						"pages": [
+							"getting-started/for-new-coders",
+							"getting-started/installing-dev-essentials"
+						]
+					}
 				]
 			},
 			{
 				"group": "Improving Your Prompting Skills",
-				"pages": ["prompting/prompt-engineering-guide", "prompting/cline-memory-bank"]
+				"pages": [
+					"prompting/prompt-engineering-guide",
+					"prompting/cline-memory-bank"
+				]
 			},
 			{
 				"group": "Features",
@@ -79,6 +88,8 @@
 					"features/drag-and-drop",
 					"features/plan-and-act",
 					"features/slash-commands/workflows",
+					"features/focus-chain",
+					"features/auto-compact",
 					"features/editing-messages",
 					{
 						"group": "@ Mentions",
@@ -97,7 +108,8 @@
 							"features/slash-commands/new-task",
 							"features/slash-commands/new-rule",
 							"features/slash-commands/smol",
-							"features/slash-commands/report-bug"
+							"features/slash-commands/report-bug",
+							"features/slash-commands/deep-planning"
 						]
 					},
 					{
@@ -146,21 +158,32 @@
 				"pages": [
 					"provider-config/anthropic",
 					"provider-config/claude-code",
-					"provider-config/aws-bedrock-with-apikey-authentication",
-					"provider-config/aws-bedrock-with-credentials-authentication",
-					"provider-config/aws-bedrock-with-profile-authentication",
+					{
+						"group": "AWS Bedrock",
+						"pages": [
+							"provider-config/aws-bedrock/api-key",
+							"provider-config/aws-bedrock/iam-credentials",
+							"provider-config/aws-bedrock/cli-profile"
+						]
+					},
 					"provider-config/gcp-vertex-ai",
 					"provider-config/litellm-and-cline-using-codestral",
 					"provider-config/vscode-language-model-api",
 					"provider-config/xai-grok",
 					"provider-config/mistral-ai",
 					"provider-config/deepseek",
+					"provider-config/groq",
+					"provider-config/cerebras",
+					"provider-config/doubao",
+					"provider-config/fireworks",
+					"provider-config/zai",
 					"provider-config/ollama",
 					"provider-config/openai",
 					"provider-config/openai-compatible",
 					"provider-config/openrouter",
-					"provider-config/requesty",
-					"provider-config/sap-aicore"
+					"provider-config/sap-aicore",
+					"provider-config/vercel-ai-gateway",
+					"provider-config/requesty"
 				]
 			},
 			{
@@ -173,11 +196,16 @@
 			},
 			{
 				"group": "Troubleshooting",
-				"pages": ["troubleshooting/terminal-quick-fixes", "troubleshooting/terminal-integration-guide"]
+				"pages": [
+					"troubleshooting/terminal-quick-fixes",
+					"troubleshooting/terminal-integration-guide"
+				]
 			},
 			{
 				"group": "More Info",
-				"pages": ["more-info/telemetry"]
+				"pages": [
+					"more-info/telemetry"
+				]
 			}
 		]
 	},
@@ -188,10 +216,19 @@
 			"discord": "https://discord.gg/cline"
 		}
 	},
+	"anchors": [
+		{
+			"name": "What is Cline",
+			"icon": "house",
+			"url": "getting-started/what-is-cline"
+		}
+	],
 	"search": {
 		"prompt": "Search Cline documentation..."
 	},
 	"contextual": {
-		"options": ["copy"]
+		"options": [
+			"copy"
+		]
 	}
 }

docs/features/auto-compact.mdx

@@ -0,0 +1,75 @@
+---
+title: "Automatic Context Summarization"
+sidebarTitle: "Auto Compact"
+---
+
+When your conversation approaches the model's context window limit, Cline automatically summarizes it to free up space and keep working. 
+
+<Frame>
+	<img
+		src="https://storage.googleapis.com/cline_public_images/docs/assets/condensing.png"
+		alt="Auto-compact feature condensing conversation context"
+	/>
+</Frame>
+
+## How It Works
+
+Cline monitors token usage during your conversation. When you're getting close to the limit, he:
+
+1. Creates a comprehensive summary of everything that's happened
+2. Preserves all the technical details, code changes, and decisions
+3. Replaces the conversation history with the summary
+4. Continues exactly where he left off
+
+You'll see a summarization tool call when this happens, showing the total cost like any other api call in the chat view.
+
+## Why This Matters
+
+Previously, Cline would truncate older messages when hitting context limits. This meant losing important context from earlier in the conversation.
+
+Now with summarization:
+- All technical decisions and code patterns are preserved
+- File changes and project context remain intact
+- Cline remembers everything he's done
+- You can work on much larger projects without interruption
+
+<Tip>
+Context Summarization synergizes beautifully with [Focus Chain](/features/focus-chain). When Focus Chain is enabled, todo lists persist across summarizations. This means Cline can work on long-horizon tasks that span multiple context windows while staying on track with the todo list guiding him through each reset.
+</Tip>
+
+## Technical Details
+
+The summarization happens through your configured API provider using the same model you're already using. It leverages prompt caching to minimize costs.
+
+1. Cline uses a [summarization prompt](https://github.com/cline/cline/blob/main/src/core/prompts/contextManagement.ts) to request a summary of the conversation.
+
+2. Once the summary is generated, Cline replaces the conversation history with a [continuation prompt](https://github.com/cline/cline/blob/main/src/core/prompts/contextManagement.ts#L69) that asks Cline to keep working and provides the summary as context.
+
+Different models have different context window thresholds for when auto-summarization kicks in. You can see how thresholds are determined in [context-window-utils.ts](https://github.com/cline/cline/blob/main/src/core/context/context-management/context-window-utils.ts). 
+
+## Cost Considerations
+
+Summarization leverages your existing prompt cache from the conversation, so it costs about the same as any other tool call.
+
+Since most input tokens are already cached, you're primarily paying for the summary generation (output tokens), making it very cost-effective.
+
+## Restoring Context with Checkpoints
+
+You can use [checkpoints](/features/checkpoints) to restore your task state from before a summarization occurred. This means you never truly lose context - you can always roll back to previous versions of your conversation. 
+
+<Note>
+  Editing a message before a summarization tool call will work similarly to a checkpoint, allowing you to restore the conversation to that point.
+</Note>
+
+## Next Generation Model Support
+
+Auto Compact uses advanced LLM-based summarization which we've found works significantly better for next-generation models. We currently support this feature for the following models:
+
+- **Claude 4 series**
+- **Gemini 2.5 series**
+- **GPT-5**
+- **Grok 4**
+
+<Note>
+When using other models, Cline automatically falls back to the standard rule-based context truncation method, even if Auto Compact is enabled in settings.
+</Note>

docs/features/drag-and-drop.mdx

@@ -5,10 +5,28 @@ sidebarTitle: "Drag & Drop"
 
 Dragging and dropping files into Cline is a quick way to add images, code, and other files to your conversations.
 
+<Frame>
+	<img
+		src="https://storage.googleapis.com/cline_public_images/docs/assets/dran-n-drop.gif"
+		alt="Dragging and dropping files into Cline chat"
+	/>
+</Frame>
+
 <Note>Due to VS Code quirks, to drag and drop files into the Cline chat input, you need to hold `Shift` while dragging.</Note>
 
 Dragging and dropping workspace files into Cline will automatically create a [file mention](/features/at-mentions/file-mentions). This allows you to reference the file in your conversation without needing to type out the path.
 
+### Dragging from Finder/File Explorer
+
+You can drag files directly from your system's file manager into Cline:
+
+<Frame>
+	<img
+		src="https://storage.googleapis.com/cline_public_images/docs/assets/drag-n-drop-finder.gif"
+		alt="Dragging files from Finder into Cline"
+	/>
+</Frame>
+
 ### Supported File Types
 
-Cline supports dragging external images from your file system, as well as files from your workspace.
+Cline supports dragging external images, pdfs, csv, excel, and other text files from your file system, as well as files from your workspace.

docs/features/focus-chain.mdx

@@ -0,0 +1,303 @@
+---
+title: "Focus Chain"
+sidebarTitle: "Focus Chain"
+---
+
+Focus Chain is a task management enhancement feature in Cline that provides automatic todo list management with real-time progress tracking throughout your tasks. 
+
+<Frame>
+	<img
+		src="https://storage.googleapis.com/cline_public_images/docs/assets/2dos.gif"
+		alt="Focus Chain todo list management with real-time progress tracking"
+	/>
+</Frame>
+
+This enables Cline to work on long-horizon tasks, seamlessly managing the context sent to LLMs, and keeping Cline on track across many context window resets.
+
+<Tip>
+  Focus Chain works particularly well with Cline's [Deep Planning slash command](/features/slash-commands/deep-planning), providing seamless progress tracking for implementation tasks created through the [planning process](/features/plan-and-act).
+</Tip>
+
+## Key Features
+
+### Automatic Todo List Generation
+
+Cline analyzes your task and automatically creates a comprehensive todo list with:
+- Clear, actionable items in markdown checklist format
+- Logical breakdown of complex tasks into manageable steps
+- Real-time updates as work progresses
+
+### User-Editable Todo Lists
+
+Todo lists are stored as editable markdown files:
+- Direct editing through your preferred markdown editor
+- Automatic detection of changes you make
+- Seamless integration back into Cline's workflow
+- Quick access through the edit button in the task header
+
+### Visual Progress Tracking
+
+The task header displays clear progress indicators:
+- **Step counters** showing current progress (e.g., "3/8")
+- **Completed items** clearly marked with checkmarks
+- **Current work** highlighted with indicators
+- **Expandable view** to see the full todo list
+
+### Smart Reminder System
+
+Configurable reminders ensure todo lists stay current:
+- Default reminder every 6 messages (customizable 1-100)
+- Automatic prompts when switching from Plan Mode to Act Mode
+- User-triggered updates when todo lists are manually edited
+
+
+## Getting Started
+
+<Steps>
+  <Step title="Open Cline Settings">
+      - Click the gear icon in the Cline sidebar
+      - Navigate to the "Features" section
+  </Step>
+  <Step title="Enable Focus Chain">
+      - Check "Enable Focus Chain"
+      - Optionally adjust "Remind Cline Interval" (default: 6 messages)
+  </Step>
+  <Step title="Start a New Task">
+      - Begin a new task
+      - Cline will automatically start creating and managing todo lists
+  </Step>
+</Steps>
+
+| Setting | Default | Range | Description |
+|---------|---------|-------|-------------|
+| Enable Focus Chain | Disabled | On/Off | Enables enhanced task progress tracking |
+| Remind Cline Interval | 6 | 1-100 messages | How often Cline updates the todo list |
+
+## Usage Examples
+
+#### 1. Task Initiation
+
+When you start a new task with Focus Chain enabled:
+
+``` markdown User Request
+User: "Create a user authentication system for my React app"
+
+Cline: [Analyzes request and creates todo list]
+```
+
+#### 2. Todo List Created
+
+Cline creates a comprehensive plan for the task, stored in a markdown file:
+
+```markdown Todo List Created
+- [ ] Set up project structure
+- [ ] Install authentication dependencies
+- [ ] Create user registration component
+- [ ] Implement login functionality
+- [ ] Add password validation
+- [ ] Set up user database schema
+- [ ] Write authentication tests
+- [ ] Deploy to staging environment
+```
+
+#### 3. Progress Tracking
+
+As Cline works, the task header shows real-time progress:
+
+```markdown Todo List Header
+[3/8] Implement login functionality     ⌄
+```
+
+Click to expand and see the full list:
+
+```markdown Full Todo List
+✓ Set up project structure
+✓ Install authentication dependencies
+✓ Create user registration component
+○ Implement login functionality  ← Currently working
+○ Add password validation
+○ Set up user database schema
+○ Write authentication tests
+○ Deploy to staging environment
+```
+
+#### 4. User Editing
+
+Need to tweak the todo list? No problem.
+
+<Steps>
+  <Step title="Open the todo list">
+    Click the edit button in the expanded todo view
+  </Step>
+  <Step title="Edit the markdown file">
+    A markdown file opens in your editor:
+
+    ```markdown Editing Todo List
+    # Focus Chain Todo List for Task abc123
+
+    <!-- Edit this markdown file to update your focus chain todo list -->
+    <!-- Use - [ ] for incomplete items and - [x] for completed items -->
+
+    - [x] Set up project structure
+    - [x] Install authentication dependencies (e.g., Firebase Auth)
+    - [x] Create user registration component
+    - [ ] Implement login functionality
+    - [ ] Add password reset feature
+    - [ ] Set up protected routes
+    - [ ] Implement logout functionality
+    - [ ] Add user profile page
+    - [ ] Write authentication tests
+    - [ ] Deploy to staging environment
+
+    <!-- Save this file to update the task's todo list -->
+    ```
+  </Step>
+  <Step title="Make your changes">
+    Add, remove, or reorder items as needed
+  </Step>
+  <Step title="Save the file">
+    Cline automatically detects and uses your updates
+  </Step>
+</Steps>
+
+## File Structure
+
+### Todo List Storage
+
+Todo lists are stored as markdown files in your task directory:
+
+``` markdown
+<VSCode Global Storage>/
+  tasks/
+    <taskId>/
+      focus_chain_taskid_<taskId>.md
+      ... other task files
+```
+
+### Markdown Format
+
+Todo files use standard markdown checklist syntax:
+
+```markdown Example Todo Syntax
+# Focus Chain Todo List for Task abc123
+
+<!-- Edit this markdown file to update your focus chain todo list -->
+<!-- Use the format: - [ ] for incomplete items and - [x] for completed items -->
+
+- [x] Set up project structure
+- [x] Install authentication dependencies
+- [ ] Create user registration component
+- [ ] Implement login functionality
+- [ ] Add password validation
+- [ ] Set up user database schema
+- [ ] Write authentication tests
+- [ ] Deploy to staging environment
+
+<!-- Save this file and the todo list will be updated in the task -->
+```
+
+
+## Integration with Plan/Act Mode
+
+Focus Chain works seamlessly with Cline's [Plan/Act mode](/features/plan-and-act):
+
+- **Plan Mode**: Optional todo lists for presenting concrete steps
+- **Act Mode**: Automatic todo creation when switching from Plan Mode
+
+<Tip>
+  For complex projects, start in Plan Mode to discuss and refine your approach before switching to Act Mode for implementation.
+</Tip>
+
+## Best Practices
+
+<AccordionGroup>
+  <Accordion title="For Effective Todo Lists">
+    1. **Start with Clear Requests**
+      - Provide detailed initial task descriptions
+      - Include specific requirements and constraints
+      - Mention any preferred technologies or approaches
+
+    2. **Review Generated Lists**
+      - Check that Cline's breakdown aligns with your expectations
+      - Verify that all important steps are included
+      - Ensure the order makes sense for your project
+
+    3. **Edit When Needed**
+      - Add missing steps you identify
+      - Remove unnecessary items
+      - Reorder steps for better workflow
+      - Add more specific details to general items
+  </Accordion>
+  <Accordion title="For Complex Projects">
+    1. **Use Plan Mode First**
+      - Discuss the approach before implementation
+      - Refine requirements through conversation
+      - Switch to Act Mode when ready to begin work
+
+    2. **Break Down Large Tasks**
+      - Split complex projects into smaller, manageable tasks
+      - Create separate todo lists for different components
+      - Focus on one major area at a time
+
+    3. **Regular Reviews**
+      - Check progress periodically during long tasks
+      - Update todo lists as requirements evolve
+      - Communicate changes to Cline through edits
+  </Accordion>
+  <Accordion title="For Collaboration">
+    1. **Share Todo Files**
+      - Todo markdown files can be shared with team members
+      - Include in version control for project documentation
+      - Use as basis for project planning discussions
+
+    2. **Consistent Format**
+      - Follow the standard markdown checklist format
+      - Keep item descriptions clear and actionable
+      - Use consistent terminology across todo lists
+  </Accordion>
+</AccordionGroup>
+
+
+## Troubleshooting
+
+Having issues? Try these quick fixes:
+
+<AccordionGroup>
+  <Accordion title="Todo list not updating?">
+    - Check that Focus Chain is enabled in settings
+    - Focus Chain may not work as well with smaller, less capable models
+    - Ensure file permissions are correct in the task directory
+  </Accordion>
+  <Accordion title="Can't edit todo file?">
+    - Verify your editor supports markdown
+    - Check VSCode has write permissions for the directory
+  </Accordion>
+  <Accordion title="Progress not displaying?">
+    - Ensure todo items use correct syntax (`- [ ]` and `- [x]`)
+    - Verify the markdown file is properly formatted
+  </Accordion>
+</AccordionGroup>
+
+Still stuck? Use the [/reportbug](/features/slash-commands/report-bug) command in Cline to get help.
+
+## Technical Details (for the curious)
+
+<AccordionGroup>
+  <Accordion title="File Monitoring">
+    - Real-time file watching detects changes to todo markdown files
+    - Automatic synchronization between file edits and UI updates
+    - Graceful handling of file creation, modification, and deletion
+  </Accordion>
+  <Accordion title="Progress Calculation">
+    - Dynamic counting of completed vs. total todo items
+    - Support for both `- [x]` and `- [X]` completion syntax
+    - Unicode symbols (✓, ○) for enhanced visual display
+  </Accordion>
+  <Accordion title="Privacy Considerations">
+    - Todo lists stored locally in VSCode workspace
+    - No todo content transmitted to external services
+    - Usage telemetry (can be disabled in settings)
+  </Accordion>
+</AccordionGroup>
+
+Focus Chain turns Cline into your personal project manager, keeping you on track and your tasks organized. Give it a try on your next project!

docs/features/slash-commands/deep-planning.mdx

@@ -0,0 +1,160 @@
+---
+title: "Deep Planning Command"
+sidebarTitle: "/deep-planning"
+---
+
+`/deep-planning` transforms Cline into a meticulous architect who investigates your codebase, asks clarifying questions, and creates a comprehensive implementation plan before writing a single line of code. 
+
+<Frame>
+	<img
+		src="https://storage.googleapis.com/cline_public_images/docs/assets/deep-planning.png"
+		alt="Deep Planning command in action showing investigation and planning process"
+	/>
+</Frame>
+
+When you use `/deep-planning`, Cline follows a four-step process that mirrors how senior developers approach complex features: thorough investigation, discussion & clarification of requirements, detailed planning, and structured task creation with progress tracking.
+
+## The Four-Step Process
+
+### Step 1: Silent Investigation
+
+Cline becomes a detective, silently exploring your codebase to understand its structure, patterns, and constraints. He examines source files, analyzes import patterns, discovers class hierarchies, and identifies technical debt markers. No commentary, no narration - just focused research.
+
+During this phase, Cline runs commands like:
+- Finding all class and function definitions across your codebase
+- Analyzing import patterns to understand dependencies
+- Discovering project structure and file organization
+- Identifying TODOs and technical debt
+
+### Step 2: Discussion and Questions
+
+Once Cline understands your codebase, he asks targeted questions that will shape the implementation. These aren't generic questions - they're specific to your project and the feature you're building.
+
+Questions might cover:
+- Clarifying ambiguous requirements
+- Choosing between equally valid implementation approaches
+- Confirming assumptions about system behavior
+- Understanding preferences for technical decisions
+
+### Step 3: Implementation Plan Document
+
+Cline creates a structured markdown document (`implementation_plan.md`) that serves as your implementation blueprint. This isn't a vague outline - it's a detailed specification with exact file paths, function signatures, and implementation order.
+
+The plan includes eight comprehensive sections:
+- **Overview**: The goal and high-level approach
+- **Types**: Complete type definitions and data structures
+- **Files**: Exact files to create, modify, or delete
+- **Functions**: New and modified functions with signatures
+- **Classes**: Class modifications and inheritance details
+- **Dependencies**: Package requirements and versions
+- **Testing**: Validation strategies and test requirements
+- **Implementation Order**: Step-by-step execution sequence
+
+### Step 4: Implementation Task Creation
+
+Cline creates a new task that references the plan document and includes trackable implementation steps. The task comes with specific commands to read each section of the plan, ensuring the implementing agent (whether that's you or Cline in Act Mode) can navigate the blueprint efficiently.
+
+<Tip>
+  Deep Planning works beautifully with [Focus Chain](/features/focus-chain). The implementation steps automatically become a todo list with real-time progress tracking, keeping complex projects organized and on track.
+</Tip>
+
+## Using Deep Planning
+
+Start a deep planning session by typing `/deep-planning` followed by your feature description:
+
+```
+/deep-planning Add user authentication with JWT tokens and role-based access control
+```
+
+Cline will begin his investigation immediately. You'll see him reading files and running commands to understand your codebase. Once he's gathered enough context, he'll engage you in discussion before creating the plan.
+
+## Example Workflow
+
+Here's how I use `/deep-planning` for a real feature:
+
+<Steps>
+  <Step title="Initiate Planning">
+    I type `/deep-planning implement a caching layer for API responses`
+  </Step>
+  <Step title="Silent Investigation">
+    Cline explores my codebase, examining:
+    - Current API structure and endpoints
+    - Existing data flow patterns
+    - Database queries and performance bottlenecks
+    - Configuration and environment setup
+  </Step>
+  <Step title="Targeted Discussion">
+    Cline asks me:
+    - "Should we use Redis or in-memory caching?"
+    - "What's the acceptable cache staleness for user data?"
+    - "Do you need cache invalidation webhooks?"
+  </Step>
+  <Step title="Plan Creation">
+    Cline generates `implementation_plan.md` with:
+    - Cache service class specifications
+    - Redis connection configuration
+    - Modified API endpoints with caching logic
+    - Cache key generation strategies
+    - TTL configurations for different data types
+  </Step>
+  <Step title="Task Generation">
+    Cline creates a new task with:
+    - Reference to the implementation plan
+    - Commands to read specific sections
+    - Trackable todo items for each implementation step
+    - Request to switch to Act Mode for execution
+  </Step>
+</Steps>
+
+## Integration with Plan/Act Mode
+
+Deep Planning is designed to work seamlessly with [Plan/Act Mode](/features/plan-and-act):
+
+- Use `/deep-planning` in Plan Mode for the investigation and planning phases
+- The generated task requests switching to Act Mode for implementation
+- Focus Chain automatically tracks progress through the implementation steps
+
+This separation ensures planning stays focused on architecture while implementation stays focused on execution.
+
+## Best Practices
+
+### When to Use Deep Planning
+
+Use `/deep-planning` for:
+- Features touching multiple parts of your codebase
+- Architectural changes requiring careful coordination
+- Complex integrations with external services
+- Refactoring efforts that need systematic execution
+- Any feature where you'd normally spend time whiteboarding
+
+### Making the Most of Investigation
+
+Let Cline complete his investigation thoroughly. The quality of the plan directly correlates with how well he understands your codebase. If you have specific areas he should examine, mention them in your initial request.
+
+### Reviewing the Plan
+
+Always review `implementation_plan.md` before starting implementation. The plan is comprehensive but not immutable - you can edit it directly if needed. Think of it as a collaborative document between you and Cline.
+
+### Tracking Progress
+
+With Focus Chain enabled, your implementation progress displays in the task header. Each completed step gets checked off automatically as Cline works through the plan, giving you real-time visibility into complex implementations.
+
+## Inspiration
+
+I use `/deep-planning` whenever I'm about to build something that would normally require a design document. Recent examples from my workflow:
+
+- **Migrating authentication systems**: Deep Planning mapped every endpoint, identified all authentication touchpoints, and created a migration plan that avoided breaking changes.
+
+- **Adding real-time features**: The plan covered WebSocket integration, event handling, state synchronization, and fallback mechanisms for disconnections.
+
+- **Database schema refactoring**: Cline identified all affected queries, created migration scripts, and planned the rollout to minimize downtime.
+
+- **API versioning implementation**: The plan detailed route changes, backward compatibility layers, deprecation notices, and client migration paths.
+
+The power of `/deep-planning` is that it forces thoughtful architecture before implementation. It's like having a senior developer review your approach before you write code, except that developer has perfect knowledge of your entire codebase.
+
+<Note>
+  Deep Planning requires models with strong reasoning capabilities. It works best with the latest generation of models, like GPT-5, Claude 4, Gemini 2.5, or Grok 4. Smaller models may struggle with the comprehensive analysis required.
+</Note>
+
+For simpler tasks that don't require extensive planning, consider using [/newtask](/features/slash-commands/new-task) to create focused tasks with context, or jump straight into implementation if the path forward is clear.

