fix models.proto types

* .

* Remove unused imports

.changeset/chatty-turkeys-wait.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+fix: respect setting litellm models for plan and act

.changeset/cool-comics-change.md

@@ -1,5 +0,0 @@
----
-"claude-dev": minor
----
-
-Ollama: Use a filterable dropdown instead of radio selection

.changeset/curvy-bulldogs-jog.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": minor
+---
+
+Migrate didBecomeVisible to protobus

.changeset/forty-singers-yawn.md

@@ -1,5 +0,0 @@
----
-"claude-dev": patch
----
-
-fix: Address memory leak by bypassing subscribeToState gRPC stream for state updates

.changeset/lazy-cherries-search.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+Prevent reading IS_DEV from the users environment

.changeset/pretty-bikes-tickle.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+fix(bedrock): remove custom Model encode

.changeset/silver-trains-learn.md

@@ -1,5 +0,0 @@
----
-"claude-dev": patch
----
-
-adding support for streamable mcp server

.changeset/thirty-donkeys-whisper.md

@@ -0,0 +1,5 @@
+---
+"claude-dev": patch
+---
+
+The close task button and delete task button in the task header are now correctly announced by screen readers.

.eslintrc.json

@@ -20,8 +20,15 @@
 		"no-throw-literal": "warn",
 		"semi": "off",
 		"react-hooks/exhaustive-deps": "off",
-		"eslint-rules/no-protobuf-object-literals": "warn",
-		"eslint-rules/no-grpc-client-object-literals": "error"
+		"eslint-rules/no-protobuf-object-literals": "error",
+		"eslint-rules/no-grpc-client-object-literals": "error",
+		"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/ISSUE_TEMPLATE/config.yml

@@ -1,4 +1,4 @@
-blank_issues_enabled: true
+blank_issues_enabled: false
 contact_links:
     - name: ✨ Feature Request
       url: https://github.com/cline/cline/discussions/categories/feature-requests?discussions_q=is%3Aopen+category%3A%22Feature+Requests%22+sort%3Atop
@@ -6,6 +6,3 @@ contact_links:
     - name: 👋 Cline Discord
       url: https://discord.gg/cline
       about: Join our Discord community for discussions and support
-    - name: ❓ Other Questions?
-      url: https://x.com/sdrzn
-      about: Contact the developer on X @sdrzn for other inquiries

.github/ISSUE_TEMPLATE/feature_proposal.yml

@@ -0,0 +1,130 @@
+name: 📝 Detailed Feature Proposal
+description: Propose a new feature or improvement
+labels: ["proposal"]
+body:
+    - type: markdown
+      attributes:
+          value: |
+              **Feature Proposal for Cline**
+
+              Thank you for creating a feature proposal for Cline! This template is for clear, actionable proposals that define a specific problem and a high-confidence solution. Please provide enough detail to enable fast prioritization, discussion, and execution.
+
+              Detailed proposals will be prioritized, while vague proposals may be closed or require extensive back and forth communication.
+
+              Before submitting:
+              - Search existing [Issues](https://github.com/cline/cline/issues) and [Discussions](https://github.com/cline/cline/discussions) to avoid duplicates
+              - Don’t start implementation until the proposal is reviewed and approved
+
+    - type: textarea
+      id: problem-description
+      attributes:
+          label: What problem does this solve?
+          description: |
+              Describe the problem clearly from a user's point of view. Focus on why this matters, who it affects, and when it occurs.
+
+              ✅ Solid:
+              - "LLM provider returns 400 error when nearing the context window instead of truncating"
+              - "Submit button is invisible in dark mode"
+
+              ❌ Avoid:
+              - "Performance is bad"
+              - "UI needs work"
+
+              Your description should include:
+              - Who is affected?
+              - When does it happen?
+              - What's the current vs expected behavior?
+              - What is the impact?
+          placeholder: Be specific about the problem, who it affects, and the impact.
+      validations:
+          required: true
+
+    - type: textarea
+      id: proposed-solution
+      attributes:
+          label: What’s the proposed solution?
+          description: |
+              Describe how the problem should be solved. Be specific about UX, system behavior, and any flows that would change.
+
+              ✅ Solid:
+              - "Add error handling immediately after attempting to create the llm stream and retry after manually truncating"
+              - "Update button styling to ensure contrast in all themes"
+
+              ❌ Avoid:
+              - "Improve performance"
+              - "Fix the bug"
+
+              Your solution should include:
+              - What exactly will change?
+              - How will users interact with it?
+              - What’s the expected outcome?
+          placeholder: Describe the proposed changes and how they solve the problem.
+      validations:
+          required: true
+
+    - type: textarea
+      id: acceptance-criteria
+      attributes:
+          label: How will we know it works? (Acceptance Criteria)
+          description: Define clear, testable success criteria.
+          placeholder: Provide specific and testable conditions for success.
+      validations:
+          required: true
+
+    - type: textarea
+      id: estimated-effort
+      attributes:
+          label: Estimated effort and complexity
+          description: |
+              Help us understand scope and risks. Include:
+
+              - Size estimate (XS/S/M/L/XL or hours/days)
+              - Why this size? What’s technically involved?
+              - Any tricky parts, refactors, or risks?
+              - Performance or compatibility concerns?
+              - Any dependencies on systems, teams, or libraries?
+          placeholder: Size, reasoning, risks, and dependencies.
+      validations:
+          required: true
+
+    - type: textarea
+      id: technical-considerations
+      attributes:
+          label: Technical considerations, tradeoffs, and/or risks (optional)
+          description: |
+              Include any technical context that helps us evaluate or implement the proposal more effectively.
+
+              You may include:
+              - Architectural changes or required refactors
+              - Performance implications or system-level impacts
+              - Known limitations, risks, or tricky edge cases
+              - Compatibility concerns or migration steps
+              - Alternative approaches you considered and why they were not chosen
+              - Dependencies on other systems, teams, or libraries
+              - Were other approaches considered? Why is this one preferred?
+          placeholder: Technical considerations, tradeoffs, and/or risks.
+
+    - type: textarea
+      id: additional-context
+      attributes:
+          label: Additional context (optional)
+          description: Diagrams, mockups, logs, links, or anything else that helps explain or justify the proposal.
+          placeholder: Diagrams, mockups, logs, links, or anything else that helps explain or justify the proposal.
+
+    - type: checkboxes
+      id: checklist
+      attributes:
+          label: Proposal checklist
+          options:
+              - label: I’ve checked for existing issues or related proposals
+                required: true
+              - label: I understand this needs review before implementation can start
+                required: true
+
+    - type: checkboxes
+      id: willingness-to-contribute
+      attributes:
+          label: Interested in implementing this?
+          description: Optional
+          options:
+              - label: Yes, I’d like to help implement this feature