docs/getting-started/for-new-coders.mdx

@@ -3,9 +3,9 @@ title: "For New Coders"
 description: "Welcome to Cline, your AI-powered coding companion! This guide will help you quickly set up your development environment and begin your coding journey with ease."
 ---
 
-> 💡 **Tip:** If you're completely new to coding, take your time with each step. There's no rush — Cline is here to guide you!
+> **Tip:** If you're completely new to coding, take your time with each step. There's no rush — Cline is here to guide you!
 
-### 🚀 Getting Started
+### Getting Started
 
 Before you jump into coding, make sure you have these essentials ready:
 
@@ -15,9 +15,9 @@ A popular, free, and powerful code editor.
 
 -   [<u>Download VS Code</u>](https://code.visualstudio.com/)
 
-📺 **Recommended YouTube Tutorial:** [<u>How to Install VS Code</u>](https://www.youtube.com/watch?v=MlIzFUI1QGA)
+**Recommended YouTube Tutorial:** [<u>How to Install VS Code</u>](https://www.youtube.com/watch?v=MlIzFUI1QGA)
 
-> ✅ **Pro Tip:** Install VS Code in your Applications folder (macOS) or Program Files (Windows) for easy access from your dock or start menu.
+> **Pro Tip:** Install VS Code in your Applications folder (macOS) or Program Files (Windows) for easy access from your dock or start menu.
 
 #### 2. **Organize Your Projects**
 
@@ -31,7 +31,7 @@ Inside your `Cline` folder, structure projects clearly:
 -   `Documents/Cline/workout-app` _(e.g., for a fitness tracking app)_
 -   `Documents/Cline/portfolio-website` _(e.g., to showcase your work)_
 
-> 💡 **Tip:** Keeping your projects organized from the start will save you time and confusion later!
+> **Tip:** Keeping your projects organized from the start will save you time and confusion later!
 
 #### 3. **Install the Cline VS Code Extension**
 
@@ -39,9 +39,9 @@ Enhance your coding workflow by installing the Cline extension directly within V
 
 -   Get Started with Cline Extension Tutorial
 
-📺 **Recommended YouTube Tutorial:** [<u>How To Install Extensions in VS Code</u>](https://www.youtube.com/watch?v=E7trgwZa-mk)
+**Recommended YouTube Tutorial:** [<u>How To Install Extensions in VS Code</u>](https://www.youtube.com/watch?v=E7trgwZa-mk)
 
-> ✅ **Pro Tip:** After installing, reload VS Code to ensure the extension is activated properly.
+> **Pro Tip:** After installing, reload VS Code to ensure the extension is activated properly.
 
 #### 4. **Essential Development Tools**
 
@@ -51,9 +51,9 @@ Basic software required for coding efficiently:
 -   Node.js
 -   Git
 
-👉 [<u>Follow our detailed guide on Installing Essential Development Tools with step-by-step help from Cline.</u>](https://docs.cline.bot/getting-started/installing-dev-essentials#installing-dev-essentials)
+[<u>Follow our detailed guide on Installing Essential Development Tools with step-by-step help from Cline.</u>](https://docs.cline.bot/getting-started/installing-dev-essentials#installing-dev-essentials)
 
-📺 **Recommended YouTube Tutorials for Manual Installation:**
+**Recommended YouTube Tutorials for Manual Installation:**
 
 -   **For macOS:**
     -   [<u>Install Homebrew on Mac</u>](https://www.youtube.com/watch?v=hwGNgVbqasc)
@@ -63,6 +63,6 @@ Basic software required for coding efficiently:
     -   [<u>Install Git on Windows 10/11 (2024)</u>](https://www.youtube.com/watch?v=yjxv1HuRQy0)
     -   [<u>Install Node.js in Windows 10/11</u>](https://www.youtube.com/watch?v=uCgAuOYpJd0)
 
-> ⚠️ **Note:** If you run into permission issues during installation, try running your terminal or command prompt as an administrator.
+> **Note:** If you run into permission issues during installation, try running your terminal or command prompt as an administrator.
 
-🎉 You're all set! Dive in and start coding smarter and faster with **Cline**.
+You're all set! Dive in and start coding smarter and faster with **Cline**.

docs/getting-started/installing-cline-jetbrains.mdx

@@ -0,0 +1,135 @@
+---
+title: "Installing Cline for JetBrains (Early Access)"
+description: "Get early access to Cline in your favorite JetBrains IDE with the same powerful AI assistance you know from VSCode."
+---
+
+<Frame>
+	<img
+		src="https://storage.googleapis.com/cline_public_images/docs/assets/jetbrains-logo.svg"
+		alt="JetBrains logo"
+		style={{ width: "200px", height: "auto", margin: "0 auto 20px auto", display: "block" }}
+	/>
+</Frame>
+
+Cline for JetBrains works almost identically to Cline in VSCode. All the core features work properly: diff editing, using tools, logging in with different providers, MCP servers, Cline rules and workflows, and more.
+
+<Frame>
+	<img
+		src="https://storage.googleapis.com/cline_public_images/docs/assets/jetbrains-demo-hifi.gif"
+		alt="Cline running in JetBrains IDE showing AI assistance"
+	/>
+</Frame>
+
+<Note>Cline for JetBrains is in early access. All core features are functional, with ongoing improvements based on user feedback.</Note>
+
+## Installation
+
+As part of our early access program, Cline for JetBrains is available through direct download before its official marketplace release. You'll need to install it manually from a downloaded file:
+
+### Manual Installation from Disk
+
+1. **Download the Plugin:**
+   - Go to [https://plugins.jetbrains.com/plugin/28247-cline/versions/eap](https://plugins.jetbrains.com/plugin/28247-cline/versions/eap)
+   - Click **Download** to get the `.zip` file
+
+<Frame>
+	<img
+		src="https://storage.googleapis.com/cline_public_images/docs/assets/cline-jetbrains-marketplace-download.png"
+		alt="JetBrains plugin marketplace showing Cline download page"
+	/>
+</Frame>
+
+2. **Install from Disk:**
+   - Open your JetBrains IDE
+   - Go to **IntelliJ IDEA** (or whichever IDE you are in) → **Settings**
+
+<Frame>
+	<img
+		src="https://storage.googleapis.com/cline_public_images/docs/assets/cline-jetbrains-settings.png"
+		alt="JetBrains IDE settings dialog"
+	/>
+</Frame>
+
+   - Select **Plugins** from the left sidebar
+   - Click the gear icon ⚙️ and select **Install Plugin from Disk...**
+
+<Frame>
+	<img
+		src="https://storage.googleapis.com/cline_public_images/docs/assets/cline-jetbrains-install-disk.png"
+		alt="JetBrains IDE settings showing Install Plugin from Disk option"
+	/>
+</Frame>
+
+   - Select the downloaded `.zip` file
+
+<Frame>
+	<img
+		src="https://storage.googleapis.com/cline_public_images/docs/assets/cline-jetbrains-zip-file.png"
+		alt="File selection dialog showing Cline plugin zip file"
+	/>
+</Frame>
+
+   - Restart your IDE when prompted
+
+## Getting Started with Cline
+
+After installation, you'll find Cline in your IDE:
+
+1. **Open Cline:**
+   - Look for the Cline tool window (usually on the right side)
+   - Or go to **View** → **Tool Windows** → **Cline**
+
+2. **Sign In (optional, BYOK is also available):**
+   - Click **Sign In** in the Cline panel
+   - You'll be taken to [app.cline.bot](https://app.cline.bot) to create your account
+   - No credit card needed to get started with free credits
+
+3. **Start Coding:**
+   - Try this first prompt: "Hey Cline! Can you help me create a simple Hello World program in this project?"
+
+## Key Differences from VSCode
+
+While Cline for JetBrains includes all the same powerful features, there's one important difference to be aware of:
+
+**Terminal Integration:** The terminal inside JetBrains isn't integrated with Cline the same way it is in VSCode. Cline can execute commands, but the output will only appear in the webview if you expand the **Command Output** section.
+
+This means:
+- Commands still run successfully
+- You can see the output by clicking to expand Command Output in the chat
+- Terminal commands work the same way, just with a different display
+
+## What Works
+
+Everything else works exactly like VSCode:
+
+- **Diff Editing:** Cline can read, write, and edit files with the same precision
+- **Tool Usage:** All of Cline's tools (file operations, web browsing, etc.) work identically  
+- **API Providers:** Connect to Anthropic, OpenAI, local models, and more
+- **MCP Servers:** Full support for Model Context Protocol servers
+- **Cline Rules:** Custom instructions and workflows work the same way
+- **@ Mentions:** Reference files, folders, problems, and more
+- **Drag & Drop:** Add files and images to conversations
+
+## Tips for JetBrains Users
+
+- **Project Context:** Cline automatically understands your project structure, just like in VSCode
+- **Language Support:** Cline works with any language your JetBrains IDE supports
+- **Debugging Help:** Share error messages and stack traces directly in the chat
+- **Code Review:** Ask Cline to review your code changes before committing
+
+## Troubleshooting
+
+If you don't see the Cline tool window after installation:
+- Restart your IDE completely
+- Check **View** → **Tool Windows** → **Cline**
+- Ensure the plugin is enabled in **Settings** → **Plugins**
+
+Having other issues? Join our [Discord community](https://discord.gg/cline) for help from the team and other users.
+
+## Next Steps
+
+Now that you have Cline installed, you might want to:
+- Learn about [model selection](/getting-started/model-selection-guide) to choose the best AI provider
+- Explore [@ mentions](/features/at-mentions/overview) to reference files and context efficiently  
+- Set up [Cline rules](/features/cline-rules) for your specific workflow
+- Try [MCP servers](/mcp/mcp-overview) to extend Cline's capabilities

docs/getting-started/installing-cline.mdx

@@ -9,13 +9,13 @@ description: "Cline is a VS Code extension that brings AI-powered coding assista
 -   **VS Code Marketplace (Recommended):** Fastest method for standard VS Code and Cursor users.
 -   **Open VSX Registry:** For VS Code-compatible editors like VSCodium.
 
-### 🛠️ VS Code Marketplace: Step-by-Step Setup
+### VS Code Marketplace: Step-by-Step Setup
 
 Follow these steps to get Cline up and running:
 
 1.  **Open VS Code:** Launch the VS Code application.
 
-    > ⚠️ **Note:** If VS Code shows "Running extensions might...", click "Allow".
+    > **Note:** If VS Code shows "Running extensions might...", click "Allow".
 
 2.  **Open Your Cline Folder:** In VS Code, open the Cline folder you created in Documents.
 3.  **Navigate to Extensions:** Click on the Extensions icon in the Activity Bar on the side of VS Code (`Ctrl + Shift + X` or `Cmd + Shift + X`).
@@ -34,9 +34,9 @@ Follow these steps to get Cline up and running:
     - Or, use the command palette (`Ctrl/Cmd + Shift + P`) and type "Cline: Open In New Tab" for a better view.
 3. **Troubleshooting:** If you don't see the Cline icon, try restarting VS Code.
 
-> ✅ **Pro Tip:** You should see the Cline chat window appear in your VS Code editor!
+> **Pro Tip:** You should see the Cline chat window appear in your VS Code editor!
 
-### 🌐 Open VSX Registry
+### Open VSX Registry
 
 For VS Code-compatible editors without Marketplace access (like VSCodium and Windsurf):
 
@@ -46,7 +46,7 @@ For VS Code-compatible editors without Marketplace access (like VSCodium and Win
 4. Select "Cline" by saoudrizwan and click **Install**.
 5. Reload if prompted.
 
-### 👤 Creating Your Cline Account
+### Creating Your Cline Account
 
 Now that you have Cline installed, let's get you set up with your account:
 
@@ -61,7 +61,7 @@ Now that you have Cline installed, let's get you set up with your account:
     - Google Gemini 2.0 Flash
     - And more — all through your Cline account.
 
-### 💻 Your First Interaction with Cline
+### Your First Interaction with Cline
 
 You're ready to start building! Copy and paste this prompt into the Cline chat window:
 
@@ -69,15 +69,15 @@ You're ready to start building! Copy and paste this prompt into the Cline chat w
 Hey Cline! Could you help me create a new project folder called "hello-world" in my Cline directory and make a simple webpage that says "Hello World" in big blue text?
 ```
 
-> ✅ **Pro Tip:** Cline will help you create the project folder and set up your first webpage!
+> **Pro Tip:** Cline will help you create the project folder and set up your first webpage!
 
-### 🧩 Tips for Working with Cline
+### Tips for Working with Cline
 
 -   **Ask Questions:** If you're unsure about something, ask Cline!
 -   **Use Screenshots:** Cline can understand images — show him what you're working on.
 -   **Copy and Paste Errors:** Share error messages in the chat for solutions.
 -   **Speak Plainly:** Use your own words — Cline will translate them into code.
 
-### 🫂 Still Struggling?
+### Still Struggling?
 
 Join our Discord community and engage with our team and other Cline users directly.

docs/getting-started/installing-dev-essentials.mdx

@@ -6,7 +6,7 @@ description: >-
     guided way.
 ---
 
-### 🧰 The Essential Tools
+### The Essential Tools
 
 Here are the core tools you'll need for development:
 
@@ -17,9 +17,9 @@ Here are the core tools you'll need for development:
     -   Chocolatey for Windows
     -   apt/yum for Linux
 
-> 💡 **Tip:** These tools are the foundation of your developer toolkit. Installing them properly will set you up for success!
+> **Tip:** These tools are the foundation of your developer toolkit. Installing them properly will set you up for success!
 
-### 🚀 Let Cline Install Everything
+### Let Cline Install Everything
 
 Copy one of these prompts based on your operating system and paste it into **Cline**:
 
@@ -41,9 +41,9 @@ Hello Cline! I need help setting up my Windows PC for software development. Coul
 Hello Cline! I need help setting up my Linux system for software development. Could you please help me install the essential development tools like Node.js, Git, and any other core utilities that are commonly needed for coding? I'd like you to guide me through the process step-by-step.
 ```
 
-> ✅ **Pro Tip:** Cline will show you each command before running it. You stay in control the entire time!
+> **Pro Tip:** Cline will show you each command before running it. You stay in control the entire time!
 
-### 🔍 What Will Happen
+### What Will Happen
 
 Cline will guide you through the following steps:
 
@@ -52,9 +52,9 @@ Cline will guide you through the following steps:
 3. Showing you the exact command before it runs (you approve each step!)
 4. Verifying each installation is successful
 
-> ⚠️ **Note:** You might need to enter your computer's password for some installations. This is normal!
+> **Note:** You might need to enter your computer's password for some installations. This is normal!
 
-### 💡 Why These Tools Are Important
+### Why These Tools Are Important
 
 -   **Node.js & npm:**
     -   Build websites with frameworks like React or Next.js
@@ -68,15 +68,15 @@ Cline will guide you through the following steps:
     -   Quickly install and update development tools
     -   Keep your environment organized and up to date
 
-### 🧩 Notes
+### Notes
 
-> 💡 **Tip:** The installation process is interactive — Cline will guide you step by step!
+> **Tip:** The installation process is interactive — Cline will guide you step by step!
 
 -   All commands are shown to you for approval before they run.
 -   If you run into any issues, Cline will help troubleshoot them.
 -   You may need to enter your computer's password for certain steps.
 
-### 🧑‍💻 Additional Tips for New Coders
+### Additional Tips for New Coders
 
 #### Understanding the Terminal
 

docs/getting-started/model-selection-guide.mdx

@@ -1,126 +1,79 @@
 ---
 title: "Model Selection Guide"
-description: "Last updated: Feb 5, 2025."
+description: "Last updated: August 20, 2025."
 ---
 
-## Understanding Context Windows
+New models drop constantly, so this guide focuses on what's working well with Cline right now. We'll keep it updated as the landscape shifts.
 
-Think of a context window as your AI assistant's working memory - similar to RAM in a computer. It determines how much information the model can "remember" and process at once during your conversation. This includes:
+## Current Top Models
 
--   Your code files and conversations
--   The assistant's responses
--   Any documentation or additional context provided
+| Model | Context Window | Input Price* | Output Price* | Best For |
+|-------|---------------|--------------|---------------|----------|
+| **Claude Sonnet 4** | 1M tokens | $3-6 | $15-22.50 | Reliable tool usage, complex codebases |
+| **Qwen3 Coder** | 256K tokens | $0.20 | $0.80 | Coding tasks, open source flexibility |
+| **Gemini 2.5 Pro** | 1M+ tokens | TBD | TBD | Large codebases, document analysis |
+| **GPT-5** | 400K tokens | $1.25 | $10 | Latest OpenAI tech, three modes |
 
-Context windows are measured in tokens (roughly 3/4 of a word in English). Different models have different context window sizes:
+*Per million tokens
 
--   Claude 3.5 Sonnet: 200K tokens
--   DeepSeek Models: 128K tokens
--   Gemini Flash 2.0: 1M tokens
--   Gemini 1.5 Pro: 2M tokens
+## Budget Options
 
-When you reach the limit of your context window, older information needs to be removed to make room for new information - just like clearing RAM to run new programs. This is why sometimes AI assistants might seem to "forget" earlier parts of your conversation.
+| Model | Context Window | Input Price* | Output Price* | Notes |
+|-------|---------------|--------------|---------------|-------|
+| **DeepSeek V3** | 128K tokens | $0.14 | $0.28 | Great value for daily coding |
+| **DeepSeek R1** | 128K tokens | $0.55 | $2.19 | Budget reasoning champion |
+| **Qwen3 32B** | 128K tokens | Varies | Varies | Open source, multiple providers |
+| **Z AI GLM 4.5** | 128K tokens | TBD | TBD | MIT licensed, hybrid reasoning |
 
-Cline helps you manage this limitation with its Context Window Progress Bar, which shows:
+*Per million tokens
 
--   Input tokens (what you've sent to the model)
--   Output tokens (what the model has generated)
--   A visual representation of how much of your context window you've used
--   The total capacity for your chosen model
 
-<Frame caption="Visual representation of the context window usage in Cline">
-	<img
-		src="https://storage.googleapis.com/cline_public_images/docs/assets/image%20(11).png"
-		alt="Context window progress bar example"
-	/>
-</Frame>
+## Context Window Guide
 
-This visibility helps you work more effectively with Cline by letting you know when you might need to start fresh or break tasks into smaller chunks.
+| Size | Word Count | Use Case |
+|------|------------|----------|
+| 32K tokens | ~24,000 words | Single files, small projects |
+| 128K tokens | ~96,000 words | Most coding projects |
+| 200K tokens | ~150,000 words | Large codebases |
+| 400K+ tokens | ~300,000+ words | Entire applications |
 
-### Model Comparison
+**Performance note**: Most models start dropping in quality around 400-500K tokens, even if they claim higher limits.
 
-## LLM Model Comparison for Cline (Feb 2025)
+## Open Source vs Closed Source
 
-| Model             | Input Cost\* | Output Cost\* | Context Window | Best For                            |
-| ----------------- | ------------ | ------------- | -------------- | ----------------------------------- |
-| Claude 3.5 Sonnet | $3.00        | $15.00        | 200K           | Best code implementation & tool use |
-| DeepSeek R1       | $0.55        | $2.19         | 128K           | Planning & reasoning champion       |
-| DeepSeek V3       | $0.14        | $0.28         | 128K           | Value code implementation           |
-| o3-mini           | $1.10        | $4.40         | 200K           | Flexible use, strong planning       |
-| Gemini Flash 2.0  | $0.00        | $0.00         | 1M             | Strong all-rounder                  |
-| Gemini 1.5 Pro    | $0.00        | $0.00         | 2M             | Large context processing            |
+### Open Source Advantages
+- **Multiple providers** compete to host them
+- **Cheaper pricing** due to competition  
+- **Provider choice** - switch if one goes down
+- **Faster innovation** cycles
 
-\*Costs per million tokens
+### Open Source Models Available
+- **Qwen3 Coder** (Apache 2.0)
+- **Z AI GLM 4.5** (MIT)
+- **Kimi K2** (Open source)
+- **DeepSeek series** (Various licenses)
 
-### Top Picks for 2025
+## Quick Decision Matrix
 
-1. **Claude 3.5 Sonnet**
-    - Best overall code implementation
-    - Most reliable tool usage
-    - Expensive but worth it for critical code
-2. **DeepSeek R1**
-    - Exceptional planning & reasoning
-    - Great value pricing
-3. **o3-mini**
-    - Strong for planning with adjustable reasoning
-    - Three reasoning modes for different needs
-    - Requires OpenAI Tier 3 API access
-    - 200K context window
-4. **DeepSeek V3**
-    - Reliable code implementation
-    - Great for daily coding
-    - Cost-effective for implementation
-5. **Gemini Flash 2.0**
-    - Massive 1M context window
-    - Improved speed and performance
-    - Good all-around capabilities
+| If you want... | Use this |
+|----------------|----------|
+| Something that just works | Claude Sonnet 4 |
+| To save money | DeepSeek V3 or Qwen3 variants |
+| Huge context windows | Gemini 2.5 Pro or Claude Sonnet 4 |
+| Open source | Qwen3 Coder, Z AI GLM 4.5, or Kimi K2 |
+| Latest tech | GPT-5 |
+| Speed | Qwen3 Coder on Cerebras (fastest available) |
 
-### Best Models by Mode (Plan or Act)
+## What Others Are Using
 
-#### Planning
+Check [OpenRouter's Cline usage stats](https://openrouter.ai/apps?url=https%3A%2F%2Fcline.bot%2F) to see real usage patterns from the community.
 
-1. **DeepSeek R1**
-    - Best reasoning capabilities in class
-    - Excellent at breaking down complex tasks
-    - Strong math/algorithm planning
-    - MoE architecture helps with reasoning
-2. **o3-mini (high reasoning)**
-    - Three reasoning levels:
-        - High: Complex planning
-        - Medium: Daily tasks
-        - Low: Quick ideas
-    - 200K context helps with large projects
-3. **Gemini Flash 2.0**
-    - Massive context window for complex planning
-    - Strong reasoning capabilities
-    - Good with multi-step tasks
+## Context Management
 
-#### Acting (coding)
+Cline automatically handles context limits with [auto-compact](/features/auto-compact). When you approach your model's limit, Cline summarizes the conversation to keep working. You don't need to micromanage this.
 
-1. **Claude 3.5 Sonnet**
-    - Best code quality
-    - Most reliable with Cline tools
-    - Worth the premium for critical code
-2. **DeepSeek V3**
-    - Nearly Sonnet-level code quality
-    - Better API stability than R1
-    - Great for daily coding
-    - Strong tool usage
-3. **Gemini 1.5 Pro**
-    - 2M context window
-    - Good with complex codebases
-    - Reliable API
-    - Strong multi-file understanding
+## The Bottom Line
 
-### A Note on Local Models
+Start with **Claude Sonnet 4** if you want reliability. Experiment with **open source options** once you're comfortable to find the best fit for your workflow and budget.
 
-While running models locally might seem appealing for cost savings, we currently don't recommend any local models for use with Cline. [Local models are significantly less reliable](https://docs.cline.bot/running-models-locally/read-me-first) at using Cline's essential tools and typically retain only 1-26% of the original model's capabilities. The full cloud version of DeepSeek-R1, for example, is 671B parameters - local versions are drastically simplified copies that struggle with complex tasks and tool usage. Even with high-end hardware (RTX 3070+, 32GB+ RAM), you'll experience slower responses, less reliable tool execution, and reduced capabilities. For the best development experience, we recommend sticking with the cloud models listed above.
-
-### Key Takeaways
-
-1. **Plan vs Act Matters**: Choose models based on task type
-2. **Real Performance > Benchmarks**: Focus on actual Cline performance
-3. **Mix & Match**: Use different models for planning and implementation
-4. **Cost vs Quality**: Premium models worth it for critical code
-5. **Keep Backups**: Have alternatives ready for API issues
-
-_\*Note: Based on real usage patterns and community feedback rather than just benchmarks. Your experience may vary. This is not an exhaustive list of all the models available for use within Cline._
+The landscape moves fast - these recommendations reflect what's working now, but keep an eye on new releases.

docs/getting-started/understanding-context-management.mdx

@@ -3,7 +3,7 @@ title: "Context Management"
 description: "Context is key to getting the most out of Cline"
 ---
 
-> 💡 **Quick Reference**
+> **Quick Reference**
 >
 > -   Context = The information Cline knows about your project
 > -   Context Window = How much information Cline can hold at once
@@ -38,7 +38,7 @@ Cline actively builds context in two ways:
     - Guide focus areas
     - Share design thoughts and requirements
 
-💡 **Key Point**: Cline isn't passive - it actively seeks to understand your project. You can either let it explore or guide its focus, especially in [Plan](https://docs.cline.bot/features/plan-and-act) mode.
+**Key Point**: Cline isn't passive - it actively seeks to understand your project. You can either let it explore or guide its focus, especially in [Plan Mode](/features/plan-and-act).
 
 ### Context & Context Windows
 
@@ -53,12 +53,13 @@ Think of context like a whiteboard you and Cline share:
 -   **Context Window** is the size of the whiteboard itself:
     -   Measured in tokens (1 token ≈ 3/4 of an English word)
     -   Each model has a fixed size:
-        -   Claude 3.5 Sonnet: 200,000 tokens
-        -   DeepSeek: 64,000 tokens
-    -   When the whiteboard is full, you need to erase (clear context) to write more
-        -   [How Cline manages context under the hood](https://cline.bot/blog/understanding-the-new-context-window-progress-bar-in-cline)
+        -   Claude Sonnet 4: 1,000,000 tokens
+        -   Qwen3 Coder: 256,000 tokens
+        -   Gemini 2.5 Pro: 1,000,000+ tokens
+        -   GPT-5: 400,000 tokens
+    -   When the whiteboard is full, Cline automatically summarizes the conversation to free up space
 
-⚠️ **Important**: Having a large context window (like Claude's 200k tokens) doesn't mean you should fill it completely. Just like a cluttered whiteboard, too much information can make it harder to focus on what's important.
+**Important**: Having a large context window doesn't mean you should fill it completely. Models start degrading around 400-500K tokens even if they claim higher limits. Just like a cluttered whiteboard, too much information can make it harder to focus on what's important.
 
 ## Understanding the Context Window Progress Bar
 
@@ -76,7 +77,7 @@ Cline provides a visual way to monitor your context window usage through a progr
 -   ↑ shows input tokens (what you've sent to the LLM)
 -   ↓ shows output tokens (what the LLM has generated)
 -   The progress bar visualizes how much of your context window you've used
--   The total shows your model's maximum capacity (e.g., 200k for Claude 3.5-Sonnet)
+-   The total shows your model's maximum capacity (e.g., 1M for Claude Sonnet 4)
 
 ### When to Watch the Bar
 
@@ -85,7 +86,33 @@ Cline provides a visual way to monitor your context window usage through a progr
 -   Before starting complex tasks
 -   When Cline seems to lose context
 
-💡 **Tip**: Consider starting a fresh session when usage reaches 70-80% to maintain optimal performance.
+**Tip**: With [Auto Compact](/features/auto-compact), Cline can now handle long conversations automatically. When combined with [Focus Chain](/features/focus-chain), you can work on complex projects that span multiple context windows without losing progress.
+
+## Automatic Context Management
+
+Cline includes intelligent features to manage context automatically:
+
+### Default Settings You Should Keep On
+
+**Focus Chain** - Enabled by default in v3.25. Cline generates a todo list at task start and keeps it in context so the thread doesn't drift. You can edit the markdown to add or reorder steps and Cline will adapt. [Learn more about Focus Chain](/features/focus-chain).
+
+**Auto Compact** - Always on. As the context window reaches its limit, Cline creates a comprehensive summary, replaces the bloated history, and continues where it left off. Decisions, code changes, and state are preserved. [Learn more about Auto Compact](/features/auto-compact).
+
+## Advanced Context Tools
+
+When you need more control over context management:
+
+### Deep Planning (`/deep-planning`)
+For substantial features, refactors, or integrations. Cline investigates your codebase, asks targeted questions, then writes `implementation_plan.md`. It creates a fresh task with distilled, high-value context. [Learn more about Deep Planning](/features/slash-commands/deep-planning).
+
+### New Task (`/newtask`)
+At natural transition points, packages only what matters into a fresh task. Clean slate for implementation after research, or crisp handoff between teammates. [Learn more about New Task](/features/slash-commands/new-task).
+
+### Smol (`/smol`)
+Compress the conversation in place to keep momentum. Ideal during debugging or exploratory work when you don't want to break flow. [Learn more about Smol](/features/slash-commands/smol).
+
+### Memory Bank + .clinerules
+For non-trivial projects. The Memory Bank captures project knowledge as Markdown in your repo. `.clinerules` are version-controlled instructions that align Cline's behavior with your team. [Learn more about Memory Bank](/prompting/cline-memory-bank) and [Cline Rules](/features/cline-rules).
 
 ## Working with Context Files
 
@@ -93,12 +120,12 @@ Context files help maintain understanding across sessions. They serve as documen
 
 #### Approaches to Context Files
 
-1. **Evergreen Project Context (i.e.** [**Memory Bank**](https://docs.cline.bot/prompting/cline-memory-bank)**)**
+1. **Evergreen Project Context (Memory Bank)**
     - Living documentation that evolves with your project
     - Updated as architecture and patterns emerge
     - Example: The Memory Bank pattern maintains files like `techContext.md` and `systemPatterns.md`
     - Useful for long-running projects and teams
-2. **Task-Specific Context (i.e.** [**Structured Approach**](https://cline.bot/blog/building-advanced-software-with-cline-a-structured-approach)**)**
+2. **Task-Specific Context**
 
     - Created for specific implementation tasks
     - Document requirements, constraints, and decisions
@@ -151,9 +178,19 @@ Context files help maintain understanding across sessions. They serve as documen
     - Use Plan mode for complex discussions
     - Start fresh sessions when needed
 3. **Team Projects**
-    - Share common context files (consider using [.clinerules](https://docs.cline.bot/features/cline-rules) files in project roots)
+    - Share common context files (consider using [.clinerules](/features/cline-rules) files in project roots)
     - Document architectural decisions
     - Maintain consistent patterns
     - Keep documentation current
 
-Remember: The goal is to help Cline maintain consistent understanding of your project across sessions.
+## Bonus Context Tips
+
+- You can @ links and have the webpage's context added to Cline (docs, blogs, etc.)
+- Utilize MCP servers to pull in context from your external knowledge bases
+- Screenshots can be used as context for models that support image inputs
+
+## The Bottom Line
+
+Cline already does a lot of context work for you - [Focus Chain](/features/focus-chain), [Auto Compact](/features/auto-compact), and the planning flow are designed to keep the thread intact across long horizons. The goal is to help Cline maintain consistent understanding of your project across sessions.
+
+Remember: The goal is to keep only what matters in view, at every step.

docs/getting-started/what-is-cline.mdx

@@ -3,6 +3,70 @@ title: "What is Cline?"
 description: "An introduction to Cline, your AI-powered development assistant in VS Code."
 ---
 
-Cline is an AI development assistant which integrates with Microsoft Visual Studio Code. It provides an interface between your IDE and LLMs facilitating code development, increasing productivity and lowering the barrier to entry for new coders. Depending on permissions, Cline can read/write files, execute commands, use your web browser, and expand its capabilities with Model Context Protocol servers.
+Cline is an open source AI coding agent that brings frontier AI models directly to your VS Code editor. Unlike autocomplete tools, Cline is a true coding agent that can understand entire codebases, plan complex changes, and execute multi-step tasks.
 
-What makes Cline distinctive is its thoughtful approach to code generation and its extensive integration capabilities. Rather than simply generating code snippets, Cline collaborates with developers by planning solutions step-by-step, maintaining awareness of the entire development environment, and requiring explicit approval for all changes. It can understand large codebases, accelerate onboarding for new engineers, and connect with hundreds of tools through its Model Context Protocol Marketplace, enabling everything from streamlined project deployments to automated incident response—all through natural language commands.
+## Open Source AI Coding, Uncompromised
+
+Cline gives you direct, transparent access to frontier AI with no limits, no surprises, and no model ecosystem lock-in. See every decision. Choose any model. Control your costs.
+
+### Complete Transparency
+
+Watch in real-time as Cline reads files, considers approaches, and proposes changes. Every decision is visible, every edit reviewable before it's made. This isn't just "explainable AI" - it's complete transparency.
+
+### Your Models, Your Control
+
+Use Claude for complex reasoning, Gemini for massive contexts, or Qwen3 Coder for efficiency. Switch instantly as new models launch. Your API keys, your choice. No gatekeeping innovation.
+
+### Built for Real Engineering
+
+Cline can:
+- **Read and write files** across your entire codebase
+- **Execute terminal commands** and debug errors
+- **Plan complex features** before writing code
+- **Connect to external systems** through MCP servers
+- **Understand large codebases** with intelligent context management
+
+## Plan & Act Mode
+
+Cline explores your codebase and works with you to create comprehensive plans before writing a single line of code, ensuring it understands the full context of your project.
+
+**Plan Mode** for complex tasks - Cline explores, asks questions, and creates detailed implementation plans.
+
+**Act Mode** for execution - Cline implements the plan with full transparency and control.
+
+## Zero Trust by Design
+
+Your code never touches our servers. Cline runs entirely client-side with your API keys, making it the only option for enterprises with strict security requirements.
+
+**Open source** means your security team can review every line. See exactly how Cline works, what it sends to AI providers, and how decisions are made.
+
+## Key Features
+
+### Focus Chain
+Automatic todo list management with real-time progress tracking throughout your tasks. Keeps Cline on track across long projects.
+
+### Auto Compact
+When conversations get long, Cline automatically summarizes to preserve context while freeing up space to continue working.
+
+### Deep Planning
+For complex features, Cline investigates your codebase, asks clarifying questions, and creates comprehensive implementation plans.
+
+### MCP Integration
+Connect to databases, APIs, and documentation through the Model Context Protocol. Cline becomes your bridge to any external system.
+
+### .clinerules
+Define project-specific instructions that Cline follows including coding standards, architecture patterns, or team conventions.
+
+## Why Developers Choose Cline
+
+**100% Open Source** - Every line of code on GitHub. 48k+ stars from developers who've read it, improved it, and trust it with their work.
+
+**No Inference Games** - We don't profit from AI usage. While others limit context or route to cheaper models, we give you unrestricted access to any model's full capabilities.
+
+**Future-Proof by Design** - New model released? Use it immediately. Cline works with any AI provider, any model.
+
+**True Visibility** - See every file read, every decision considered, every token used.
+
+## Getting Started
+
+Ready to experience AI coding without limits? [Install Cline](/getting-started/installing-cline) and start with our [Model Selection Guide](/getting-started/model-selection-guide) to choose the right AI model for your needs.

docs/provider-config/anthropic.mdx

@@ -16,6 +16,7 @@ description: "Learn how to configure and use Anthropic Claude models with Cline.
 
 Cline supports the following Anthropic Claude models:
 
+-   `claude-opus-4-1-20250805`
 -   `claude-opus-4-20250514`
 -   `claude-opus-4-20250514:thinking` (Extended Thinking variant)
 -   `claude-sonnet-4-20250514` (Recommended)

docs/provider-config/aws-bedrock/api-key.mdx

@@ -1,6 +1,7 @@
 ---
-title: "AWS Bedrock"
-description: "Learn how to set up AWS Bedrock with Cline using credentials authentication. This guide covers AWS environment setup, regional access verification, and secure integration with the Cline VS Code extension."
+title: "API Key (Simple Setup)"
+sidebarTitle: "API Key"
+description: "Set up AWS Bedrock with Cline using Bedrock API Keys. Simplest setup for individual developers to access frontier models."
 ---
 
 ### Overview
@@ -121,14 +122,14 @@ You can create a custom IAM policy with these permissions and attach it to your
 
 ### Conclusion
 
-By following these steps, your enterprise team can securely integrate AWS Bedrock with the Cline VS Code extension to accelerate development:
+By following these steps, you can quickly integrate AWS Bedrock with the Cline VS Code extension to accelerate development:
 
-1. **Prepare Your AWS Environment:** Create or use a secure IAM role/user, attach the `AmazonBedrockLimitedAccess` policy, and ensure necessary permissions.
+1. **Prepare Your AWS Environment:** Create a Bedrock API Key with the necessary permissions.
 2. **Verify Region and Model Access:** Confirm that your selected region supports your required models.
-3. **Configure Cline in VS Code:** Install and set up Cline with your AWS credentials and choose an appropriate model.
+3. **Configure Cline in VS Code:** Install and set up Cline with your AWS API Key and choose an appropriate model.
 4. **Implement Security and Monitoring:** Use best practices for IAM, network security, monitoring, and cost management.
 
-For further details, consult the [AWS Bedrock Documentation](https://docs.aws.amazon.com/bedrock/latest/userguide/what-is-bedrock.html) and coordinate with your internal cloud team. Happy coding!
+For further details, consult the [AWS Bedrock Documentation](https://docs.aws.amazon.com/bedrock/latest/userguide/what-is-bedrock.html). Happy coding!
 