.github/pull_request_template.md

@@ -1,10 +1,47 @@
+<!--
+Thank you for contributing to Cline!
+
+⚠️ Important: Before submitting this PR, please ensure you have:
+- Opened an issue and discussed your proposed changes with the community / contributors
+- Received approval from a core Cline contributor prior to proceeding with the implementation
+- Link the associated issue in the "Related Issue" section
+
+Limited exceptions:
+Small bug fixes, typo corrections, minor wording improvements, or simple type fixes that don't change functionality may be submitted directly.
+
+Why this requirement?
+We deeply appreciate all community contributions - they are the core reason we're able to operate successfully and keep innovating! We welcome community input and want to make it as easy as possible for people to submit quality work. This process helps our core maintainers review new ideas faster and saves contributor time by ensuring you have the go-ahead before spending time on implementation.
+-->
+
+### Related Issue
+
+<!-- Replace XXXX with the issue number that this PR addresses -->
+**Issue:** #XXXX
+
 ### Description
 
-<!-- Describe your changes in detail. What problem does this PR solve? -->
+<!-- 
+Help reviewers understand your changes by making this PR readable and well-organized:
+
+- What problem does this PR solve?
+- Why were these changes introduced and what purpose do they serve?
+- For larger changes, provide context about your approach and reasoning
+
+Small PRs may need minimal description, but larger changes benefit from explaining where you're coming from. Much of this context can be in the linked issue above, so feel free to reference it rather than repeating everything here.
+-->
 
 ### Test Procedure
 
-<!-- How did you test this? Are you confident that it will not introduce bugs? If so, why? -->
+<!-- 
+Please walk us through your testing approach and thought process. This helps reviewers understand that you've thoroughly considered the impact of your changes:
+
+- How did you test this change?
+- What could potentially break and how did you verify it doesn't?
+- What existing functionality might be affected and how did you check it still works?
+- Why are you confident this is ready for merge?
+
+We're not looking for exhaustive documentation - just evidence that you've thought through the implications of your changes and tested accordingly.
+-->
 
 ### Type of Change
 
@@ -29,7 +66,15 @@
 
 ### Screenshots
 
-<!-- For UI changes, add screenshots here -->
+<!-- 
+Help reviewers quickly understand your changes:
+
+- **UI Changes**: Please include screenshots showing before/after states
+- **Complex Workflows**: Consider uploading a screen recording (video) if your changes involve multiple steps or state transitions
+- **Backend Changes**: Not required, but feel free to include terminal output or other evidence that demonstrates functionality
+
+This helps reviewers see what you've built without having to pull down and test your branch first.
+-->
 
 ### Additional Notes
 

.github/workflows/publish.yml

@@ -11,6 +11,10 @@ on:
                 options:
                     - pre-release
                     - release
+            tag:
+                description: "Enter existing tag to publish (e.g., v3.1.2)"
+                required: true
+                type: string
 
 permissions:
     contents: write
@@ -30,6 +34,8 @@ jobs:
 
         steps:
             - uses: actions/checkout@v4
+              with:
+                  ref: ${{ github.event.inputs.tag }}
 
             - name: Setup Node.js
               uses: actions/setup-node@v4
@@ -69,14 +75,20 @@ jobs:
                   VERSION=$(node -p "require('./package.json').version")
                   echo "version=$VERSION" >> $GITHUB_OUTPUT
 
-            - name: Create Git Tag
-              id: create_tag
+            - name: Validate Tag
+              id: validate_tag
               run: |
-                  VERSION=v${{ steps.get_version.outputs.version }}
-                  echo "tag=$VERSION" >> $GITHUB_OUTPUT
-                  echo "Tagging with $VERSION"
-                  git tag "$VERSION"
-                  git push origin "$VERSION"
+                  TAG="${{ github.event.inputs.tag }}"
+                  echo "tag=$TAG" >> $GITHUB_OUTPUT
+                  echo "Using existing tag: $TAG"
+
+                  # Verify the tag exists
+                  if ! git rev-parse "$TAG" >/dev/null 2>&1; then
+                    echo "Error: Tag '$TAG' does not exist in the repository"
+                    exit 1
+                  fi
+
+                  echo "Tag '$TAG' validated successfully"
 
             - name: Package and Publish Extension
               env:
@@ -106,7 +118,7 @@ jobs:
             - name: Create GitHub Release
               uses: softprops/action-gh-release@v1
               with:
-                  tag_name: ${{ steps.create_tag.outputs.tag }}
+                  tag_name: ${{ steps.validate_tag.outputs.tag }}
                   files: "*.vsix"
                   # body: ${{ steps.changelog.outputs.content }}
                   generate_release_notes: true

.github/workflows/test.yml

@@ -68,6 +68,12 @@ jobs:
               if: steps.webview-cache.outputs.cache-hit != 'true'
               run: cd webview-ui && npm ci
 
+            - 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: |
@@ -86,8 +92,9 @@ jobs:
             - name: Build Tests and Extension
               run: npm run pretest
 
-            - name: Unit Tests
-              run: npm run test:unit
+            # Unit Tests disabled due to module system conflicts between backend and webview-ui
+            # - name: Unit Tests
+            #   run: npm run test:unit
 
             # Run extension tests with coverage
             - name: Extension Tests with Coverage

.gitignore

@@ -21,10 +21,18 @@ coverage
 
 *evals.env
 
-# Generated proto files
-src/shared/proto/*.ts
+# Generated files
+src/generated/
+# Core
 src/core/controller/*/methods.ts
 src/core/controller/*/index.ts
 src/core/controller/grpc-service-config.ts
+# Shared
+src/shared/proto/*.ts
+src/shared/proto/host/*.ts
+# Webview
 webview-ui/src/services/grpc-client.ts
-src/standalone/server-setup.ts
+# Host bridge
+src/hosts/vscode/*/methods.ts
+src/hosts/vscode/*/index.ts
+src/hosts/vscode/host-grpc-service-config.ts

.husky/pre-commit

@@ -9,9 +9,9 @@ npm run lint || {
 
 # Run Prettier
 echo "Running Prettier..."
-npm run format || {
-    echo "❌ Prettier check failed. Run 'npm run format:fix' to automatically fix formatting issues."
-    exit 1
-}
+npx lint-staged --verbose || {
+     echo "❌ Prettier failed. Please fix the errors and try committing again."
+     exit 1
+ }
 
 echo "✅ All checks passed!"

.vscode/launch.json

@@ -42,7 +42,7 @@
 		{
 			"type": "node",
 			"request": "launch",
-			"name": "Run Standalone Extension",
+			"name": "Run Standalone Service",
 			"skipFiles": ["<node_internals>/**"],
 			"sourceMaps": true,
 			"resolveSourceMapLocations": ["${workspaceFolder}/**", "!**/node_modules/**"],
@@ -52,7 +52,9 @@
 			"env": {
 				"GRPC_TRACE": "all",
 				"GRPC_VERBOSITY": "DEBUG",
-				"NODE_PATH": "${workspaceFolder}/dist-standalone/node_modules"
+				"NODE_PATH": "${workspaceFolder}/dist-standalone/node_modules",
+				"CLINE_DIR": "${userHome}/.cline-standalone",
+				"HOST_BRIDGE_ADDRESS": "localhost:50052"
 			},
 			"program": "standalone.js"
 		}

.vscode/tasks.json

@@ -128,7 +128,25 @@
 			"type": "npm",
 			"script": "watch:esbuild",
 			"group": "build",
-			"problemMatcher": "$esbuild-watch",
+			"problemMatcher": {
+				"pattern": [
+					{
+						"regexp": "^✘ \\[ERROR\\] (.*)$",
+						"message": 1
+					},
+					{
+						"regexp": "^\\s+(.*):(\\d+):(\\d+):$",
+						"file": 1,
+						"line": 2,
+						"column": 3
+					}
+				],
+				"background": {
+					"activeOnStart": true,
+					"beginsPattern": "^\\[watch\\] build started$",
+					"endsPattern": "^\\[watch\\] build finished$"
+				}
+			},
 			"isBackground": true,
 			"label": "npm: watch:esbuild",
 			"dependsOn": ["npm: protos"],
@@ -146,7 +164,25 @@
 			"type": "npm",
 			"script": "watch:esbuild:test",
 			"group": "build",
-			"problemMatcher": "$esbuild-watch",
+			"problemMatcher": {
+				"pattern": [
+					{
+						"regexp": "^✘ \\[ERROR\\] (.*)$",
+						"message": 1
+					},
+					{
+						"regexp": "^\\s+(.*):(\\d+):(\\d+):$",
+						"file": 1,
+						"line": 2,
+						"column": 3
+					}
+				],
+				"background": {
+					"activeOnStart": true,
+					"beginsPattern": "^\\[watch\\] build started$",
+					"endsPattern": "^\\[watch\\] build finished$"
+				}
+			},
 			"isBackground": true,
 			"label": "npm: watch:esbuild:test",
 			"dependsOn": ["npm: protos"],