 ---
 

docs/provider-config/aws-bedrock/cli-profile.mdx

@@ -1,6 +1,7 @@
 ---
-title: "AWS Bedrock w/ Profile Authentication"
-description: "Learn how to configure AWS Bedrock to use AWS Profiles for authentication with Cline, focusing on SSO/Federated roles for secure access."
+title: "CLI Profile (SSO)"
+sidebarTitle: "CLI Profile (SSO)"
+description: "Configure AWS Bedrock to use AWS CLI profiles for authentication with Cline. Best for SSO/federated roles and secure enterprise access."
 ---
 
 ### Overview

docs/provider-config/aws-bedrock/iam-credentials.mdx

@@ -1,6 +1,7 @@
 ---
-title: "AWS Bedrock"
-description: "Learn how to set up AWS Bedrock with Cline using credentials authentication. This guide covers AWS environment setup, regional access verification, and secure integration with the Cline VS Code extension."
+title: "IAM Credentials"
+sidebarTitle: "IAM Credentials"
+description: "Set up AWS Bedrock with Cline using IAM Access Key and Secret Key credentials. Best for enterprise environments with established IAM policies."
 ---
 
 ### Overview

docs/provider-config/cerebras.mdx

@@ -0,0 +1,96 @@
+---
+title: "Cerebras"
+description: "Learn how to configure and use Cerebras's ultra-fast inference with Cline. Experience up to 2,600 tokens per second with wafer-scale chip architecture and real-time reasoning models."
+---
+
+Cerebras delivers the world's fastest AI inference through their revolutionary wafer-scale chip architecture. Unlike traditional GPUs that shuttle model weights from external memory, Cerebras stores entire models on-chip, eliminating bandwidth bottlenecks and achieving speeds up to 2,600 tokens per second—often 20x faster than GPUs.
+
+**Website:** [https://cloud.cerebras.ai/](https://cloud.cerebras.ai/)
+
+### Getting an API Key
+
+1.  **Sign Up/Sign In:** Go to [Cerebras Cloud](https://cloud.cerebras.ai/) and create an account or sign in.
+2.  **Navigate to API Keys:** Access the API keys section in your dashboard.
+3.  **Create a Key:** Generate a new API key. Give it a descriptive name (e.g., "Cline").
+4.  **Copy the Key:** Copy the API key immediately. Store it securely.
+
+### Supported Models
+
+Cline supports the following Cerebras models:
+
+-   `qwen-3-coder-480b-free` (Free tier) - High-performance coding model at no cost
+-   `qwen-3-coder-480b` - Flagship 480B parameter coding model
+-   `qwen-3-235b-a22b-instruct-2507` - Advanced instruction-following model
+-   `qwen-3-235b-a22b-thinking-2507` - Reasoning model with step-by-step thinking
+-   `llama-3.3-70b` - Meta's Llama 3.3 model optimized for speed
+-   `qwen-3-32b` - Compact yet powerful model for general tasks
+
+### Configuration in Cline
+
+1.  **Open Cline Settings:** Click the settings icon (⚙️) in the Cline panel.
+2.  **Select Provider:** Choose "Cerebras" from the "API Provider" dropdown.
+3.  **Enter API Key:** Paste your Cerebras API key into the "Cerebras API Key" field.
+4.  **Select Model:** Choose your desired model from the "Model" dropdown.
+5.  **(Optional) Custom Base URL:** Most users won't need to adjust this setting.
+
+### Cerebras's Wafer-Scale Advantage
+
+Cerebras has fundamentally reimagined AI hardware architecture to solve the inference speed problem:
+
+#### Wafer-Scale Architecture
+Traditional GPUs use separate chips for compute and memory, forcing them to constantly shuttle model weights back and forth. Cerebras built the world's largest AI chip—a wafer-scale engine that stores entire models on-chip. No external memory, no bandwidth bottlenecks, no waiting.
+
+#### Revolutionary Speed
+- **Up to 2,600 tokens per second** - often 20x faster than GPUs
+- **Single-second reasoning** - what used to take minutes now happens instantly
+- **Real-time applications** - reasoning models become practical for interactive use
+- **No bandwidth limits** - entire models stored on-chip eliminate memory bottlenecks
+
+#### The Cerebras Scaling Law
+Cerebras discovered that **faster inference enables smarter AI**. Modern reasoning models generate thousands of tokens as "internal monologue" before answering. On traditional hardware, this takes too long for real-time use. Cerebras makes reasoning models fast enough for everyday applications.
+
+#### Quality Without Compromise
+Unlike other speed optimizations that sacrifice accuracy, Cerebras maintains full model quality while delivering unprecedented speed. You get the intelligence of frontier models with the responsiveness of lightweight ones.
+
+Learn more about Cerebras's technology in their blog posts:
+- [The Cerebras Scaling Law: Faster Inference Is Smarter AI](https://www.cerebras.ai/blog/the-cerebras-scaling-law-faster-inference-is-smarter-ai)
+- [Introducing Cerebras Code](https://www.cerebras.ai/blog/introducing-cerebras-code)
+
+### Cerebras Code Plans
+
+Cerebras offers specialized plans for developers:
+
+#### Code Pro ($50/month)
+- Access to Qwen3-Coder with fast, high-context completions
+- Up to 24 million tokens per day
+- Ideal for indie developers and weekend projects
+- 3-4 hours of uninterrupted coding per day
+
+#### Code Max ($200/month)
+- Heavy coding workflow support
+- Up to 120 million tokens per day
+- Perfect for full-time development and multi-agent systems
+- No weekly limits, no IDE lock-in
+
+### Special Features
+
+#### Free Tier
+The `qwen-3-coder-480b-free` model provides access to high-performance inference at no cost—unique among speed-focused providers.
+
+#### Real-Time Reasoning
+Reasoning models like `qwen-3-235b-a22b-thinking-2507` can complete complex multi-step reasoning in under a second, making them practical for interactive development workflows.
+
+#### Coding Specialization
+Qwen3-Coder models are specifically optimized for programming tasks, delivering performance comparable to Claude Sonnet 4 and GPT-4.1 in coding benchmarks.
+
+#### No IDE Lock-In
+Works with any OpenAI-compatible tool—Cursor, Continue.dev, Cline, or any other editor that supports OpenAI endpoints.
+
+### Tips and Notes
+
+-   **Speed Advantage:** Cerebras excels at making reasoning models practical for real-time use. Perfect for agentic workflows that require multiple LLM calls.
+-   **Free Tier:** Start with the free model to experience Cerebras speed before upgrading to paid plans.
+-   **Context Windows:** Models support context windows ranging from 64K to 128K tokens for including substantial code context.
+-   **Rate Limits:** Generous rate limits designed for development workflows. Check your dashboard for current limits.
+-   **Pricing:** Competitive pricing with significant speed advantages. Visit [Cerebras Cloud](https://cloud.cerebras.ai/) for current rates.
+-   **Real-Time Applications:** Ideal for applications where AI response time matters—code generation, debugging, and interactive development.

docs/provider-config/claude-code.mdx

@@ -52,6 +52,7 @@ If you're not sure where Claude Code is installed:
 The Claude Code provider supports these models:
 
 -   `claude-sonnet-4-20250514` (Recommended)
+-   `claude-opus-4-1-20250805`
 -   `claude-opus-4-20250514`
 -   `claude-3-7-sonnet-20250219`
 -   `claude-3-5-sonnet-20241022`

docs/provider-config/doubao.mdx

@@ -0,0 +1,87 @@
+---
+title: "Doubao"
+description: "Learn how to configure and use ByteDance's Doubao AI models with Cline. Experience advanced reasoning, multimodal capabilities, and cost-effective inference with Chinese language optimization."
+---
+
+Doubao is ByteDance's flagship AI model series, featuring innovative sparse Mixture-of-Experts (MoE) architecture that delivers performance equivalent to much larger models while maintaining cost efficiency. With over 13 million users and advanced multimodal capabilities, Doubao offers competitive alternatives to Western AI systems with particular strength in Chinese language processing.
+
+**Website:** [https://www.volcengine.com/](https://www.volcengine.com/)
+
+### Getting an API Key
+
+1.  **Sign Up/Sign In:** Visit the [Volcano Engine Console](https://console.volcengine.com/). Create an account or sign in.
+2.  **Navigate to Model Service:** Access the AI model service section in the console.
+3.  **Create API Key:** Generate a new API key for the Doubao service.
+4.  **Copy the Key:** Copy the API key immediately and store it securely. You may not be able to view it again.
+
+### Supported Models
+
+Cline supports the following Doubao models:
+
+-   `doubao-seed-1-6-250615` (Default) - General purpose model with balanced performance
+-   `doubao-seed-1-6-thinking-250715` - Enhanced reasoning model with step-by-step thinking
+-   `doubao-seed-1-6-flash-250715` - Speed-optimized model for fast inference
+
+All models feature:
+-   **128,000 token context window** for extensive document processing
+-   **32,768 max output tokens** for comprehensive responses
+-   **Image input support** for multimodal applications
+-   **Prompt caching** with 80% discount on cached reads
+
+### Configuration in Cline
+
+1.  **Open Cline Settings:** Click the settings icon (⚙️) in the Cline panel.
+2.  **Select Provider:** Choose "Doubao" from the "API Provider" dropdown.
+3.  **Enter API Key:** Paste your Doubao API key into the "Doubao API Key" field.
+4.  **Select Model:** Choose your desired model from the "Model" dropdown.
+
+**Note:** Doubao uses the base URL `https://ark.cn-beijing.volces.com/api/v3` and servers are located in Beijing, China.
+
+### ByteDance's AI Innovation
+
+Doubao represents ByteDance's strategic entry into the AI model space with several key innovations:
+
+#### Sparse Mixture-of-Experts Architecture
+Doubao 1.5 Pro employs an innovative sparse MoE framework where 20 billion activated parameters deliver performance equivalent to a 140-billion-parameter dense model. This architecture significantly reduces operational costs while maintaining high performance standards.
+
+#### Extended Context Processing
+With context windows ranging from 32,000 to 256,000 tokens, Doubao excels at processing long-form content including legal documents, academic research, market reports, and creative content generation.
+
+#### Multimodal Excellence
+- **Advanced Visual Processing:** Enhanced visual reasoning, document recognition, and fine-grained information understanding
+- **Integrated Speech:** Seamless speech and text token integration with superior emotional continuity
+- **Document Analysis:** Comprehensive document summarization and content processing capabilities
+
+#### Chinese Language Optimization
+Doubao was specifically trained for Chinese language fluency and cultural relevance, providing significant advantages for Chinese-speaking users and applications requiring deep cultural context understanding.
+
+#### Cost Efficiency
+Doubao maintains pricing approximately **half the cost of comparable OpenAI offerings**, making advanced AI more accessible while establishing competitive market positioning.
+
+### Special Features
+
+#### Reasoning Models
+The `doubao-seed-1-6-thinking-250715` model offers enhanced reasoning capabilities with step-by-step thinking processes, making it ideal for complex problem-solving tasks.
+
+#### Multimodal Capabilities
+Unlike traditional cascaded approaches, Doubao integrates speech and text processing seamlessly, enabling more natural voice interactions and comprehensive document analysis.
+
+#### Prompt Caching
+All models support prompt caching with significant cost savings (80% discount on cached reads), making repeated queries more economical.
+
+#### ByteDance Ecosystem Integration
+Doubao integrates vertically with ByteDance properties including TikTok (Douyin), Toutiao, and Feishu, enabling seamless workflow integration across the ecosystem.
+
+### Performance and Benchmarks
+
+Doubao-1.5 Pro-AS1 Preview has demonstrated superior performance compared to OpenAI's O1-preview on specific benchmarks, including surpassing O1 models on AIME tests. The model continues to improve through reinforcement learning, with performance expected to enhance over time.
+
+### Tips and Notes
+
+-   **Regional Advantage:** Optimized for Chinese language and cultural contexts, making it ideal for Chinese-speaking users and markets.
+-   **Cost Effectiveness:** Approximately 50% lower cost than comparable Western AI models while maintaining competitive performance.
+-   **Context Windows:** Large context windows (up to 256K tokens) enable processing of extensive documents and codebases.
+-   **Multimodal Applications:** Strong visual and speech processing capabilities make it suitable for diverse multimedia applications.
+-   **Server Location:** Servers located in Beijing, China - consider latency implications for global users.
+-   **Ecosystem Benefits:** Integration with ByteDance services provides additional workflow advantages for users of TikTok, Toutiao, and Feishu.
+-   **Pricing:** Check the Volcano Engine console for current pricing information and regional availability.

docs/provider-config/fireworks-ai.mdx

@@ -0,0 +1,51 @@
+---
+title: "Fireworks AI"
+description: "Learn how to configure and use Fireworks AI models with Cline. Access high-performance open-source language models with fast, cost-effective APIs."
+---
+
+Cline supports accessing models through the Fireworks AI platform, which offers fast, cost-effective access to a wide range of state-of-the-art open-source language models. Built for speed and reliability, Fireworks AI provides serverless deployment options with OpenAI-compatible APIs and context windows up to 256,000 tokens.
+
+**Website:** [https://fireworks.ai/](https://fireworks.ai/)
+
+### Getting an API Key
+
+1. **Sign Up/Sign In:** Go to [Fireworks AI](https://fireworks.ai/) and create an account or sign in.
+2. **Navigate to API Keys:** After logging in, go to the [API Keys page](https://app.fireworks.ai/settings/users/api-keys) in the account settings.
+3. **Create a Key:** Click "Create API key" and give your key a descriptive name (e.g., "Cline").
+4. **Copy the Key:** **Important:** Copy the API key _immediately_. You will not be able to see it again. Store it securely.
+
+### Supported Models
+
+Cline supports the following Fireworks AI models:
+
+-   `accounts/fireworks/models/kimi-k2-instruct` (Default)
+-   `accounts/fireworks/models/qwen3-235b-a22b-instruct-2507`
+-   `accounts/fireworks/models/qwen3-coder-480b-a35b-instruct`
+-   `accounts/fireworks/models/deepseek-r1-0528`
+-   `accounts/fireworks/models/deepseek-v3`
+
+**Model Details:**
+
+| Model | Context Window | Best For | Pricing (per 1M tokens) |
+|-------|----------------|----------|-------------------------|
+| Kimi K2 | 128K | General tasks, agentic capabilities | \$0.60 input, \$2.50 output |
+| Qwen3 235B | 256K | Cost-effective general use | \$0.22 input, \$0.88 output |
+| Qwen3 Coder | 256K | Code generation and debugging | \$0.45 input, \$1.80 output |
+| DeepSeek R1 | 160K | Complex reasoning, function calling | \$3.00 input, \$8.00 output |
+| DeepSeek V3 | 128K | Strong general performance | \$0.90 input, \$0.90 output |
+
+### Configuration in Cline
+
+1. **Open Cline Settings:** Click the settings icon (⚙️) in the Cline panel.
+2. **Select Provider:** Choose "Fireworks AI" from the "API Provider" dropdown.
+3. **Enter API Key:** Paste your Fireworks AI API key into the "Fireworks AI API Key" field.
+4. **Select Model:** Choose your desired model from the "Model" dropdown. The default model is Kimi K2.
+
+### Tips and Notes
+
+-   **Cost-Effective:** Fireworks AI offers significantly lower pricing than proprietary models while maintaining competitive performance.
+-   **Large Context Windows:** Most models support 128K-256K tokens, suitable for processing large documents and maintaining extended conversations.
+-   **OpenAI Compatibility:** The provider uses an OpenAI-compatible API format with streaming support and usage tracking.
+-   **Rate Limits:** Fireworks AI has usage-based rate limits. Monitor your usage in the dashboard and consider upgrading your plan if needed.
+-   **API Keys:** Stored locally on your machine for security.
+-   **Pricing:** See the [Fireworks AI pricing page](https://fireworks.ai/pricing) for current rates. Prices shown are per million tokens.

docs/provider-config/fireworks.mdx

@@ -0,0 +1,131 @@
+---
+title: "Fireworks AI"
+description: "Learn how to configure and use Fireworks AI's lightning-fast inference platform with Cline. Experience up to 4x faster inference speeds with optimized models and competitive pricing."
+---
+
+Fireworks AI is a leading infrastructure platform for generative AI that focuses on delivering exceptional performance through optimized inference capabilities. With up to 4x faster inference speeds than alternative platforms and support for over 40 different AI models, Fireworks eliminates the operational complexity of running AI models at scale.
+
+**Website:** [https://fireworks.ai/](https://fireworks.ai/)
+
+### Getting an API Key
+
+1.  **Sign Up/Sign In:** Go to [Fireworks AI](https://fireworks.ai/) and create an account or sign in.
+2.  **Navigate to API Keys:** Access the API keys section in your dashboard.
+3.  **Create a Key:** Generate a new API key. Give it a descriptive name (e.g., "Cline").
+4.  **Copy the Key:** Copy the API key immediately. Store it securely.
+
+### Supported Models
+
+Fireworks AI supports a wide variety of models across different categories. Popular models include:
+
+**Text Generation Models:**
+-   Llama 3.1 series (8B, 70B, 405B)
+-   Mixtral 8x7B and 8x22B
+-   Qwen 2.5 series
+-   DeepSeek models with reasoning capabilities
+-   Code Llama models for programming tasks
+
+**Vision Models:**
+-   Llama 3.2 Vision models
+-   Qwen 2-VL models
+
+**Embedding Models:**
+-   Various text embedding models for semantic search
+
+The platform curates, optimizes, and deploys models with custom kernels and inference optimizations for maximum performance.
+
+### Configuration in Cline
+
+1.  **Open Cline Settings:** Click the settings icon (⚙️) in the Cline panel.
+2.  **Select Provider:** Choose "Fireworks" from the "API Provider" dropdown.
+3.  **Enter API Key:** Paste your Fireworks API key into the "Fireworks API Key" field.
+4.  **Enter Model ID:** Specify the model you want to use (e.g., "accounts/fireworks/models/llama-v3p1-70b-instruct").
+5.  **Configure Tokens:** Optionally set max completion tokens and context window size.
+
+### Fireworks AI's Performance Focus
+
+Fireworks AI's competitive advantages center on performance optimization and developer experience:
+
+#### Lightning-Fast Inference
+- **Up to 4x faster inference** than alternative platforms
+- **250% higher throughput** compared to open source inference engines
+- **50% faster speed** with significantly reduced latency
+- **6x lower cost** than HuggingFace Endpoints with 2.5x generation speed
+
+#### Advanced Optimization Technology
+- **Custom kernels** and inference optimizations increase throughput per GPU
+- **Multi-LoRA architecture** enables efficient resource sharing
+- **Hundreds of fine-tuned model variants** can run on shared base model infrastructure
+- **Asset-light model** focuses on optimization software rather than expensive GPU ownership
+
+#### Comprehensive Model Support
+- **40+ different AI models** curated and optimized for performance
+- **Multiple GPU types** supported: A100, H100, H200, B200, AMD MI300X
+- **Pay-per-GPU-second billing** with no extra charges for start-up times
+- **OpenAI API compatibility** for seamless integration
+
+### Pricing Structure
+
+Fireworks AI uses a usage-based pricing model with competitive rates:
+
+#### Text and Vision Models (2025)
+| Parameter Count | Price per 1M Input Tokens |
+|---|---|
+| Less than 4B parameters | $0.10 |
+| 4B - 16B parameters | $0.20 |
+| More than 16B parameters | $0.90 |
+| MoE 0B - 56B parameters | $0.50 |
+
+#### Fine-Tuning Services
+| Base Model Size | Price per 1M Training Tokens |
+|---|---|
+| Up to 16B parameters | $0.50 |
+| 16.1B - 80B parameters | $3.00 |
+| DeepSeek R1 / V3 | $10.00 |
+
+#### Dedicated Deployments
+| GPU Type | Price per Hour |
+|---|---|
+| A100 80GB | $2.90 |
+| H100 80GB | $5.80 |
+| H200 141GB | $6.99 |
+| B200 180GB | $11.99 |
+| AMD MI300X | $4.99 |
+
+### Special Features
+
+#### Fine-Tuning Capabilities
+Fireworks offers sophisticated fine-tuning services accessible through CLI interface, supporting JSON-formatted data from databases like MongoDB Atlas. Fine-tuned models cost the same as base models for inference.
+
+#### Developer Experience
+- **Browser playground** for direct model interaction
+- **REST API** with OpenAI compatibility
+- **Comprehensive cookbook** with ready-to-use recipes
+- **Multiple deployment options** from serverless to dedicated GPUs
+
+#### Enterprise Features
+- **HIPAA and SOC 2 Type II compliance** for regulated industries
+- **Self-serve onboarding** for developers
+- **Enterprise sales** for larger deployments
+- **Post-paid billing options** and Business tier
+
+#### Reasoning Model Support
+Advanced support for reasoning models with `<think>` tag processing and reasoning content extraction, making complex multi-step reasoning practical for real-time applications.
+
+### Performance Advantages
+
+Fireworks AI's optimization delivers measurable improvements:
+- **250% higher throughput** vs open source engines
+- **50% faster speed** with reduced latency
+- **6x cost reduction** compared to alternatives
+- **2.5x generation speed** improvement per request
+
+### Tips and Notes
+
+-   **Model Selection:** Choose models based on your specific use case - smaller models for speed, larger models for complex reasoning.
+-   **Performance Focus:** Fireworks excels at making AI inference fast and cost-effective through advanced optimizations.
+-   **Fine-Tuning:** Leverage fine-tuning capabilities to improve model accuracy with your proprietary data.
+-   **Compliance:** HIPAA and SOC 2 Type II compliance enables use in regulated industries.
+-   **Pricing Model:** Usage-based pricing scales with your success rather than traditional seat-based models.
+-   **Developer Resources:** Extensive documentation and cookbook recipes accelerate implementation.
+-   **GPU Options:** Multiple GPU types available for dedicated deployments based on performance needs.