CHANGELOG.md

@@ -1,5 +1,57 @@
 # Changelog
 
+## [3.17.14]
+
+-   Add Claude Code as a new API provider, allowing integration with Anthropic's Claude Code CLI tool and Claude Max Plan (Thanks @BarreiroT!)
+-   Add SAP AI Core as a new API provider with support for Claude and GPT models (Thanks @schardosin!)
+-   Add configurable default terminal profile setting, allowing users to specify which terminal Cline should use (Thanks @valinha!)
+-   Add terminal output size constraint setting to limit how much terminal output is processed
+-   Add MCP Rich Display settings to the settings page for persistent configuration (Thanks @Vl4diC0de!)
+-   Improve copy button functionality with refactored reusable components (Thanks @shouhanzen!)
+-   Improve AWS Bedrock provider by removing deprecated dependency and using standard AWS SDK (Thanks @watany-dev!)
+-   Fix list_files tool to properly return files when targeting hidden directories
+-   Fix search and replace edge case that could cause file deletion, making the algorithm more lenient for models using different diff formats
+-   Fix task restoration issues that could occur when resuming interrupted tasks
+-   Fix checkpoint saving to properly track all file changes
+-   Improve file context warnings to reduce diff edit errors when resuming restored tasks
+-   Clear chat input when switching between Plan/Act modes within a task
+-   Exclude .clinerules files from checkpoint tracking
+
+## [3.17.13]
+
+-   Add Thinking UX for Gemini models, providing visual feedback during model reasoning
+-   Add support for Notifications MCP integration with Cline
+-   Add prompt caching indicator for Grok 3 models
+-   Sort MCP marketplace by newest listings by default for easier discovery of recent servers
+-   Update O3 model family pricing to reflect latest OpenAI rates
+-   Remove '-beta' suffix from Grok model identifiers
+-   Fix AWS Bedrock provider by removing deprecated Anthropic-Bedrock SDK (Thanks @watany-dev!)
+-   Fix menu display issue for terminal timeout settings
+-   Improve chat input field styling and behavior
+
+## [3.17.12]
+
+-   **Free Grok Model Available!** Access Grok 3 completely free through the Cline provider
+-   Add collapsible MCP response panels to keep conversations focused on the main AI responses while still allowing access to detailed MCP output (Thanks @valinha!)
+-   Prioritize active files (open tabs) at the top of the file context menu when using @ mentions (Thanks @abeatrix!)
+-   Fix context menu to properly default to "File" option instead of incorrectly selecting "Git Commits"
+-   Fix diff editing to handle out-of-order SEARCH/REPLACE blocks, improving reliability with models that don't follow strict ordering
+-   Fix telemetry warning popup appearing repeatedly for users who have telemetry disabled
+
+## [3.17.11]
+
+-   Add support for Gemini 2.5 Pro Preview 06-05 model to Vertex AI and Google Gemini providers
+
+## [3.17.10]
+
+-   Add support for Qwen 3 series models with thinking mode options (Thanks @Jonny-china!)
+-   Add new AskSage models: Claude 4 Sonnet, Claude 4 Opus, GPT 4.1, Gemini 2.5 Pro (Thanks @swhite24!)
+-   Add VSCode walkthrough to help new users get started with Cline
+-   Add support for streamable MCP servers
+-   Improve Ollama model selection with filterable dropdown instead of radio buttons (Thanks @paulgear!)
+-   Add setting to disable aggressive terminal reuse to help users experiencing task lockout issues
+-   Fix settings dialog applying changes even when cancel button is clicked
+
 ## [3.17.9]
 
 -   Aligning Cline to work with Claude 4 model family (Experimental)

CONTRIBUTING.md

@@ -18,6 +18,18 @@ We also welcome contributions to our [documentation](https://github.com/cline/cl
 
 If you're planning to work on a bigger feature, please create a [feature request](https://github.com/cline/cline/discussions/categories/feature-requests?discussions_q=is%3Aopen+category%3A%22Feature+Requests%22+sort%3Atop) first so we can discuss whether it aligns with Cline's vision.
 
+## Before Contributing
+
+All contributions must begin with a GitHub Issue, unless the change is for small bug fixes, typo corrections, minor wording improvements, or simple type fixes that don't change functionality.
+
+- **Check existing issues**: Search [GitHub Issues](https://github.com/cline/cline/issues).
+- **Create an issue**: Use appropriate templates:
+    - **Bugs:** "Bug Report" template.
+    - **Features:** "Detailed Feature Proposal" template. Approval from a core Cline contributor required before starting.
+- **Claim issues**: Comment your interest.
+
+**PRs without approved issues may be closed.**
+
 ## Development Setup
 
 1. **VS Code Extensions**

buf.yaml

@@ -0,0 +1,22 @@
+version: v2
+modules:
+    - path: proto
+      name: cline/cline/lint
+
+lint:
+    use:
+        - STANDARD
+
+    except: # Add exceptions for current patterns that contradict STANDARD settings
+        - RPC_PASCAL_CASE # rpcs are camel case (start with lowercase)
+        - PACKAGE_DIRECTORY_MATCH # the protos in the cline package are not in a dir named cline.
+        - RPC_REQUEST_RESPONSE_UNIQUE # request messages are not unique.
+        - RPC_REQUEST_STANDARD_NAME # request messages dont all end with Request
+        - RPC_RESPONSE_STANDARD_NAME # response messages dont all end with Response
+        - PACKAGE_VERSION_SUFFIX # package name does not contain version.
+        - ENUM_VALUE_PREFIX # enum values dont start with the enum name.
+        - ENUM_ZERO_VALUE_SUFFIX # first value does not have to be UNSPECIFIED.
+
+# breaking:
+#   use:
+#     - WIRE_JSON # Detect changes that break the json wire format (this is the minimum recommended level.)

esbuild.js

@@ -125,9 +125,11 @@ const baseConfig = {
 	minify: production,
 	sourcemap: !production,
 	logLevel: "silent",
-	define: {
-		"process.env.IS_DEV": JSON.stringify(!production),
-	},
+	define: production
+		? {
+				"process.env.IS_DEV": JSON.stringify(!production),
+			}
+		: undefined,
 	tsconfig: path.resolve(__dirname, "tsconfig.json"),
 	plugins: [
 		copyWasmFiles,

evals/.gitignore

@@ -1,3 +1,6 @@
 repositories
 
-results/evals.db
+results/evals.db
+
+diff_editing/test_cases/
+diff_editing/test_outputs/

evals/cli/package-lock.json

@@ -9,7 +9,7 @@
 			"version": "0.1.0",
 			"license": "MIT",
 			"dependencies": {
-				"better-sqlite3": "^8.0.0",
+				"better-sqlite3": "^11.10.0",
 				"chalk": "^4.1.2",
 				"commander": "^9.4.1",
 				"execa": "^5.1.1",
@@ -217,10 +217,11 @@
 			]
 		},
 		"node_modules/better-sqlite3": {
-			"version": "8.7.0",
-			"resolved": "https://registry.npmjs.org/better-sqlite3/-/better-sqlite3-8.7.0.tgz",
-			"integrity": "sha512-99jZU4le+f3G6aIl6PmmV0cxUIWqKieHxsiF7G34CVFiE+/UabpYqkU0NJIkY/96mQKikHeBjtR27vFfs5JpEw==",
+			"version": "11.10.0",
+			"resolved": "https://registry.npmjs.org/better-sqlite3/-/better-sqlite3-11.10.0.tgz",
+			"integrity": "sha512-EwhOpyXiOEL/lKzHz9AW1msWFNzGc/z+LzeB3/jnFJpxu+th2yqvzsSWas1v9jgs9+xiXJcD5A8CJxAG2TaghQ==",
 			"hasInstallScript": true,
+			"license": "MIT",
 			"dependencies": {
 				"bindings": "^1.5.0",
 				"prebuild-install": "^7.1.1"
@@ -1607,9 +1608,9 @@
 			"integrity": "sha512-AKpaYlHn8t4SVbOHCy+b5+KKgvR4vrsD8vbvrbiQJps7fKDTkjkDry6ji0rUJjC0kzbNePLwzxq8iypo41qeWA=="
 		},
 		"better-sqlite3": {
-			"version": "8.7.0",
-			"resolved": "https://registry.npmjs.org/better-sqlite3/-/better-sqlite3-8.7.0.tgz",
-			"integrity": "sha512-99jZU4le+f3G6aIl6PmmV0cxUIWqKieHxsiF7G34CVFiE+/UabpYqkU0NJIkY/96mQKikHeBjtR27vFfs5JpEw==",
+			"version": "11.10.0",
+			"resolved": "https://registry.npmjs.org/better-sqlite3/-/better-sqlite3-11.10.0.tgz",
+			"integrity": "sha512-EwhOpyXiOEL/lKzHz9AW1msWFNzGc/z+LzeB3/jnFJpxu+th2yqvzsSWas1v9jgs9+xiXJcD5A8CJxAG2TaghQ==",
 			"requires": {
 				"bindings": "^1.5.0",
 				"prebuild-install": "^7.1.1"

evals/cli/package.json

@@ -17,7 +17,7 @@
 	"author": "",
 	"license": "MIT",
 	"dependencies": {
-		"better-sqlite3": "^8.0.0",
+		"better-sqlite3": "^11.10.0",
 		"chalk": "^4.1.2",
 		"commander": "^9.4.1",
 		"execa": "^5.1.1",