docs/provider-config/groq.mdx

@@ -0,0 +1,80 @@
+---
+title: "Groq"
+description: "Learn how to configure and use Groq's lightning-fast inference with Cline. Access models from OpenAI, Meta, DeepSeek, and more on Groq's purpose-built LPU architecture."
+---
+
+Groq provides ultra-fast AI inference through their custom LPU™ (Language Processing Unit) architecture, purpose-built for inference rather than adapted from training hardware. Groq hosts open-source models from various providers including OpenAI, Meta, DeepSeek, Moonshot AI, and others.
+
+**Website:** [https://groq.com/](https://groq.com/)
+
+### Getting an API Key
+
+1.  **Sign Up/Sign In:** Go to [Groq](https://groq.com/) and create an account or sign in.
+2.  **Navigate to Console:** Go to the [Groq Console](https://console.groq.com/) to access your dashboard.
+3.  **Create a Key:** Navigate to the API Keys section and create a new API key. Give your key a descriptive name (e.g., "Cline").
+4.  **Copy the Key:** Copy the API key immediately. You will not be able to see it again. Store it securely.
+
+### Supported Models
+
+Cline supports the following Groq models:
+
+-   `llama-3.3-70b-versatile` (Meta) - Balanced performance with 131K context
+-   `llama-3.1-8b-instant` (Meta) - Fast inference with 131K context  
+-   `openai/gpt-oss-120b` (OpenAI) - Featured flagship model with 131K context
+-   `openai/gpt-oss-20b` (OpenAI) - Featured compact model with 131K context
+-   `moonshotai/kimi-k2-instruct` (Moonshot AI) - 1 trillion parameter model with prompt caching
+-   `deepseek-r1-distill-llama-70b` (DeepSeek/Meta) - Reasoning-optimized model
+-   `qwen/qwen3-32b` (Alibaba Cloud) - Enhanced for Q&A tasks
+-   `meta-llama/llama-4-maverick-17b-128e-instruct` (Meta) - Latest Llama 4 variant
+-   `meta-llama/llama-4-scout-17b-16e-instruct` (Meta) - Latest Llama 4 variant
+
+### Configuration in Cline
+
+1.  **Open Cline Settings:** Click the settings icon (⚙️) in the Cline panel.
+2.  **Select Provider:** Choose "Groq" from the "API Provider" dropdown.
+3.  **Enter API Key:** Paste your Groq API key into the "Groq API Key" field.
+4.  **Select Model:** Choose your desired model from the "Model" dropdown.
+
+### Groq's Speed Revolution
+
+Groq's LPU architecture delivers several key advantages over traditional GPU-based inference:
+
+#### LPU Architecture
+Unlike GPUs that are adapted from training workloads, Groq's LPU is purpose-built for inference. This eliminates architectural bottlenecks that create latency in traditional systems.
+
+#### Unmatched Speed
+- **Sub-millisecond latency** that stays consistent across traffic, regions, and workloads
+- **Static scheduling** with pre-computed execution graphs eliminates runtime coordination delays
+- **Tensor parallelism** optimized for low-latency single responses rather than high-throughput batching
+
+#### Quality Without Tradeoffs
+- **TruePoint numerics** reduce precision only in areas that don't affect accuracy
+- **100-bit intermediate accumulation** ensures lossless computation
+- **Strategic precision control** maintains quality while achieving 2-4× speedup over BF16
+
+#### Memory Architecture
+- **SRAM as primary storage** (not cache) with hundreds of megabytes on-chip
+- **Eliminates DRAM/HBM latency** that plagues traditional accelerators
+- **Enables true tensor parallelism** by splitting layers across multiple chips
+
+Learn more about Groq's technology in their [LPU architecture blog post](https://groq.com/blog/inside-the-lpu-deconstructing-groq-speed).
+
+### Special Features
+
+#### Prompt Caching
+The Kimi K2 model supports prompt caching, which can significantly reduce costs and latency for repeated prompts.
+
+#### Vision Support
+Select models support image inputs and vision capabilities. Check the model details in the Groq Console for specific capabilities.
+
+#### Reasoning Models
+Some models like DeepSeek variants offer enhanced reasoning capabilities with step-by-step thought processes.
+
+### Tips and Notes
+
+-   **Model Selection:** Choose models based on your specific use case and performance requirements.
+-   **Speed Advantage:** Groq excels at single-request latency rather than high-throughput batch processing.
+-   **OSS Model Provider:** Groq hosts open-source models from multiple providers (OpenAI, Meta, DeepSeek, etc.) on their fast infrastructure.
+-   **Context Windows:** Most models offer large context windows (up to 131K tokens) for including substantial code and context.
+-   **Pricing:** Groq offers competitive pricing with their speed advantages. Check the [Groq Pricing](https://groq.com/pricing) page for current rates.
+-   **Rate Limits:** Groq has generous rate limits, but check their documentation for current limits based on your usage tier.

docs/provider-config/openai-compatible.mdx

@@ -43,7 +43,6 @@ While the "OpenAI Compatible" provider type allows connecting to various endpoin
 -   `o1`
 -   `o1-preview`
 -   `o1-mini`
--   `gpt-4.5-preview`
 -   `gpt-4o`
 -   `gpt-4o-mini`
 

docs/provider-config/openai.mdx

@@ -26,7 +26,6 @@ Cline is compatible with a variety of OpenAI models, including but not limited t
 -   `o1`
 -   `o1-preview`
 -   `o1-mini`
--   `gpt-4.5-preview`
 -   `gpt-4o`
 -   `gpt-4o-mini`
 -   'gpt-4.1'

docs/provider-config/requesty.mdx

@@ -10,7 +10,7 @@ Cline supports accessing models through the [Requesty](https://www.requesty.ai/)
 ### Getting an API Key
 
 1.  **Sign Up/Sign In:** Go to the [Requesty website](https://www.requesty.ai/) and create an account or sign in.
-2.  **Get API Key:** You can get an API key from the [API Management](https://app.requesty.ai/manage-api) section of your Requesty dashboard.
+2.  **Get API Key:** You can get an API key from the [API Management](https://app.requesty.ai/api-keys) section of your Requesty dashboard.
 
 ### Supported Models
 
@@ -26,7 +26,7 @@ Requesty provides access to a wide range of models. Cline will automatically fet
 ### Tips and Notes
 
 -   **Optimizations**: Requesty offers a range of in-flight cost optimizations to lower your costs.
--   **Unified and simplified billing**: Unrestricted access to all providers and models, automatic balance top ups and more via a single [API key](https://app.requesty.ai/manage-api).
+-   **Unified and simplified billing**: Unrestricted access to all providers and models, automatic balance top ups and more via a single [API key](https://app.requesty.ai/api-keys).
 -   **Cost tracking**: Track cost per model, coding language, changed file, and more via the [Cost dashboard](https://app.requesty.ai/cost-management) or the [Requesty VS Code extension](https://marketplace.visualstudio.com/items?itemName=Requesty.requesty).
 -   **Stats and logs**: See your [coding stats dashboard](https://app.requesty.ai/usage-stats) or go through your [LLM interaction logs](https://app.requesty.ai/logs).
 -   **Fallback policies**: Keep your LLM working for you with fallback policies when providers are down.

docs/provider-config/sap-aicore.mdx

@@ -7,12 +7,13 @@ SAP AI Core and the generative AI hub help you to integrate LLMs and AI into new
 
 **Website:** [SAP Help Portal](https://help.sap.com/docs/sap-ai-core/sap-ai-core-service-guide/what-is-sap-ai-core)
 
-### Getting a Service Binding
 
 > 💡 **Information**
 >
 > SAP AI Core, and Generative AI Hub, are offerings from SAP BTP.
-> You need an active SAP BTP contract and a existing subaccount with a SAP AI Core instance to perform these steps.
+> You need an active SAP BTP contract and a existing subaccount with a SAP AI Core instance with the `extended` service plan (For more details about SAP AI Core service plans and their capabilities, see the [Service Plans documentation](https://help.sap.com/docs/sap-ai-core/sap-ai-core-service-guide/service-plans)) to perform these steps.
+
+### Getting a Service Binding
 
 1. **Access:** Go to your subaccount via [BTP Cloud Cockpit](cockpit.btp.cloud.sap/cockpit)
 2. **Create a Service Binding:** Go to "Instances and Subscriptions", select your SAP AI Core service instance and click on Service Bindings > Create.
@@ -32,8 +33,44 @@ Refer to the [Generative AI Hub Supported Models page](https://me.sap.com/notes/
 5.  **Enter Base URL:** Add the `.serviceurls.AI_API_URL` field from the service binding into the "AI Core Base URL" field.
 6.  **Enter Auth URL:** Add the `.url` field from the service binding into the "AI Core Auth URL" field.
 7.  **Enter Resource Group:** Add the resource group where you have your model deployments. See [Create a Deployment for a Generative AI Model](https://help.sap.com/docs/sap-ai-core/sap-ai-core-service-guide/create-deployment-for-generative-ai-model-in-sap-ai-core).
-8.  **Select Model:** Choose your desired model from the "Model" dropdown.
+8.  **Configure Orchestration Mode:** If you have an `extended` service plan, the "Orchestration Mode" checkbox will automatically appear.
+9.  **Select Model:** Choose your desired model from the "Model" dropdown.
+
+### Orchestration Mode vs Native API
+
+**Orchestration Mode:**
+- **Simplified usage:** Provides access to all available models without requiring individual deployments using the [Harmonized API](https://help.sap.com/docs/sap-ai-core/sap-ai-core-service-guide/harmonized-api)
+
+**Native API Mode:**
+- **Manual deployments:** Requires manual model deployment and management in your SAP AI Core service instance
 
 ### Tips and Notes
 
--   **Model Selection:** SAP AI Core offers a wide range of models. You won't be able to use the model, even if selected, if a deployment doesn't exist in the provided resource group.
+- **Service Plan Requirement:** You must have the SAP AI Core `extended` service plan to use LLMs with Cline. Other service plans do not provide access to Generative AI Hub.
+
+- **Orchestration Mode (Recommended):** Keep Orchestration Mode enabled for the simplest setup. It provides automatic access to all available models without requiring manual deployments.
+
+- **Native API Mode:** Only disable Orchestration Mode if you have specific requirements that necessitate direct AI Core API access or need features not supported by the orchestration mode.
+
+- **When using Native API Mode:**
+    -   **Model Selection:** The model dropdown displays models in two separate lists:
+        -   **Deployed Models:** These models are already deployed in your specified resource group and are ready to use immediately.
+        -   **Not Deployed Models:** These models don't have active deployments in your specified resource group. You won't be able to use these models until you create deployments for them in SAP AI Core.
+    -   **Creating Deployments:** To use a model that has not been deployed yet, you'll need to create a deployment in your SAP AI Core service instance. See [Create a Deployment for a Generative AI Model](https://help.sap.com/docs/sap-ai-core/sap-ai-core-service-guide/create-deployment-for-generative-ai-model-in-sap-ai-core) for instructions.
+
+#### Configuring Reasoning Effort for OpenAI Models
+
+When using OpenAI reasoning models (such as o1, o3, o3-mini, o4-mini) through SAP AI Core, you can control the reasoning effort to balance performance and cost:
+
+1. **Open Cline Settings:** Click the settings icon (⚙️) in the Cline panel.
+2. **Navigate to Features:** Go to the "Features" section in the settings.
+3. **Find OpenAI Reasoning Effort:** Locate the "OpenAI Reasoning Effort" setting.
+4. **Choose Effort Level:** Select between:
+   - **Low:** Faster responses with lower token usage, suitable for simpler tasks
+   - **Medium:** Balanced performance and token usage for most tasks
+   - **High:** More thorough analysis with higher token usage, better for complex reasoning tasks
+
+> 💡 **Note**
+>
+> This setting only applies when using OpenAI reasoning models (o1, o3, o3-mini, o4-mini, gpt-5, etc.) deployed through SAP AI Core. Other models will ignore this setting.
+

docs/provider-config/vercel-ai-gateway.mdx

@@ -0,0 +1,98 @@
+---
+title: "Vercel AI Gateway"
+description: "Use Vercel AI Gateway in Cline to reach 100+ models from one endpoint with routing, retries, and spend observability."
+---
+
+Vercel AI Gateway gives you a single API to access models from many providers. You switch by model id without swapping SDKs or juggling multiple keys. Cline integrates directly so you can pick a Gateway model in the dropdown, use it like any other provider, and see token and cache usage in the stream.
+
+Useful links:
+- Team dashboard: https://vercel.com/d?to=%2F%5Bteam%5D%2F%7E%2Fai
+- Models catalog: https://vercel.com/ai-gateway/models
+- Docs: https://vercel.com/docs/ai-gateway
+
+## What you get
+
+- One endpoint for 100+ models with a single key
+- Automatic retries and fallbacks that you configure on the dashboard
+- Spend monitoring with requests by model, token counts, cache usage, latency percentiles, and cost
+- OpenAI-compatible surface so existing clients work
+
+## Getting an API Key
+
+1. Sign in at https://vercel.com
+2. Dashboard → AI Gateway → API Keys → Create key
+3. Copy the key
+
+For more on authentication and OIDC options, see https://vercel.com/docs/ai-gateway/authentication
+
+## Configuration in Cline
+
+1. Open Cline settings
+2. Select **Vercel AI Gateway** as the API Provider
+3. Paste your Gateway API Key
+4. Pick a model from the list. Cline fetches the catalog automatically. You can also paste an exact id
+
+Notes:
+- Model ids often follow `provider/model`. Copy the exact id from the catalog  
+  Examples:
+  - `openai/gpt-5`
+  - `anthropic/claude-sonnet-4`
+  - `google/gemini-2.5-pro`
+  - `groq/llama-3.1-70b`
+  - `deepseek/deepseek-v3`
+
+## Observability you can act on
+
+<Frame>
+  <img src="https://assets.vercel.com/image/upload/v1753121283/gateway-overhead-dark_zhqwwj.svg" alt="Vercel AI Gateway observability with requests by model, tokens, cache, latency, and cost." />
+</Frame>
+
+What to watch:
+- Requests by model - confirm routing and adoption
+- Tokens - input vs output, including reasoning if exposed
+- Cache - cached input and cache creation tokens
+- Latency - p75 duration and p75 time to first token
+- Cost - per project and per model
+
+Use it to:
+- Compare output tokens per request before and after a model change
+- Validate cache strategy by tracking cache reads and write creation
+- Catch TTFT regressions during experiments
+- Align budgets with real usage
+
+## Supported models
+
+The gateway supports a large and changing set of models. Cline pulls the list from the Gateway API and caches it locally. For the current catalog, see https://vercel.com/ai-gateway/models
+
+## Tips
+
+<Tip>
+Use separate gateway keys per environment (dev, staging, prod). It keeps dashboards clean and budgets isolated.
+</Tip>
+
+<Note>
+Pricing is pass-through at provider list price. Bring-your-own key has 0% markup. You still pay provider and processing fees.
+</Note>
+
+<Info>
+Vercel does not add rate limits. Upstream providers may. New accounts receive $5 credits every 30 days until the first payment.
+</Info>
+
+## Troubleshooting
+
+- 401 - send the Gateway key to the Gateway endpoint, not an upstream URL
+- 404 model - copy the exact id from the Vercel catalog
+- Slow first token - check p75 TTFT in the dashboard and try a model optimized for streaming
+- Cost spikes - break down by model in the dashboard and cap or route traffic
+
+## Inspiration
+
+- Multi-model evals - swap only the model id in Cline and compare latency and output tokens
+- Progressive rollout - route a small percent to a new model in the dashboard and ramp with metrics
+- Budget enforcement - set per-project limits without code changes
+
+## Crosslinks
+
+- OpenAI-Compatible setup: /provider-config/openai-compatible
+- Model Selection Guide: /getting-started/model-selection-guide
+- Understanding Context Management: /getting-started/understanding-context-management

docs/provider-config/zai.mdx

@@ -0,0 +1,124 @@
+---
+title: "Z AI (Zhipu AI)"
+description: "Learn how to configure and use Z AI's GLM-4.5 models with Cline. Experience advanced hybrid reasoning, agentic capabilities, and open-source excellence with regional optimization."
+---
+
+Z AI (formerly Zhipu AI) offers the groundbreaking GLM-4.5 series, featuring hybrid reasoning capabilities and agentic AI design. Released in July 2025, these models excel in unified reasoning, coding, and intelligent agent applications while maintaining open-source accessibility under MIT license.
+
+**Website:** [https://z.ai/model-api](https://z.ai/model-api) (International) | [https://open.bigmodel.cn/](https://open.bigmodel.cn/) (China)
+
+### Getting an API Key
+
+#### International Users
+1.  **Sign Up/Sign In:** Go to [https://z.ai/model-api](https://z.ai/model-api). Create an account or sign in.
+2.  **Navigate to API Keys:** Access your account dashboard and find the API keys section.
+3.  **Create a Key:** Generate a new API key for your application.
+4.  **Copy the Key:** Copy the API key immediately and store it securely.
+
+#### China Mainland Users
+1.  **Sign Up/Sign In:** Go to [https://open.bigmodel.cn/](https://open.bigmodel.cn/). Create an account or sign in.
+2.  **Navigate to API Keys:** Access your account dashboard and find the API keys section.
+3.  **Create a Key:** Generate a new API key for your application.
+4.  **Copy the Key:** Copy the API key immediately and store it securely.
+
+### Supported Models
+
+Z AI provides different model catalogs based on your selected region:
+
+#### GLM-4.5 Series
+-   **GLM-4.5** - Flagship model with 355B total parameters, 32B active parameters
+-   **GLM-4.5-Air** - Compact model with 106B total parameters, 12B active parameters
+
+#### GLM-4.5 Hybrid Reasoning Models
+-   **GLM-4.5 (Thinking Mode)** - Advanced reasoning with step-by-step analysis
+-   **GLM-4.5-Air (Thinking Mode)** - Efficient reasoning for mainstream hardware
+
+All models feature:
+-   **128,000 token context window** for extensive document processing
+-   **Mixture of Experts (MoE) architecture** for optimal performance
+-   **Agent-native design** integrating reasoning, coding, and tool usage
+-   **Open-source availability** under MIT license
+
+### Configuration in Cline
+
+1.  **Open Cline Settings:** Click the settings icon (⚙️) in the Cline panel.
+2.  **Select Provider:** Choose "Z AI" from the "API Provider" dropdown.
+3.  **Select Region:** Choose your region:
+    -   "International" for global access
+    -   "China" for mainland China access
+4.  **Enter API Key:** Paste your Z AI API key into the "Z AI API Key" field.
+5.  **Select Model:** Choose your desired model from the "Model" dropdown.
+
+### Z AI's Hybrid Intelligence
+
+Z AI's GLM-4.5 series introduces revolutionary capabilities that set it apart from conventional language models:
+
+#### Hybrid Reasoning Architecture
+GLM-4.5 operates in two distinct modes:
+- **Thinking Mode:** Designed for complex reasoning tasks and tool usage, engaging in deeper analytical processes
+- **Non-Thinking Mode:** Provides immediate responses for straightforward queries, optimizing efficiency
+
+This dual-mode architecture represents an "agent-native" design philosophy that adapts processing intensity based on query complexity.
+
+#### Exceptional Performance
+GLM-4.5 achieves a comprehensive score of **63.2** across 12 benchmarks spanning agentic tasks, reasoning, and coding challenges, securing **3rd place** among all proprietary and open-source models. GLM-4.5-Air maintains competitive performance with a score of **59.8** while delivering superior efficiency.
+
+#### Mixture of Experts Excellence
+The sophisticated MoE architecture optimizes performance while maintaining computational efficiency:
+- **GLM-4.5:** 355B total parameters with 32B active parameters
+- **GLM-4.5-Air:** 106B total parameters with 12B active parameters
+
+#### Extended Context Capabilities
+The 128,000-token context window enables comprehensive understanding of lengthy documents and codebases, with real-world testing confirming effective processing of nearly 2,000-line codebases while maintaining remarkable performance.
+
+#### Open-Source Leadership
+Released under MIT license, GLM-4.5 provides researchers and developers with access to state-of-the-art capabilities without proprietary restrictions, including base models, hybrid reasoning versions, and optimized FP8 variants.
+
+### Regional Optimization
+
+#### API Endpoints
+- **International:** Uses `https://api.z.ai/api/paas/v4`
+- **China:** Uses `https://open.bigmodel.cn/api/paas/v4`
+
+#### Model Availability
+The region setting determines both API endpoint and available models, with automatic filtering to ensure compatibility with your selected region.
+
+### Special Features
+
+#### Agentic Capabilities
+GLM-4.5's unified architecture makes it particularly suitable for complex intelligent agent applications requiring integrated reasoning, coding, and tool utilization capabilities.
+
+#### Comprehensive Benchmarking
+Performance evaluation encompasses:
+- **3 agentic task benchmarks**
+- **7 reasoning benchmarks** 
+- **2 coding benchmarks**
+
+This comprehensive assessment demonstrates versatility across diverse AI applications.
+
+#### Developer Integration
+Models support integration through multiple frameworks:
+- **transformers**
+- **vLLM**
+- **SGLang**
+
+Complete with dedicated model code, tool parser, and reasoning parser implementations.
+
+### Performance Comparisons
+
+#### vs Claude 4 Sonnet
+GLM-4.5 shows competitive performance in agentic coding and reasoning tasks, though Claude Sonnet 4 maintains advantages in coding success rates and autonomous multi-feature application development.
+
+#### vs GPT-4.5
+GLM-4.5 ranks competitively in reasoning and agent benchmarks, with GPT-4.5 generally leading in raw task accuracy on professional benchmarks like MMLU and AIME.
+
+### Tips and Notes
+
+-   **Region Selection:** Choose the appropriate region for optimal performance and compliance with local regulations.
+-   **Model Selection:** GLM-4.5 for maximum performance, GLM-4.5-Air for efficiency and mainstream hardware compatibility.
+-   **Context Advantage:** Large 128K context window enables processing of substantial codebases and documents.
+-   **Open Source Benefits:** MIT license enables both commercial use and secondary development.
+-   **Agentic Applications:** Particularly strong for applications requiring reasoning, coding, and tool usage integration.
+-   **Hybrid Reasoning:** Use Thinking Mode for complex problems, Non-Thinking Mode for simple queries.
+-   **API Compatibility:** OpenAI-compatible API provides streaming responses and usage reporting.
+-   **Framework Support:** Multiple integration options available for different deployment scenarios.