evals/cli/src/commands/runDiffEval.ts

@@ -0,0 +1,79 @@
+import execa from "execa"
+import chalk from "chalk"
+import path from "path"
+
+interface RunDiffEvalOptions {
+	modelId: string
+	systemPromptName: string
+	numberOfRuns: number
+	parsingFunction: string
+	diffEditFunction: string
+	thinkingBudget: number
+	parallel: boolean
+	verbose: boolean
+	testPath: string
+	outputPath: string
+	replay: boolean
+}
+
+export async function runDiffEvalHandler(options: RunDiffEvalOptions) {
+	console.log(chalk.blue("Starting diff editing evaluation..."))
+
+	// Resolve the path to the TestRunner.ts script relative to the current file
+	const scriptPath = path.resolve(__dirname, "../../../diff_editing/TestRunner.ts")
+
+	// Construct the arguments array for the execa call
+	const args = [
+		"--model-id",
+		options.modelId,
+		"--system-prompt-name",
+		options.systemPromptName,
+		"--number-of-runs",
+		String(options.numberOfRuns),
+		"--parsing-function",
+		options.parsingFunction,
+		"--diff-edit-function",
+		options.diffEditFunction,
+	]
+
+	// Conditionally add the optional arguments
+	if (options.testPath) {
+		args.push("--test-path", options.testPath)
+	}
+	if (options.outputPath) {
+		args.push("--output-path", options.outputPath)
+	}
+	if (options.thinkingBudget > 0) {
+		args.push("--thinking-budget", String(options.thinkingBudget))
+	}
+
+	if (options.parallel) {
+		args.push("--parallel")
+	}
+
+	if (options.replay) {
+		args.push("--replay")
+	}
+
+	if (options.verbose) {
+		args.push("--verbose")
+	}
+
+	try {
+		console.log(chalk.gray(`Executing: npx tsx ${scriptPath} ${args.join(" ")}`))
+
+		// Execute the script as a child process
+		// We use 'inherit' to stream the stdout/stderr directly to the user's terminal
+		const subprocess = execa("npx", ["tsx", scriptPath, ...args], {
+			stdio: "inherit",
+		})
+
+		await subprocess
+
+		console.log(chalk.green("Diff editing evaluation completed successfully."))
+	} catch (error) {
+		console.error(chalk.red("An error occurred during the diff editing evaluation."))
+		// The 'inherit' stdio will have already printed the error details from the script
+		process.exit(1)
+	}
+}

evals/cli/src/index.ts

@@ -5,6 +5,7 @@ import { setupHandler } from "./commands/setup"
 import { runHandler } from "./commands/run"
 import { reportHandler } from "./commands/report"
 import { evalsEnvHandler } from "./commands/evals-env"
+import { runDiffEvalHandler } from "./commands/runDiffEval"
 
 // Create the CLI program
 const program = new Command()
@@ -77,6 +78,36 @@ program
 		}
 	})
 
+// Run-diff-eval command
+program
+	.command("run-diff-eval")
+	.description("Run the diff editing evaluation suite")
+	.option("--test-path <path>", "Path to the directory containing test case JSON files")
+	.option("--output-path <path>", "Path to the directory to save the test output JSON files")
+	.option("--model-id <model_id>", "The model ID to use for the test")
+	.option("--system-prompt-name <name>", "The name of the system prompt to use", "basicSystemPrompt")
+	.option("-n, --number-of-runs <number>", "Number of times to run each test case", "1")
+	.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("--parallel", "Run tests in parallel", false)
+	.option("--replay", "Run evaluation from a pre-recorded LLM output, skipping the API call", false)
+	.option("-v, --verbose", "Enable verbose logging", false)
+	.action(async (options) => {
+		try {
+			// The logic here simplifies slightly
+			const fullOptions = {
+				...options,
+				numberOfRuns: parseInt(options.numberOfRuns, 10),
+				thinkingBudget: parseInt(options.thinkingBudget, 10),
+			}
+			await runDiffEvalHandler(fullOptions)
+		} catch (error) {
+			console.error(chalk.red(`Error during diff eval run: ${error instanceof Error ? error.message : String(error)}`))
+			process.exit(1)
+		}
+	})
+
 // Parse command line arguments
 program.parse(process.argv)
 

evals/diff_editing/ClineWrapper.ts

@@ -0,0 +1,279 @@
+import { OpenRouterHandler } from "../../src/api/providers/openrouter"
+import { ApiHandlerOptions } from "../../src/shared/api"
+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 constructNewFileContentV1, constructNewFileContentV2 } from "./diff-apply/diff-06-06-25"
+import { constructNewFileContent as constructNewFileContentV3 } from "../../src/core/assistant-message/diff" // this defaults to the new v1 when called
+
+type ParseAssistantMessageFn = (message: string) => AssistantMessageContent[]
+type ConstructNewFileContentFn = (diff: string, original: string, strict: boolean) => Promise<string>
+
+const parsingFunctions: Record<string, ParseAssistantMessageFn> = {
+	parseAssistantMessageV1: parseAssistantMessageV1,
+	parseAssistantMessageV2: parseAssistantMessageV2,
+	parseAssistantMessageV3: parseAssistantMessageV3,
+}
+
+const diffEditingFunctions: Record<string, ConstructNewFileContentFn> = {
+	constructNewFileContentV1: constructNewFileContentV1,
+	constructNewFileContentV2: constructNewFileContentV2,
+	constructNewFileContentV3: constructNewFileContentV3, // position invariant diff
+}
+
+import { TestInput, TestResult, ExtractedToolCall } from "./types"
+export { TestInput, TestResult, ExtractedToolCall }
+
+interface StreamResult {
+	assistantMessage: string
+	reasoningMessage: string
+	usage: {
+		inputTokens: number
+		outputTokens: number
+		cacheWriteTokens: number
+		cacheReadTokens: number
+		totalCost: number
+	}
+}
+
+/**
+ * Process the stream and return full response
+ */
+async function processStream(
+	handler: OpenRouterHandler,
+	systemPrompt: string,
+	messages: Anthropic.Messages.MessageParam[],
+): Promise<StreamResult> {
+	const stream = handler.createMessage(systemPrompt, messages)
+
+	let assistantMessage = ""
+	let reasoningMessage = ""
+	let inputTokens = 0
+	let outputTokens = 0
+	let cacheWriteTokens = 0
+	let cacheReadTokens = 0
+	let totalCost = 0
+
+	for await (const chunk of stream) {
+		if (!chunk) {
+			continue
+		}
+
+		switch (chunk.type) {
+			case "usage":
+				inputTokens += chunk.inputTokens
+				outputTokens += chunk.outputTokens
+				cacheWriteTokens += chunk.cacheWriteTokens ?? 0
+				cacheReadTokens += chunk.cacheReadTokens ?? 0
+				if (chunk.totalCost) {
+					totalCost = chunk.totalCost
+				}
+				break
+			case "reasoning":
+				reasoningMessage += chunk.reasoning
+				break
+			case "text":
+				assistantMessage += chunk.text
+				break
+		}
+	}
+
+	return {
+		assistantMessage,
+		reasoningMessage,
+		usage: {
+			inputTokens,
+			outputTokens,
+			cacheWriteTokens,
+			cacheReadTokens,
+			totalCost,
+		},
+	}
+}
+
+/**
+ * Main evaluation function:
+ * 1. create and process stream
+ * 2. extract any tool calls from the stream
+ * 3. if no diff edit, considered a failure (or rerun) - otherwise attempt to apply the diff edit
+ */
+export async function runSingleEvaluation(input: TestInput): Promise<TestResult> {
+	try {
+		// Extract parameters
+		const {
+			apiKey,
+			systemPrompt,
+			messages,
+			modelId,
+			originalFile,
+			originalFilePath,
+			parsingFunction,
+			diffEditFunction,
+			thinkingBudgetTokens,
+			originalDiffEditToolCallMessage,
+		} = input
+
+		const requiredParams = {
+			systemPrompt,
+			messages,
+			modelId,
+			originalFile,
+			originalFilePath,
+			parsingFunction,
+			diffEditFunction,
+		}
+
+		const missingParams = Object.entries(requiredParams)
+			.filter(([, value]) => !value)
+			.map(([key]) => key)
+
+		if (missingParams.length > 0) {
+			return {
+				success: false,
+				error: "missing_required_parameters",
+				errorString: `Missing required parameters: ${missingParams.join(", ")}`,
+			}
+		}
+
+		const parseAssistantMessage = parsingFunctions[parsingFunction]
+		const constructNewFileContent = diffEditingFunctions[diffEditFunction]
+
+		if (!parseAssistantMessage || !constructNewFileContent) {
+			return {
+				success: false,
+				error: "invalid_functions",
+			}
+		}
+
+		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,
+			},
+		}
+
+		// Get the output of streaming output of this llm call
+		let streamResult: StreamResult
+		if (originalDiffEditToolCallMessage !== undefined) {
+			// Replay mode: mock the stream result
+			streamResult = {
+				assistantMessage: originalDiffEditToolCallMessage,
+				reasoningMessage: "",
+				usage: { inputTokens: 0, outputTokens: 0, cacheWriteTokens: 0, cacheReadTokens: 0, totalCost: 0 },
+			}
+		} else {
+			// Live mode: existing API call logic
+			try {
+				const openRouterHandler = new OpenRouterHandler(options)
+				streamResult = await processStream(openRouterHandler, systemPrompt, messages)
+			} catch (error: any) {
+				return {
+					success: false,
+					error: "llm_stream_error",
+					errorString: error.message || error.toString(),
+				}
+			}
+		}
+
+		// process the assistant message into its constituent tool calls & text blocks
+		const assistantContentBlocks: AssistantMessageContent[] = parseAssistantMessage(streamResult.assistantMessage)
+
+		const detectedToolCalls: ExtractedToolCall[] = []
+
+		for (const block of assistantContentBlocks) {
+			if (block.type === "tool_use") {
+				detectedToolCalls.push({
+					name: block.name,
+					input: block.params,
+				})
+			}
+		}
+
+		// check if there are any tool calls, if there are none then its a clear error
+		if (detectedToolCalls.length === 0) {
+			return {
+				success: false,
+				streamResult: streamResult,
+				toolCalls: detectedToolCalls,
+				error: "no_tool_calls",
+			}
+		}
+
+		// check that there is exactly one tool call, otherwise an error
+		if (detectedToolCalls.length > 1) {
+			return {
+				success: false,
+				streamResult: streamResult,
+				toolCalls: detectedToolCalls,
+				error: "multi_tool_calls",
+			}
+		}
+
+		// check that the tool call is diff edit tool call
+		if (detectedToolCalls[0].name !== "replace_in_file") {
+			return {
+				success: false,
+				streamResult: streamResult,
+				toolCalls: detectedToolCalls,
+				error: "wrong_tool_call",
+			}
+		}
+
+		const toolCall = detectedToolCalls[0]
+		const diffToolPath = toolCall.input.path
+		const diffToolContent = toolCall.input.diff
+
+		if (!diffToolPath || !diffToolContent) {
+			return {
+				success: false,
+				streamResult: streamResult,
+				toolCalls: detectedToolCalls,
+				error: "tool_call_params_undefined",
+			}
+		}
+
+		// check that we are editing the correct file path
+		if (diffToolPath !== originalFilePath) {
+			return {
+				success: false,
+				streamResult: streamResult,
+				toolCalls: detectedToolCalls,
+				error: "wrong_file_edited",
+			}
+		}
+
+		// checking if the diff edit succeeds, if it failed it will throw an error
+		let diffSuccess = true
+		try {
+			await constructNewFileContent(diffToolContent, originalFile, true)
+		} catch (error: any) {
+			diffSuccess = false
+		}
+
+		return {
+			success: true,
+			streamResult: streamResult,
+			toolCalls: detectedToolCalls,
+			diffEdit: diffToolContent,
+			diffEditSuccess: diffSuccess,
+		}
+	} catch (error: any) {
+		return {
+			success: false,
+			error: "other_error",
+			errorString: error.message || error.toString(),
+		}
+	}
+}