esbuild.mjs

@@ -1,7 +1,7 @@
 import fs from "node:fs"
-import * as esbuild from "esbuild"
 import path from "node:path"
 import { fileURLToPath } from "node:url"
+import * as esbuild from "esbuild"
 
 const __filename = fileURLToPath(import.meta.url)
 const __dirname = path.dirname(__filename)
@@ -20,7 +20,6 @@ const aliasResolverPlugin = {
 	setup(build) {
 		const aliases = {
 			"@": path.resolve(__dirname, "src"),
-			"@api": path.resolve(__dirname, "src/api"),
 			"@core": path.resolve(__dirname, "src/core"),
 			"@integrations": path.resolve(__dirname, "src/integrations"),
 			"@services": path.resolve(__dirname, "src/services"),
@@ -170,7 +169,7 @@ const standaloneConfig = {
 const e2eBuildConfig = {
 	...baseConfig,
 	entryPoints: ["src/test/e2e/utils/build.ts"],
-	outfile: `${destDir}/e2e-build.js`,
+	outfile: `${destDir}/e2e-build.mjs`,
 	external: ["@vscode/test-electron", "execa"],
 	sourcemap: false,
 	plugins: [aliasResolverPlugin, esbuildProblemMatcherPlugin],

eslint-rules/__tests__/no-direct-vscode-api.test.ts

@@ -1,123 +0,0 @@
-const { RuleTester: DirectApiRuleTester } = require("eslint")
-const noDirectVscodeApiRule = require("../no-direct-vscode-api")
-
-const directApiRuleTester = new DirectApiRuleTester({
-	parser: require.resolve("@typescript-eslint/parser"),
-	parserOptions: {
-		ecmaVersion: 2020,
-		sourceType: "module",
-		ecmaFeatures: {
-			jsx: true,
-		},
-	},
-})
-
-directApiRuleTester.run("no-direct-vscode-api", noDirectVscodeApiRule, {
-	valid: [
-		// Should allow vscode.postMessage in grpc-client-base.ts
-		{
-			code: `vscode.postMessage({ type: "grpc_request", data: {} })`,
-			filename: "grpc-client-base.ts",
-		},
-		{
-			code: `vscode.postMessage({ type: "grpc_request_cancel" })`,
-			filename: "/path/to/grpc-client-base.ts",
-		},
-		// Should allow in exception directories
-		{
-			code: `vscode.workspace.workspaceFolders`,
-			filename: "/src/hosts/vscode/host-bridge.ts",
-		},
-		{
-			code: `vscode.workspace.fs.stat(uri)`,
-			filename: "/standalone/runtime-files/helpers.ts",
-		},
-		// Should allow other vscode API calls
-		{
-			code: `vscode.window.showInformationMessage("Hello")`,
-			filename: "test.ts",
-		},
-		// Should allow postMessage calls on other objects
-		{
-			code: `window.postMessage({ type: "test" }, "*")`,
-			filename: "test.ts",
-		},
-		// Should allow variables named vscode but not calling postMessage
-		{
-			code: `const vscode = { other: "method" }; vscode.other()`,
-			filename: "test.ts",
-		},
-	],
-	invalid: [
-		// Should disallow vscode.postMessage in regular files
-		{
-			code: `vscode.postMessage({ type: "test", data: {} })`,
-			filename: "test.ts",
-			errors: [
-				{
-					messageId: "useGrpcClient",
-				},
-			],
-		},
-		// Should disallow vscode.postMessage in components
-		{
-			code: `vscode.postMessage({ type: "apiConfiguration", apiConfiguration })`,
-			filename: "ApiOptions.tsx",
-			errors: [
-				{
-					messageId: "useGrpcClient",
-				},
-			],
-		},
-		// Should disallow vscode.postMessage in test files
-		{
-			code: `vscode.postMessage({ type: "newTask", text: message.text })`,
-			filename: "test.test.ts",
-			errors: [
-				{
-					messageId: "useGrpcClient",
-				},
-			],
-		},
-		// Should disallow property access for disallowed APIs
-		{
-			code: `const folders = vscode.workspace.workspaceFolders;`,
-			filename: "workspace.ts",
-			errors: [
-				{
-					messageId: "useHostBridge",
-				},
-			],
-		},
-		// Should disallow method calls for disallowed APIs
-		{
-			code: `const relativePath = vscode.workspace.asRelativePath(filePath);`,
-			filename: "path-utils.ts",
-			errors: [
-				{
-					messageId: "usePathUtils",
-				},
-			],
-		},
-		// Should disallow nested property access
-		{
-			code: `const stats = await vscode.workspace.fs.stat(uri);`,
-			filename: "file-utils.ts",
-			errors: [
-				{
-					messageId: "useFsUtils",
-				},
-			],
-		},
-		// Should disallow getting a workspace folder
-		{
-			code: `const folder = vscode.workspace.getWorkspaceFolder(uri);`,
-			filename: "path-helper.ts",
-			errors: [
-				{
-					messageId: "usePathUtils",
-				},
-			],
-		},
-	],
-})

eslint-rules/index.js

@@ -1,16 +0,0 @@
-// eslint-rules/index.js
-const noDirectVscodeApi = require("./no-direct-vscode-api")
-
-module.exports = {
-	rules: {
-		"no-direct-vscode-api": noDirectVscodeApi,
-	},
-	configs: {
-		recommended: {
-			plugins: ["local"],
-			rules: {
-				"local/no-direct-vscode-api": "warn",
-			},
-		},
-	},
-}

eslint-rules/no-direct-vscode-api.js

@@ -1,214 +0,0 @@
-const { ESLintUtils } = require("@typescript-eslint/utils")
-const path = require("path")
-
-const createRule = ESLintUtils.RuleCreator((name) => `https://cline.bot/eslint-rules/${name}`)
-
-// Configuration of disallowed VSCode APIs and their recommended alternatives
-const disallowedApis = {
-	"vscode.postMessage": {
-		messageId: "useGrpcClient",
-	},
-	"vscode.workspace.fs.stat": {
-		messageId: "useFsUtils",
-	},
-	"vscode.workspace.fs.writeFile": {
-		messageId: "useFsUtils",
-	},
-	"vscode.workspace.workspaceFolders": {
-		messageId: "useHostBridgeWorkspace",
-	},
-	"vscode.workspace.asRelativePath": {
-		messageId: "usePathUtils",
-	},
-	"vscode.workspace.getWorkspaceFolder": {
-		messageId: "usePathUtils",
-	},
-	"vscode.window.showTextDocument": {
-		messageId: "useHostBridge",
-	},
-	"vscode.workspace.applyEdit": {
-		messageId: "useHostBridge",
-	},
-	"vscode.window.onDidChangeActiveTextEditor": {
-		messageId: "useHostBridge",
-	},
-	"vscode.env.openExternal": {
-		messageId: "useUtils",
-	},
-	// "vscode.window.showWarningMessage": {
-	// 	messageId: "useHostBridgeShowMessage",
-	// },
-	"vscode.window.showOpenDialog": {
-		messageId: "useHostBridgeShowMessage",
-	},
-	"vscode.window.showErrorMessage": {
-		messageId: "useHostBridgeShowMessage",
-	},
-	// "vscode.window.showInformationMessage": {
-	// 	messageId: "useHostBridgeShowMessage",
-	// },
-}
-
-module.exports = createRule({
-	name: "no-direct-vscode-api",
-	meta: {
-		type: "problem",
-		docs: {
-			description:
-				"Disallow direct VSCode API usage in favor of Cline's abstraction layers, except in src/hosts/vscode and standalone/runtime-files directories",
-			recommended: "error",
-		},
-		messages: {
-			useGrpcClient:
-				"Use gRPC service clients instead of vscode.postMessage().\n" +
-				"Example: AccountServiceClient.methodName(RequestType.create({...})) instead of vscode.postMessage({type: '...'}).\n" +
-				"Found: {{code}}",
-			useFsUtils:
-				"Use utilities in @/utils/fs instead of vscode.workspace.fs\n" +
-				"Example: import { isDirectory } from '@/utils/fs' or use the file system methods from the host bridge provider.\n" +
-				"Found: {{code}}",
-			usePathUtils:
-				"Use path utilities from @/utils/path instead of VSCode workspace path methods.\n" +
-				"This provides consistent path handling across different environments.\n" +
-				"Found: {{code}}",
-			useHostBridgeWorkspace:
-				"Use HostProvider.workspace.getWorkspacePaths({}) instead of vscode.workspace.workspaceFolders.\n" +
-				"This provides a consistent abstraction across VSCode and standalone environments.\n" +
-				"Found: {{code}}",
-			useHostBridgeShowMessage:
-				"Use HostProvider.window.showMessage instead of the vscode.window.showMessage.\n" +
-				"This provides a consistent abstraction across VSCode and standalone environments.\n" +
-				"Found: {{code}}",
-			useHostBridge:
-				"Use the host bridge instead of calling vscode APIs directly.\n" +
-				"This provides a consistent abstraction across VSCode and standalone environments.\n" +
-				"Found: {{code}}",
-			useUtils:
-				"Use utilities in @/utils instead of calling vscode APIs directly.\n" +
-				"This provides a consistent abstraction across VSCode and standalone environments.\n" +
-				"Found: {{code}}",
-		},
-		schema: [],
-	},
-	defaultOptions: [],
-
-	create(context) {
-		// Pattern for checking memberExpressions like vscode.workspace.fs.stat
-		function checkMemberExpression(node) {
-			if (isExcluded(context.filename)) {
-				// Skip if this file is being excluded.
-				return
-			}
-
-			// For handling nested properties like vscode.workspace.fs.stat
-			function getFullPropertyPath(node) {
-				if (node.type !== "MemberExpression") {
-					return node.name || ""
-				}
-
-				const objectPart = getFullPropertyPath(node.object)
-				const propertyPart = node.property.name || ""
-
-				return objectPart ? `${objectPart}.${propertyPart}` : propertyPart
-			}
-
-			// Check if the expression matches one of our disallowed patterns
-			if (node.object && node.object.type === "Identifier" && node.object.name === "vscode") {
-				const fullPath = `vscode.${node.property.name}`
-				checkDisallowedApi(fullPath, node)
-			}
-			// Handle nested expressions like vscode.workspace.fs.stat
-			else if (node.object && node.object.type === "MemberExpression") {
-				const fullPath = getFullPropertyPath(node)
-
-				// Only proceed if it starts with vscode
-				if (fullPath.startsWith("vscode.")) {
-					checkDisallowedApi(fullPath, node)
-				}
-			}
-		}
-
-		// Check if an expression matches a disallowed API and report if it does
-		function checkDisallowedApi(expressionPath, node) {
-			// Check exact matches
-			if (disallowedApis[expressionPath]) {
-				reportViolation(expressionPath, node)
-				return
-			}
-
-			// Check prefix matches (for nested properties)
-			for (const disallowedApi in disallowedApis) {
-				// For direct property access like vscode.workspace.workspaceFolders
-				if (expressionPath === disallowedApi) {
-					reportViolation(disallowedApi, node)
-					return
-				}
-
-				// For method calls like vscode.workspace.asRelativePath(...)
-				if (expressionPath.startsWith(`${disallowedApi}.`) || expressionPath.startsWith(`${disallowedApi}(`)) {
-					reportViolation(disallowedApi, node)
-					return
-				}
-			}
-		}
-
-		// Report a violation with the appropriate message
-		function reportViolation(disallowedApi, node) {
-			const sourceCode = context.sourceCode
-			const config = disallowedApis[disallowedApi]
-
-			// For method calls, get the whole call expression
-			let reportNode = node
-			let parentNode = sourceCode.getAncestors(node).pop()
-			if (parentNode && parentNode.type === "CallExpression" && parentNode.callee === node) {
-				reportNode = parentNode
-			}
-
-			const callText = sourceCode.getText(reportNode).trim()
-
-			context.report({
-				node: reportNode,
-				messageId: config.messageId,
-				data: {
-					code: callText,
-				},
-			})
-		}
-
-		function isExcluded(filename) {
-			// Check if current file is in an exception directory or is grpc-client-base.ts
-			if (path.basename(filename) === "grpc-client-base.ts") {
-				return true
-			}
-			// Skip checking files in src/hosts/vscode or standalone/runtime-files
-			if (filename.includes("/src/hosts/vscode/")) {
-				return true
-			}
-			if (filename.includes("/standalone/runtime-files/")) {
-				return true
-			}
-			// Skip unit tests
-			if (filename.endsWith(".test.ts")) {
-				return true
-			}
-		}
-
-		return {
-			// Detect basic member expressions (e.g., vscode.postMessage)
-			MemberExpression(node) {
-				checkMemberExpression(node)
-			},
-
-			// Detect property access through destructuring
-			VariableDeclarator(node) {
-				// Skip if this file is in an exception directory or is grpc-client-base.ts
-				if (isExcluded(context.filename)) {
-					return
-				}
-
-				// Destructuring pattern checks removed as developers don't use the API this way
-				// They always use direct imports: import * as vscode from "vscode" and direct access: vscode.thing.foo
-			},
-		}
-	},
-})

eslint-rules/package.json

@@ -1,31 +0,0 @@
-{
-	"name": "eslint-plugin-eslint-rules",
-	"version": "1.0.0",
-	"description": "Custom ESLint rules for Cline",
-	"main": "index.js",
-	"scripts": {
-		"test": "mocha --no-config --require ts-node/register __tests__/**/*.test.ts"
-	},
-	"keywords": [
-		"eslint",
-		"eslintplugin"
-	],
-	"author": "Cline Bot Inc.",
-	"license": "Apache-2.0",
-	"dependencies": {
-		"@typescript-eslint/utils": "^8.33.0"
-	},
-	"devDependencies": {
-		"@types/eslint": "^8.0.0",
-		"@types/mocha": "^10.0.7",
-		"@types/node": "^20.0.0",
-		"@typescript-eslint/parser": "^7.14.1",
-		"eslint": "^8.57.0",
-		"mocha": "^10.0.0",
-		"ts-node": "^10.9.2",
-		"typescript": "^5.4.5"
-	},
-	"peerDependencies": {
-		"eslint": ">=8.0.0"
-	}
-}

eslint-rules/tsconfig.json

@@ -1,16 +0,0 @@
-{
-	"compilerOptions": {
-		"target": "es2020",
-		"module": "commonjs",
-		"moduleResolution": "node",
-		"esModuleInterop": true,
-		"strict": true,
-		"skipLibCheck": true,
-		"forceConsistentCasingInFileNames": true,
-		"outDir": "dist",
-		"resolveJsonModule": true,
-		"declaration": true
-	},
-	"include": ["**/*.ts", "**/*.js", "**/*.tsx", "__tests__/**/*"],
-	"exclude": ["node_modules", "dist"]
-}

evals/cli/src/commands/runDiffEval.ts

@@ -10,6 +10,7 @@ interface RunDiffEvalOptions {
 	parsingFunction: string
 	diffEditFunction: string
 	thinkingBudget: number
+	provider: string
 	parallel: boolean
 	verbose: boolean
 	testPath: string
@@ -39,6 +40,8 @@ export async function runDiffEvalHandler(options: RunDiffEvalOptions) {
 		options.parsingFunction,
 		"--diff-edit-function",
 		options.diffEditFunction,
+		"--provider",
+		options.provider,
 	]
 
 	// Conditionally add the optional arguments

evals/cli/src/index.ts

@@ -92,6 +92,7 @@ program
 	.option("--parsing-function <name>", "The parsing function to use", "parseAssistantMessageV2")
 	.option("--diff-edit-function <name>", "The diff editing function to use", "constructNewFileContentV2")
 	.option("--thinking-budget <tokens>", "Set the thinking tokens budget", "0")
+	.option("--provider <provider>", "API provider to use (openrouter, openai)", "openrouter")
 	.option("--parallel", "Run tests in parallel", false)
 	.option("--replay", "Run evaluation from a pre-recorded LLM output, skipping the API call", false)
 	.option("--replay-run-id <run_id>", "The ID of the run to replay from the database")

evals/diff-edits/ClineWrapper.ts

@@ -1,11 +1,9 @@
 import { OpenRouterHandler } from "../../src/api/providers/openrouter"
-import { ApiHandlerOptions } from "../../src/shared/api"
+import { OpenAiNativeHandler } from "../../src/api/providers/openai-native"
 import { Anthropic } from "@anthropic-ai/sdk"
 
 import {
-	parseAssistantMessageV1,
 	parseAssistantMessageV2,
-	parseAssistantMessageV3,
 	AssistantMessageContent,
 } from "./parsing/parse-assistant-message-06-06-25" // "../../src/core/assistant-message"
 import { constructNewFileContent as constructNewFileContent_06_06_25 } from "./diff-apply/diff-06-06-25"
@@ -17,9 +15,7 @@ type ParseAssistantMessageFn = (message: string) => AssistantMessageContent[]
 type ConstructNewFileContentFn = (diff: string, original: string, strict: boolean) => Promise<string | any>
 
 const parsingFunctions: Record<string, ParseAssistantMessageFn> = {
-	parseAssistantMessageV1: parseAssistantMessageV1,
 	parseAssistantMessageV2: parseAssistantMessageV2,
-	parseAssistantMessageV3: parseAssistantMessageV3,
 }
 
 const diffEditingFunctions: Record<string, ConstructNewFileContentFn> = {
@@ -54,7 +50,7 @@ interface StreamResult {
  * Process the stream and return full response with timing data
  */
 async function processStream(
-	handler: OpenRouterHandler,
+	handler: OpenRouterHandler | OpenAiNativeHandler,
 	systemPrompt: string,
 	messages: Anthropic.Messages.MessageParam[],
 ): Promise<StreamResult> {
@@ -190,19 +186,7 @@ export async function runSingleEvaluation(input: TestInput): Promise<TestResult>
 			}
 		}
 
-		const options: ApiHandlerOptions = {
-			openRouterApiKey: apiKey,
-			openRouterModelId: modelId,
-			thinkingBudgetTokens: thinkingBudgetTokens,
-			openRouterModelInfo: {
-				maxTokens: 10_000,
-				contextWindow: 1_000_000,
-				supportsImages: true,
-				supportsPromptCache: true, // may need to turn this on
-				inputPrice: 0,
-				outputPrice: 0,
-			},
-		}
+		const provider = input.provider || "openrouter"
 
 		// Get the output of streaming output of this llm call
 		let streamResult: StreamResult
@@ -214,10 +198,34 @@ export async function runSingleEvaluation(input: TestInput): Promise<TestResult>
 				usage: { inputTokens: 0, outputTokens: 0, cacheWriteTokens: 0, cacheReadTokens: 0, totalCost: 0 },
 			}
 		} else {
-			// Live mode: existing API call logic
+			// Live mode: provider-specific API call logic
 			try {
-				const openRouterHandler = new OpenRouterHandler(options)
-				streamResult = await processStream(openRouterHandler, systemPrompt, messages)
+				let handler: OpenRouterHandler | OpenAiNativeHandler
+				
+				if (provider === "openai") {
+					const openAiOptions = {
+						openAiNativeApiKey: apiKey,
+						apiModelId: modelId,
+					}
+					handler = new OpenAiNativeHandler(openAiOptions)
+				} else {
+					const openRouterOptions = {
+						openRouterApiKey: apiKey,
+						openRouterModelId: modelId,
+						thinkingBudgetTokens: thinkingBudgetTokens,
+						openRouterModelInfo: {
+							maxTokens: 10_000,
+							contextWindow: 1_000_000,
+							supportsImages: true,
+							supportsPromptCache: true,
+							inputPrice: 0,
+							outputPrice: 0,
+						},
+					}
+					handler = new OpenRouterHandler(openRouterOptions)
+				}
+				
+				streamResult = await processStream(handler, systemPrompt, messages)
 			} catch (error: any) {
 				return {
 					success: false,

evals/diff-edits/TestRunner.ts

@@ -49,16 +49,25 @@ type TestResultSet = { [test_id: string]: (TestResult & { test_id?: string })[]
 
 class NodeTestRunner {
 	private apiKey: string | undefined
+	private provider: string
 	private currentRunId: string | null = null
 	private systemPromptHash: string | null = null
 	private processingFunctionsHash: string | null = null
 	private caseIdMap: Map<string, string> = new Map() // test_id -> case_id mapping
 
-	constructor(isReplay: boolean) {
+	constructor(isReplay: boolean, provider: string = "openrouter") {
+		this.provider = provider
 		if (!isReplay) {
-			this.apiKey = process.env.OPENROUTER_API_KEY
-			if (!this.apiKey) {
-				throw new Error("OPENROUTER_API_KEY environment variable not set for a non-replay run.")
+			if (provider === "openai") {
+				this.apiKey = process.env.OPENAI_API_KEY
+				if (!this.apiKey) {
+					throw new Error("OPENAI_API_KEY environment variable not set for a non-replay run with OpenAI provider.")
+				}
+			} else {
+				this.apiKey = process.env.OPENROUTER_API_KEY
+				if (!this.apiKey) {
+					throw new Error("OPENROUTER_API_KEY environment variable not set for a non-replay run with OpenRouter provider.")
+				}
 			}
 		}
 	}
@@ -635,6 +644,7 @@ class NodeTestRunner {
 			thinkingBudgetTokens: testConfig.thinking_tokens_budget,
 			originalDiffEditToolCallMessage: testConfig.replay ? testCase.original_diff_edit_tool_call_message : undefined,
 			diffApplyFile: testConfig.diff_apply_file,
+			provider: this.provider,
 			isVerbose: isVerbose,
 		}
 
@@ -927,6 +937,7 @@ async function main() {
 		.option("--parsing-function <name>", "The parsing function to use", "parseAssistantMessageV2")
 		.option("--diff-edit-function <name>", "The diff editing function to use", "diff-06-26-25")
 		.option("--thinking-budget <tokens>", "Set the thinking tokens budget", "0")
+		.option("--provider <provider>", "API provider to use (openrouter, openai)", "openrouter")
 		.option("--parallel", "Run tests in parallel", false)
 		.option("--replay", "Run evaluation from a pre-recorded LLM output, skipping the API call", false)
 		.option("--replay-run-id <run_id>", "The ID of the run to replay from the database")
@@ -959,7 +970,7 @@ async function main() {
 		? parseInt(options.maxAttemptsPerCase, 10)
 		: validAttemptsPerCase * 10;
 
-	const runner = new NodeTestRunner(options.replay || !!options.replayRunId)
+	const runner = new NodeTestRunner(options.replay || !!options.replayRunId, options.provider)
 
 	if (options.replayRunId) {
 		if (!options.diffApplyFile) {
@@ -979,7 +990,7 @@ async function main() {
 			log(isVerbose, "Warning: Could not load OpenRouter model data. Context window filtering might be affected for OpenRouter models.");
 		}
 
-		const runner = new NodeTestRunner(options.replay)
+		const runner = new NodeTestRunner(options.replay, options.provider)
 		let allLoadedTestCases = runner.loadTestCases(testPath, isVerbose) // Pass isVerbose
 		
 		const allProcessedTestCasesGlobal: ProcessedTestCase[] = allLoadedTestCases.map((tc) => ({

evals/diff-edits/dashboard/app.py

@@ -331,6 +331,42 @@ def get_performance_grade(success_rate):
     else:
         return "C", "poor"
 
+def get_error_description(error_enum, error_string=None):
+    """Map error enum values to user-friendly descriptions"""
+    error_map = {
+        1: "No tool calls - Model didn't use the replace_in_file tool",
+        2: "Multiple tool calls - Model called multiple tools instead of one", 
+        3: "Wrong tool call - Model used wrong tool (not replace_in_file)",
+        4: "Missing parameters - Tool call missing required path or diff",
+        5: "Wrong file edited - Model edited different file than expected",
+        6: "Wrong tool call - Model used wrong tool type",
+        7: "Wrong file edited - Model targeted incorrect file path",
+        8: "API/Stream error - Problem with model API connection",
+        9: "Configuration error - Invalid evaluation parameters",
+        10: "Function error - Invalid parsing/diff functions",
+        11: "Other error - Unexpected failure"
+    }
+    
+    base_description = error_map.get(error_enum, f"Unknown error (code: {error_enum})")
+    
+    if error_string:
+        return f"{base_description}: {error_string}"
+    return base_description
+
+def get_error_guidance(error_enum):
+    """Provide specific guidance based on error type"""
+    guidance_map = {
+        1: "💡 The model provided a response but didn't use the replace_in_file tool. Check the raw output to see what the model actually said.",
+        2: "💡 The model called multiple tools when it should only call replace_in_file once. Check the parsed tool call section.",
+        3: "💡 The model used a different tool instead of replace_in_file. This might indicate confusion about the task.",
+        4: "💡 The model called replace_in_file but didn't provide the required 'path' or 'diff' parameters.",
+        5: "💡 The model tried to edit a different file than expected. Check the parsed tool call to see which file it targeted.",
+        6: "💡 The model used the wrong tool type. Check the raw output to see what tool it attempted to use.",
+        7: "💡 The model tried to edit a different file path than expected. This could indicate path confusion or hallucination.",
+    }
+    
+    return guidance_map.get(error_enum, "")
+
 def render_hero_section(current_run, model_performance):
     """Render the hero section with key metrics"""
     run_title = current_run['description'] if current_run['description'] else f"Run {current_run['run_id'][:8]}..."
@@ -570,12 +606,16 @@ def render_result_detail(result):
     """Render detailed view of a single result"""
     st.markdown("### 🔬 Result Deep Dive")
     
-    # Check if this is a valid result
-    is_valid = (result['error_enum'] not in [1, 6, 7]) if not pd.isna(result['error_enum']) else True
+    # Check if this is a valid result (only invalid if no tool calls or wrong file)
+    is_valid = True
+    if not pd.isna(result['error_enum']):
+        # Only these specific errors make a result "invalid" for the benchmark:
+        # 1 = no_tool_calls, 5 = wrong_file_edited, 7 = wrong_file_edited
+        is_valid = result['error_enum'] not in [1, 5, 7]
     
     # Show validity warning if needed
     if not is_valid:
-        st.warning("⚠️ **This is an invalid result** - The model didn't properly call the diff edit tool or edited the wrong file. This result is excluded from success rate calculations.")
+        st.warning("⚠️ **This is an invalid result** - The model didn't call the replace_in_file tool or edited the wrong file. This result is excluded from success rate calculations.")
     
     # Result metadata
     col1, col2, col3, col4 = st.columns(4)
@@ -591,7 +631,10 @@ def render_result_detail(result):
         st.markdown(f"**Round Trip:** {result['time_round_trip_ms']:.0f}ms")
     
     with col4:
-        st.markdown(f"**Cost:** ${result['cost_usd']:.4f}")
+        if pd.notna(result['cost_usd']) and result['cost_usd'] is not None:
+            st.markdown(f"**Cost:** ${result['cost_usd']:.4f}")
+        else:
+            st.markdown(f"**Cost:** Free")
     
     # Tabbed interface for different views
     tab1, tab2, tab3, tab4 = st.tabs(["📄 File & Edits", "🤖 Raw Output", "🔧 Parsed Tool Call", "📊 Metrics"])
@@ -693,8 +736,46 @@ def render_file_and_edits_view(result):
             # Show error information
             st.error("❌ **Edit Failed**")
             
+            # Show detailed error reason
             if not pd.isna(result['error_enum']):
-                st.markdown(f"**Error Code:** {result['error_enum']}")
+                error_description = get_error_description(
+                    result['error_enum'], 
+                    result.get('error_string')
+                )
+                st.markdown(f"**Reason:** {error_description}")
+                
+                # Show specific guidance based on error type
+                guidance = get_error_guidance(result['error_enum'])
+                if guidance:
+                    st.info(guidance)
+            
+            # For valid results that failed, check for diff application failures
+            elif not result['succeeded']:
+                # This is a valid result that failed - likely due to diff application issues
+                raw_output = result.get('raw_model_output', '')
+                
+                # Check if we have specific error information in the raw output
+                if 'does not match anything in the file' in str(raw_output).lower():
+                    st.warning("⚠️ **Diff Application Failed**")
+                    st.info("💡 The SEARCH block in the diff didn't match any content in the original file. This usually means the model hallucinated code that doesn't exist.")
+                elif 'malformatted' in str(raw_output).lower() or 'malformed' in str(raw_output).lower():
+                    st.warning("⚠️ **Diff Format Error**")
+                    st.info("💡 The diff format was incorrect. Check the raw tool call to see the formatting issues.")
+                elif 'error:' in str(raw_output).lower():
+                    # Try to extract the specific error message
+                    lines = str(raw_output).split('\n')
+                    error_lines = [line for line in lines if 'error:' in line.lower()]
+                    if error_lines:
+                        error_msg = error_lines[0].strip()
+                        st.warning("⚠️ **Diff Application Failed**")
+                        st.info(f"💡 {error_msg}")
+                    else:
+                        st.warning("⚠️ **Diff Application Failed**")
+                        st.info("💡 The diff couldn't be applied to the original file. Check the raw output and parsed tool call for more details.")
+                else:
+                    # Generic diff application failure
+                    st.warning("⚠️ **Diff Application Failed**")
+                    st.info("💡 The model made a valid tool call but the diff couldn't be applied to the original file. This usually indicates a mismatch between the expected and actual file content.")
         else:
             # Show successful edit information
             st.success("✅ **Edit Successful**")
@@ -725,8 +806,25 @@ def render_file_and_edits_view(result):
                     if len(edited_lines) > 50:
                         st.text(f"... ({len(edited_lines) - 50} more lines)")
         
-        # Show parsed tool call if available
+        # Show raw and parsed tool calls if available
         if not pd.isna(result['parsed_tool_call_json']):
+            with st.expander("View Raw Tool Call"):
+                # Extract the raw tool call text from the model output
+                raw_output = result['raw_model_output'] if not pd.isna(result['raw_model_output']) else ""
+                
+                # Try to extract just the tool call portion
+                if raw_output and '<replace_in_file>' in raw_output:
+                    # Find the tool call block
+                    start_idx = raw_output.find('<replace_in_file>')
+                    end_idx = raw_output.find('</replace_in_file>') + len('</replace_in_file>')
+                    if start_idx != -1 and end_idx != -1:
+                        raw_tool_call = raw_output[start_idx:end_idx]
+                        st.code(raw_tool_call, language='xml')
+                    else:
+                        st.text("Tool call not found in raw output")
+                else:
+                    st.text("No raw tool call available")
+            
             with st.expander("View Parsed Tool Call"):
                 try:
                     parsed_call = json.loads(result['parsed_tool_call_json'])
@@ -795,8 +893,10 @@ def render_metrics_view(result):
         if not pd.isna(result['completion_tokens']):
             st.metric("Completion Tokens", int(result['completion_tokens']))
         
-        if not pd.isna(result['cost_usd']):
+        if pd.notna(result['cost_usd']) and result['cost_usd'] is not None:
             st.metric("Cost", f"${result['cost_usd']:.4f}")
+        else:
+            st.metric("Cost", "Free")
         
         if not pd.isna(result['tokens_in_context']):
             st.metric("Context Tokens", int(result['tokens_in_context']))

evals/diff-edits/parsing/parse-assistant-message-06-06-25.ts

@@ -70,246 +70,7 @@ export interface ToolUse {
 	partial: boolean
 }
 
-/**
- * @description **Version 1**
- * Parses an assistant message string potentially containing mixed text and tool usage blocks
- * marked with XML-like tags into an array of structured content objects.
- *
- * This version iterates through the message character by character, building an accumulator string.
- * It maintains state to track whether it's currently parsing text, a tool use block, or a specific tool parameter.
- * It detects the start and end of tool uses and parameters by checking if the accumulator ends with
- * the corresponding opening or closing tags.
- * Special handling is included for `write_to_file` and `new_rule` tool uses to correctly parse
- * the `content` parameter, which might contain the closing tag itself, by looking for the *last*
- * occurrence of the closing tag.
- * If the input string ends mid-tag or mid-content, the last block (text or tool use) is marked as partial.
- *
- * @param assistantMessage The raw string output from the assistant.
- * @returns An array of `AssistantMessageContent` objects, which can be `TextContent` or `ToolUse`.
- *          Blocks that were not fully closed by the end of the input string will have their `partial` flag set to `true`.
- */
-export function parseAssistantMessageV1(assistantMessage: string): AssistantMessageContent[] {
-	const contentBlocks: AssistantMessageContent[] = []
-	let currentTextContent: TextContent | undefined = undefined
-	let currentTextContentStartIndex = 0
-	let currentToolUse: ToolUse | undefined = undefined
-	let currentToolUseStartIndex = 0
-	let currentParamName: ToolParamName | undefined = undefined
-	let currentParamValueStartIndex = 0
-	let accumulator = ""
-
-	for (let i = 0; i < assistantMessage.length; i++) {
-		const char = assistantMessage[i]
-		accumulator += char
-
-		// --- State: Parsing a Tool Parameter ---
-		// there should not be a param without a tool use
-		if (currentToolUse && currentParamName) {
-			const currentParamValue = accumulator.slice(currentParamValueStartIndex)
-			const paramClosingTag = `</${currentParamName}>`
-			if (currentParamValue.endsWith(paramClosingTag)) {
-				// End of param value found
-				currentToolUse.params[currentParamName] = currentParamValue.slice(0, -paramClosingTag.length).trim()
-				currentParamName = undefined // Go back to parsing tool content or looking for next param
-				continue // Move to next character
-			} else {
-				// Partial param value is accumulating
-				continue // Move to next character
-			}
-		}
-
-		// --- State: Parsing a Tool Use (but not a specific parameter) ---
-		// no currentParamName
-		if (currentToolUse) {
-			const currentToolValue = accumulator.slice(currentToolUseStartIndex)
-			const toolUseClosingTag = `</${currentToolUse.name}>`
-
-			if (currentToolValue.endsWith(toolUseClosingTag)) {
-				// End of a tool use found
-				currentToolUse.partial = false
-				contentBlocks.push(currentToolUse)
-				currentToolUse = undefined // Go back to parsing text or looking for next tool
-				// Reset text start index in case text follows immediately
-				currentTextContentStartIndex = i + 1
-				continue // Move to next character
-			} else {
-				// Check if starting a new parameter within the current tool use
-				const possibleParamOpeningTags = toolParamNames.map((name) => `<${name}>`)
-				let foundParamStart = false
-				for (const paramOpeningTag of possibleParamOpeningTags) {
-					if (accumulator.endsWith(paramOpeningTag)) {
-						// Start of a new parameter found
-						currentParamName = paramOpeningTag.slice(1, -1) as ToolParamName
-						currentParamValueStartIndex = accumulator.length
-						foundParamStart = true
-						break
-					}
-				}
-				if (foundParamStart) {
-					continue // Move to next character
-				}
-
-				// Special case for write_to_file/new_rule content param allowing nested tags
-				// Check if a </content> tag appears, potentially indicating the end of the content param
-				// even if the main tool closing tag hasn't been seen yet.
-				const contentParamName: ToolParamName = "content"
-				if (
-					(currentToolUse.name === "write_to_file" || currentToolUse.name === "new_rule") &&
-					accumulator.endsWith(`</${contentParamName}>`)
-				) {
-					const toolContent = accumulator.slice(currentToolUseStartIndex)
-					const contentStartTag = `<${contentParamName}>`
-					const contentEndTag = `</${contentParamName}>`
-					const contentStartIndex = toolContent.indexOf(contentStartTag) + contentStartTag.length
-					// Use lastIndexOf to handle cases where </content> might appear within the content itself
-					const contentEndIndex = toolContent.lastIndexOf(contentEndTag)
-
-					// Ensure we found valid start/end tags and end is after start
-					if (
-						contentStartIndex !== -1 &&
-						contentEndIndex !== -1 &&
-						contentEndIndex > contentStartIndex - contentStartTag.length // Ensure end tag is after start tag begins
-					) {
-						// Check if this content param was already being parsed. If so, update it.
-						// If not, and we just found the closing tag, assign it.
-						// This handles cases where the </content> detection might fire before
-						// the <content> tag detection logic, or if the content is very short.
-						if (currentParamName === contentParamName) {
-							// Already parsing content, now we found the end tag
-							currentToolUse.params[contentParamName] = toolContent.slice(contentStartIndex, contentEndIndex).trim()
-							currentParamName = undefined // Finished with this param
-						} else if (currentParamName === undefined) {
-							// Not parsing a param, but found </content>. Assume it closes the content block.
-							currentToolUse.params[contentParamName] = toolContent.slice(contentStartIndex, contentEndIndex).trim()
-							// We stay in the "parsing tool use" state, looking for more params or the tool end tag.
-						}
-					}
-				}
-
-				// If none of the above, partial tool value is accumulating
-				continue // Move to next character
-			}
-		}
-
-		// --- State: Parsing Text (or looking for start of a tool use) ---
-		// no currentToolUse
-		let didStartToolUse = false
-		const possibleToolUseOpeningTags = toolUseNames.map((name) => `<${name}>`)
-		for (const toolUseOpeningTag of possibleToolUseOpeningTags) {
-			if (accumulator.endsWith(toolUseOpeningTag)) {
-				// Start of a new tool use found
-				const toolName = toolUseOpeningTag.slice(1, -1) as ToolUseName
-				currentToolUse = {
-					type: "tool_use",
-					name: toolName,
-					params: {},
-					partial: true,
-				}
-				currentToolUseStartIndex = accumulator.length
-
-				// This also indicates the end of the current text content block (if any)
-				if (currentTextContent) {
-					currentTextContent.partial = false
-					// Extract text content, removing the part that formed the tool opening tag
-					const textEndIndex = accumulator.length - toolUseOpeningTag.length
-					currentTextContent.content = accumulator.slice(currentTextContentStartIndex, textEndIndex).trim()
-					// Only add if there's actual content
-					if (currentTextContent.content.length > 0) {
-						contentBlocks.push(currentTextContent)
-					}
-					currentTextContent = undefined
-				} else {
-					// Check if there was text before this tool use started
-					const textEndIndex = accumulator.length - toolUseOpeningTag.length
-					const potentialText = accumulator.slice(currentTextContentStartIndex, textEndIndex).trim()
-					if (potentialText.length > 0) {
-						contentBlocks.push({
-							type: "text",
-							content: potentialText,
-							partial: false, // Ended because tool use started
-						})
-					}
-				}
-
-				didStartToolUse = true
-				break // Found tool start, stop checking for others
-			}
-		}
-
-		if (!didStartToolUse) {
-			// No tool use started, so it must be text content accumulating
-			// (or continuing after a closed tool use)
-			if (currentTextContent === undefined) {
-				// Start of a new text block
-				currentTextContentStartIndex = i - (accumulator.length - currentTextContentStartIndex - 1) // Adjust start index based on how much we've accumulated since the last block ended or the beginning
-				// If accumulator starts from 0, start index is i
-				if (contentBlocks.length === 0 && currentToolUse === undefined) {
-					currentTextContentStartIndex = accumulator.length - 1 // i
-				} else {
-					// Re-calculate based on the actual start of the current text segment
-					// Find the end of the last block
-					let lastBlockEndIndex = 0
-					if (contentBlocks.length > 0) {
-						const lastBlock = contentBlocks[contentBlocks.length - 1]
-						// Approximation: find where the accumulator matches the end of the message string representation of the last block. This is complex.
-						// Simpler: Assume text starts right after the last block ended implicitly at index i.
-						lastBlockEndIndex = i // Where the loop *was* when the last block finished processing
-						// Need a more robust way to track the end index of the *raw string* corresponding to the last block.
-						// Let's stick to the accumulator slice approach for simplicity in this version.
-						// The start index should be where the current *unmatched* text began.
-						let lastProcessedIndex = -1
-						if (contentBlocks.length > 0) {
-							// This requires knowing the raw string length of the previous block, which V1 doesn't explicitly track easily.
-							// We'll approximate based on the current accumulator and start index logic.
-							// The issue arises if a tool tag was just closed. accumulator contains everything up to i.
-							// lastBlockEndIndex should point to the character *after* the closing tag of the last block.
-						}
-						// Reset start index to the beginning of the *current* potential text block
-						currentTextContentStartIndex = accumulator.length - 1 // Start accumulating from the current character `i`
-					}
-
-					// If we just closed a tool, text starts *after* its closing tag
-					// The logic needs refinement here for accurate start index after a tool closure.
-					// Let's assume for now the start index logic inside the loop handles it via slicing.
-				}
-
-				currentTextContent = {
-					type: "text",
-					content: "", // Content will be filled by slicing accumulator
-					partial: true,
-				}
-			}
-			// Update text content based on the accumulator from its start index
-			currentTextContent.content = accumulator.slice(currentTextContentStartIndex).trimStart() // Trim start to avoid leading space if text follows tool
-		}
-	} // End of loop
-
-	// --- Finalization after loop ---
-
-	// If a tool use was open at the end
-	if (currentToolUse) {
-		// If a parameter was open within that tool use
-		if (currentParamName) {
-			// The remaining accumulator content belongs to this partial parameter
-			currentToolUse.params[currentParamName] = accumulator.slice(currentParamValueStartIndex).trim()
-		}
-		// Add the potentially partial tool use block
-		contentBlocks.push(currentToolUse)
-	}
-	// If text content was being accumulated at the end
-	// Note: Only one of currentToolUse or currentTextContent can be defined here,
-	// as starting a tool use finalizes the preceding text block.
-	else if (currentTextContent) {
-		// Update content one last time
-		currentTextContent.content = accumulator.slice(currentTextContentStartIndex).trim()
-		// Add the potentially partial text block only if it contains content
-		if (currentTextContent.content.length > 0) {
-			contentBlocks.push(currentTextContent)
-		}
-	}
-
-	return contentBlocks
-}
+// parseAssistantmessageV1 removed in https://github.com/cline/cline/pull/5425
 
 /**
  * @description **Version 2**
@@ -543,621 +304,3 @@ export function parseAssistantMessageV2(assistantMessage: string): AssistantMess
 
 	return contentBlocks
 }
-
-export function parseAssistantMessageV3(assistantMessage: string): AssistantMessageContent[] {
-	const contentBlocks: AssistantMessageContent[] = []
-	let currentTextContentStart = 0 // Index where the current text block started
-	let currentTextContent: TextContent | undefined = undefined
-	let currentToolUseStart = 0 // Index *after* the opening tag of the current tool use
-	let currentToolUse: ToolUse | undefined = undefined
-	let currentParamValueStart = 0 // Index *after* the opening tag of the current param
-	let currentParamName: ToolParamName | undefined = undefined
-
-	// Precompute tags for faster lookups
-	const toolUseOpenTags = new Map<string, ToolUseName>()
-	const toolParamOpenTags = new Map<string, ToolParamName>()
-	for (const name of toolUseNames) {
-		toolUseOpenTags.set(`<${name}>`, name)
-	}
-	for (const name of toolParamNames) {
-		toolParamOpenTags.set(`<${name}>`, name)
-	}
-
-	// Function calls format detection
-	const isFunctionCallsOpen = "<function_calls>"
-	const isFunctionCallsClose = "</function_calls>"
-	const isInvokeStart = '<invoke name="'
-	const isInvokeEnd = '">'
-	const isInvokeClose = "</invoke>"
-	const isParameterStart = '<parameter name="'
-	const isParameterNameEnd = '">'
-	const isParameterClose = "</parameter>"
-
-	// Variables for function calls parsing
-	let inFunctionCalls = false
-	let currentInvokeName = ""
-	let currentParameterName = ""
-
-	const len = assistantMessage.length
-	for (let i = 0; i < len; i++) {
-		const currentCharIndex = i
-
-		// --- State: Parsing Function Calls ---
-		// Check for opening function_calls tag
-		if (
-			!inFunctionCalls &&
-			currentCharIndex >= isFunctionCallsOpen.length - 1 &&
-			assistantMessage.startsWith(isFunctionCallsOpen, currentCharIndex - isFunctionCallsOpen.length + 1)
-		) {
-			// End current text block if one was active
-			if (currentTextContent) {
-				currentTextContent.content = assistantMessage
-					.slice(currentTextContentStart, currentCharIndex - isFunctionCallsOpen.length + 1)
-					.trim()
-				currentTextContent.partial = false
-				if (currentTextContent.content.length > 0) {
-					contentBlocks.push(currentTextContent)
-				}
-				currentTextContent = undefined
-			}
-
-			inFunctionCalls = true
-			continue
-		}
-
-		// Check for invoke start within function_calls
-		if (
-			inFunctionCalls &&
-			currentInvokeName === "" &&
-			!currentToolUse && // Don't create a new tool if we already have one
-			currentCharIndex >= isInvokeStart.length - 1 &&
-			assistantMessage.startsWith(isInvokeStart, currentCharIndex - isInvokeStart.length + 1)
-		) {
-			// Find the end of the invoke name
-			const nameEndPos = assistantMessage.indexOf(isInvokeEnd, currentCharIndex + 1)
-			if (nameEndPos !== -1) {
-				// Extract the invoke name
-				currentInvokeName = assistantMessage.slice(currentCharIndex + 1, nameEndPos)
-				i = nameEndPos + isInvokeEnd.length - 1 // Skip to after the '">
-
-				// If this is an LS invoke, create a list_files tool
-				if (currentInvokeName === "LS") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "list_files",
-						params: {},
-						partial: true,
-					}
-				}
-
-				// If this is a Grep invoke, create a search_files tool
-				if (currentInvokeName === "Grep") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "search_files",
-						params: {},
-						partial: true,
-					}
-				}
-
-				if (currentInvokeName === "Bash") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "execute_command",
-						params: {},
-						partial: true,
-					}
-				}
-
-				if (currentInvokeName === "Read") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "read_file",
-						params: {},
-						partial: true,
-					}
-				}
-
-				if (currentInvokeName === "Write") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "write_to_file",
-						params: {},
-						partial: true,
-					}
-				}
-
-				if (currentInvokeName === "WebFetch") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "web_fetch",
-						params: {},
-						partial: true,
-					}
-				}
-
-				if (currentInvokeName === "AskQuestion") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "ask_followup_question",
-						params: {},
-						partial: true,
-					}
-				}
-
-				if (currentInvokeName === "UseMCPTool") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "use_mcp_tool",
-						params: {},
-						partial: true,
-					}
-				}
-
-				if (currentInvokeName === "AccessMCPResource") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "access_mcp_resource",
-						params: {},
-						partial: true,
-					}
-				}
-
-				if (currentInvokeName === "ListCodeDefinitionNames") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "list_code_definition_names",
-						params: {},
-						partial: true,
-					}
-				}
-
-				if (currentInvokeName === "PlanModeRespond") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "plan_mode_respond",
-						params: {},
-						partial: true,
-					}
-				}
-
-				if (currentInvokeName === "LoadMcpDocumentation") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "load_mcp_documentation",
-						params: {},
-						partial: true,
-					}
-				}
-
-				if (currentInvokeName === "AttemptCompletion") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "attempt_completion",
-						params: {},
-						partial: true,
-					}
-				}
-
-				if (currentInvokeName === "BrowserAction") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "browser_action",
-						params: {},
-						partial: true,
-					}
-				}
-
-				if (currentInvokeName === "NewTask") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "new_task",
-						params: {},
-						partial: true,
-					}
-				}
-
-				// If this is a MultiEdit invoke, create a replace_in_file tool
-				if (currentInvokeName === "MultiEdit") {
-					currentToolUse = {
-						type: "tool_use",
-						name: "replace_in_file",
-						params: {},
-						partial: true,
-					}
-				}
-
-				continue
-			}
-		}
-
-		// Check for parameter start within invoke
-		if (
-			inFunctionCalls &&
-			currentInvokeName !== "" &&
-			currentParameterName === "" &&
-			currentCharIndex >= isParameterStart.length - 1 &&
-			assistantMessage.startsWith(isParameterStart, currentCharIndex - isParameterStart.length + 1)
-		) {
-			// Find the end of the parameter name
-			const nameEndPos = assistantMessage.indexOf(isParameterNameEnd, currentCharIndex + 1)
-			if (nameEndPos !== -1) {
-				// Extract the parameter name
-				currentParameterName = assistantMessage.slice(currentCharIndex + 1, nameEndPos)
-				currentParamValueStart = nameEndPos + isParameterNameEnd.length
-				i = nameEndPos + isParameterNameEnd.length - 1 // Skip to after the '">'
-				continue
-			}
-		}
-
-		// Check for parameter end
-		if (
-			inFunctionCalls &&
-			currentInvokeName !== "" &&
-			currentParameterName !== "" &&
-			currentCharIndex >= isParameterClose.length - 1 &&
-			assistantMessage.startsWith(isParameterClose, currentCharIndex - isParameterClose.length + 1)
-		) {
-			// Extract parameter value
-			const value = assistantMessage.slice(currentParamValueStart, currentCharIndex - isParameterClose.length + 1).trim()
-
-			// Map parameter to tool params
-			if (currentToolUse && currentInvokeName === "LS" && currentParameterName === "path") {
-				currentToolUse.params["path"] = value
-				// Default recursive to false - only show top level
-				currentToolUse.params["recursive"] = "false"
-			}
-
-			if (currentToolUse && currentInvokeName === "Read" && currentParameterName === "file_path") {
-				currentToolUse.params["path"] = value
-			}
-
-			if (currentToolUse && currentInvokeName === "PlanModeRespond" && currentParameterName === "response") {
-				currentToolUse.params["response"] = value
-			}
-
-			if (currentToolUse && currentInvokeName === "WebFetch" && currentParameterName === "url") {
-				currentToolUse.params["url"] = value
-			}
-
-			if (currentToolUse && currentInvokeName === "ListCodeDefinitionNames" && currentParameterName === "path") {
-				currentToolUse.params["path"] = value
-			}
-
-			if (currentToolUse && currentInvokeName === "NewTask" && currentParameterName === "context") {
-				currentToolUse.params["context"] = value
-			}
-
-			// Map parameter to tool params for Grep
-			if (currentToolUse && currentInvokeName === "Grep") {
-				if (currentParameterName === "pattern") {
-					currentToolUse.params["regex"] = value
-				} else if (currentParameterName === "path") {
-					currentToolUse.params["path"] = value
-				} else if (currentParameterName === "include") {
-					currentToolUse.params["file_pattern"] = value
-				}
-			}
-
-			if (currentToolUse && currentInvokeName === "Bash") {
-				if (currentParameterName === "command") {
-					currentToolUse.params["command"] = value
-				} else if (currentParameterName === "requires_approval") {
-					currentToolUse.params["requires_approval"] = value === "true" ? "true" : "false"
-				}
-			}
-
-			if (currentToolUse && currentInvokeName === "Write") {
-				if (currentParameterName === "file_path") {
-					currentToolUse.params["path"] = value
-				} else if (currentParameterName === "content") {
-					currentToolUse.params["content"] = value
-				}
-			}
-
-			if (currentToolUse && currentInvokeName === "AskQuestion") {
-				if (currentParameterName === "question") {
-					currentToolUse.params["question"] = value
-				} else if (currentParameterName === "options") {
-					currentToolUse.params["options"] = value
-				}
-			}
-
-			if (currentToolUse && currentInvokeName === "UseMCPTool") {
-				if (currentParameterName === "server_name") {
-					currentToolUse.params["server_name"] = value
-				} else if (currentParameterName === "tool_name") {
-					currentToolUse.params["tool_name"] = value
-				} else if (currentParameterName === "arguments") {
-					currentToolUse.params["arguments"] = value
-				}
-			}
-
-			if (currentToolUse && currentInvokeName === "AccessMCPResource") {
-				if (currentParameterName === "server_name") {
-					currentToolUse.params["server_name"] = value
-				} else if (currentParameterName === "uri") {
-					currentToolUse.params["uri"] = value
-				}
-			}
-
-			if (currentToolUse && currentInvokeName === "AttemptCompletion") {
-				if (currentParameterName === "result") {
-					currentToolUse.params["result"] = value
-				}
-				if (currentParameterName === "command") {
-					currentToolUse.params["command"] = value
-				}
-			}
-
-			if (currentToolUse && currentInvokeName === "BrowserAction") {
-				if (currentParameterName === "action") {
-					currentToolUse.params["action"] = value
-				} else if (currentParameterName === "url") {
-					currentToolUse.params["url"] = value
-				} else if (currentParameterName === "coordinate") {
-					currentToolUse.params["coordinate"] = value
-				} else if (currentParameterName === "text") {
-					currentToolUse.params["text"] = value
-				}
-			}
-
-			// Map parameter to tool params for MultiEdit
-			if (currentToolUse && currentInvokeName === "MultiEdit") {
-				if (currentParameterName === "file_path") {
-					currentToolUse.params["path"] = value
-				} else if (currentParameterName === "edits") {
-					// Save the value to the diff parameter for replace_in_file
-					currentToolUse.params["diff"] = value
-				}
-			}
-
-			currentParameterName = ""
-			continue
-		}
-
-		// Check for invoke end
-		if (
-			inFunctionCalls &&
-			currentInvokeName !== "" &&
-			currentCharIndex >= isInvokeClose.length - 1 &&
-			assistantMessage.startsWith(isInvokeClose, currentCharIndex - isInvokeClose.length + 1)
-		) {
-			// If we have a tool use from this invoke, finalize it
-			if (
-				currentToolUse &&
-				(currentInvokeName === "LS" ||
-					currentInvokeName === "Grep" ||
-					currentInvokeName === "Bash" ||
-					currentInvokeName === "Read" ||
-					currentInvokeName === "Write" ||
-					currentInvokeName === "WebFetch" ||
-					currentInvokeName === "AskQuestion" ||
-					currentInvokeName === "UseMCPTool" ||
-					currentInvokeName === "AccessMCPResource" ||
-					currentInvokeName === "ListCodeDefinitionNames" ||
-					currentInvokeName === "PlanModeRespond" ||
-					currentInvokeName === "LoadMcpDocumentation" ||
-					currentInvokeName === "AttemptCompletion" ||
-					currentInvokeName === "BrowserAction" ||
-					currentInvokeName === "NewTask" ||
-					currentInvokeName === "MultiEdit")
-			) {
-				currentToolUse.partial = false
-				contentBlocks.push(currentToolUse)
-				currentToolUse = undefined
-			}
-			currentInvokeName = ""
-			continue
-		}
-
-		// Check for function_calls end
-		if (
-			inFunctionCalls &&
-			currentCharIndex >= isFunctionCallsClose.length - 1 &&
-			assistantMessage.startsWith(isFunctionCallsClose, currentCharIndex - isFunctionCallsClose.length + 1)
-		) {
-			inFunctionCalls = false
-			currentTextContentStart = currentCharIndex + 1
-			// Start a new text content block for any text after function_calls
-			currentTextContent = {
-				type: "text",
-				content: "",
-				partial: true,
-			}
-			continue
-		}
-
-		// Skip normal parsing when inside function_calls
-		if (inFunctionCalls) {
-			continue
-		}
-
-		// --- State: Parsing a Tool Parameter ---
-		if (currentToolUse && currentParamName) {
-			const closeTag = `</${currentParamName}>`
-			// Check if the string *ending* at index `i` matches the closing tag
-			if (
-				currentCharIndex >= closeTag.length - 1 &&
-				assistantMessage.startsWith(
-					closeTag,
-					currentCharIndex - closeTag.length + 1, // Start checking from potential start of tag
-				)
-			) {
-				// Found the closing tag for the parameter
-				const value = assistantMessage
-					.slice(
-						currentParamValueStart, // Start after the opening tag
-						currentCharIndex - closeTag.length + 1, // End before the closing tag
-					)
-					.trim()
-				currentToolUse.params[currentParamName] = value
-				currentParamName = undefined // Go back to parsing tool content
-				// We don't continue loop here, need to check for tool close or other params at index i
-			} else {
-				continue // Still inside param value, move to next char
-			}
-		}
-
-		// --- State: Parsing a Tool Use (but not a specific parameter) ---
-		if (currentToolUse && !currentParamName) {
-			// Ensure we are not inside a parameter already
-			// Check if starting a new parameter
-			let startedNewParam = false
-			for (const [tag, paramName] of toolParamOpenTags.entries()) {
-				if (currentCharIndex >= tag.length - 1 && assistantMessage.startsWith(tag, currentCharIndex - tag.length + 1)) {
-					currentParamName = paramName
-					currentParamValueStart = currentCharIndex + 1 // Value starts after the tag
-					startedNewParam = true
-					break
-				}
-			}
-			if (startedNewParam) {
-				continue // Handled start of param, move to next char
-			}
-
-			// Check if closing the current tool use
-			const toolCloseTag = `</${currentToolUse.name}>`
-			if (
-				currentCharIndex >= toolCloseTag.length - 1 &&
-				assistantMessage.startsWith(toolCloseTag, currentCharIndex - toolCloseTag.length + 1)
-			) {
-				// End of the tool use found
-				// Special handling for content params *before* finalizing the tool
-				const toolContentSlice = assistantMessage.slice(
-					currentToolUseStart, // From after the tool opening tag
-					currentCharIndex - toolCloseTag.length + 1, // To before the tool closing tag
-				)
-
-				// Check if content parameter needs special handling (write_to_file/new_rule)
-				// This check is important if the closing </content> tag was missed by the parameter parsing logic
-				// (e.g., if content is empty or parsing logic prioritizes tool close)
-				const contentParamName: ToolParamName = "content"
-				if (
-					currentToolUse.name === "write_to_file" /* || currentToolUse.name === "new_rule" */ &&
-					toolContentSlice.includes(`<${contentParamName}>`)
-				) {
-					const contentStartTag = `<${contentParamName}>`
-					const contentEndTag = `</${contentParamName}>`
-					const contentStart = toolContentSlice.indexOf(contentStartTag)
-					// Use lastIndexOf for robustness against nested tags
-					const contentEnd = toolContentSlice.lastIndexOf(contentEndTag)
-
-					if (contentStart !== -1 && contentEnd !== -1 && contentEnd > contentStart) {
-						const contentValue = toolContentSlice.slice(contentStart + contentStartTag.length, contentEnd).trim()
-						currentToolUse.params[contentParamName] = contentValue
-					}
-				}
-
-				currentToolUse.partial = false // Mark as complete
-				contentBlocks.push(currentToolUse)
-				currentToolUse = undefined // Reset state
-				currentTextContentStart = currentCharIndex + 1 // Potential text starts after this tag
-				continue // Move to next char
-			}
-			// If not starting a param and not closing the tool, continue accumulating tool content implicitly
-			continue
-		}
-
-		// --- State: Parsing Text / Looking for Tool Start ---
-		if (!currentToolUse) {
-			// Check if starting a new tool use
-			let startedNewTool = false
-			for (const [tag, toolName] of toolUseOpenTags.entries()) {
-				if (currentCharIndex >= tag.length - 1 && assistantMessage.startsWith(tag, currentCharIndex - tag.length + 1)) {
-					// End current text block if one was active
-					if (currentTextContent) {
-						currentTextContent.content = assistantMessage
-							.slice(
-								currentTextContentStart, // From where text started
-								currentCharIndex - tag.length + 1, // To before the tool tag starts
-							)
-							.trim()
-						currentTextContent.partial = false // Ended because tool started
-						if (currentTextContent.content.length > 0) {
-							contentBlocks.push(currentTextContent)
-						}
-						currentTextContent = undefined
-					} else {
-						// Check for any text between the last block and this tag
-						const potentialText = assistantMessage
-							.slice(
-								currentTextContentStart, // From where text *might* have started
-								currentCharIndex - tag.length + 1, // To before the tool tag starts
-							)
-							.trim()
-						if (potentialText.length > 0) {
-							contentBlocks.push({
-								type: "text",
-								content: potentialText,
-								partial: false,
-							})
-						}
-					}
-
-					// Start the new tool use
-					currentToolUse = {
-						type: "tool_use",
-						name: toolName,
-						params: {},
-						partial: true, // Assume partial until closing tag is found
-					}
-					currentToolUseStart = currentCharIndex + 1 // Tool content starts after the opening tag
-					startedNewTool = true
-					break
-				}
-			}
-
-			if (startedNewTool) {
-				continue // Handled start of tool, move to next char
-			}
-
-			// If not starting a tool, it must be text content
-			if (!currentTextContent) {
-				// Start a new text block if we aren't already in one
-				currentTextContentStart = currentCharIndex // Text starts at the current character
-				// Check if the current char is the start of potential text *immediately* after a tag
-				// This needs the previous state - simpler to let slicing handle it later.
-				// Resetting start index accurately is key.
-				// It should be the index *after* the last processed tag.
-				// The logic managing currentTextContentStart after closing tags handles this.
-
-				currentTextContent = {
-					type: "text",
-					content: "", // Will be determined by slicing at the end or when a tool starts
-					partial: true,
-				}
-			}
-			// Continue accumulating text implicitly; content is extracted later.
-		}
-	} // End of loop
-
-	// --- Finalization after loop ---
-
-	// Finalize any open parameter within an open tool use
-	if (currentToolUse && currentParamName) {
-		currentToolUse.params[currentParamName] = assistantMessage
-			.slice(currentParamValueStart) // From param start to end of string
-			.trim()
-		// Tool use remains partial
-	}
-
-	// Finalize any open tool use (which might contain the finalized partial param)
-	if (currentToolUse) {
-		// Tool use is partial because the loop finished before its closing tag
-		contentBlocks.push(currentToolUse)
-	}
-	// Finalize any trailing text content
-	// Only possible if a tool use wasn't open at the very end
-	else if (currentTextContent) {
-		currentTextContent.content = assistantMessage
-			.slice(currentTextContentStart) // From text start to end of string
-			.trim()
-		// Text is partial because the loop finished
-		if (currentTextContent.content.length > 0) {
-			contentBlocks.push(currentTextContent)
-		}
-	}
-
-	return contentBlocks
-}

evals/diff-edits/types.ts

@@ -104,5 +104,6 @@ export interface TestInput {
 	thinkingBudgetTokens: number
 	originalDiffEditToolCallMessage?: string
 	diffApplyFile?: string
+	provider?: string
 	isVerbose: boolean
 }