evals/diff_editing/TestRunner.ts

@@ -0,0 +1,391 @@
+import { runSingleEvaluation, TestInput, TestResult } from "./ClineWrapper"
+import { basicSystemPrompt } from "./prompts/basicSystemPrompt-06-06-25"
+import { claude4SystemPrompt } from "./prompts/claude4SystemPrompt-06-06-25"
+import { formatResponse } from "./helpers"
+import { Anthropic } from "@anthropic-ai/sdk"
+import * as fs from "fs"
+import * as path from "path"
+import { Command } from "commander"
+import { InputMessage, ProcessedTestCase, TestCase, TestConfig, SystemPromptDetails, ConstructSystemPromptFn } from "./types"
+
+function log(isVerbose: boolean, message: string) {
+	if (isVerbose) {
+		console.log(message)
+	}
+}
+
+const systemPromptGeneratorLookup: Record<string, ConstructSystemPromptFn> = {
+	basicSystemPrompt: basicSystemPrompt,
+	claude4SystemPrompt: claude4SystemPrompt,
+}
+
+type TestResultSet = { [test_id: string]: (TestResult & { test_id?: string })[] }
+
+class NodeTestRunner {
+	private apiKey: string | undefined
+
+	constructor(isReplay: boolean) {
+		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.")
+			}
+		}
+	}
+
+	/**
+	 * convert our messages array into a properly formatted Anthropic messages array
+	 */
+	transformMessages(messages: InputMessage[]): Anthropic.Messages.MessageParam[] {
+		return messages.map((msg) => {
+			// Use TextBlockParam here for constructing the input message
+			const content: (Anthropic.TextBlockParam | Anthropic.ImageBlockParam)[] = []
+
+			if (msg.text) {
+				// This object now correctly matches the TextBlockParam type
+				content.push({ type: "text", text: msg.text })
+			}
+
+			if (msg.images && Array.isArray(msg.images)) {
+				const imageBlocks = formatResponse.imageBlocks(msg.images)
+				content.push(...imageBlocks)
+			}
+
+			return {
+				role: msg.role,
+				content: content,
+			}
+		})
+	}
+
+	/**
+	 * Generate the system prompt on the fly
+	 */
+	constructSystemPrompt(systemPromptDetails: SystemPromptDetails, systemPromptName: string) {
+		const systemPromptGenerator = systemPromptGeneratorLookup[systemPromptName]
+
+		const { cwd_value, browser_use, width, height, os_value, shell_value, home_value, mcp_string, user_custom_instructions } =
+			systemPromptDetails
+
+		const systemPrompt = systemPromptGenerator(
+			cwd_value,
+			browser_use,
+			width,
+			height,
+			os_value,
+			shell_value,
+			home_value,
+			mcp_string,
+			user_custom_instructions,
+		)
+
+		return systemPrompt
+	}
+
+	/**
+	 * Loads our test cases from a directory of json files
+	 */
+	loadTestCases(testDirectoryPath: string): TestCase[] {
+		const testCasesArray: TestCase[] = []
+		const dirents = fs.readdirSync(testDirectoryPath, { withFileTypes: true })
+
+		for (const dirent of dirents) {
+			if (dirent.isFile() && dirent.name.endsWith(".json")) {
+				const testFilePath = path.join(testDirectoryPath, dirent.name)
+				const fileContent = fs.readFileSync(testFilePath, "utf8")
+				const testCase: TestCase = JSON.parse(fileContent)
+
+				// Use the filename (without extension) as the test_id if not provided
+				if (!testCase.test_id) {
+					testCase.test_id = path.parse(dirent.name).name
+				}
+				testCasesArray.push(testCase)
+			}
+		}
+
+		return testCasesArray
+	}
+
+	/**
+	 * Saves the test results to the specified output directory.
+	 */
+	saveTestResults(results: TestResultSet, outputPath: string) {
+		// Ensure output directory exists
+		if (!fs.existsSync(outputPath)) {
+			fs.mkdirSync(outputPath, { recursive: true })
+		}
+
+		// Write each test result to its own file
+		for (const testId in results) {
+			const outputFilePath = path.join(outputPath, `${testId}.json`)
+			const testResult = results[testId]
+			fs.writeFileSync(outputFilePath, JSON.stringify(testResult, null, 2))
+		}
+	}
+
+	/**
+	 * Run a single test example
+	 */
+	async runSingleTest(testCase: ProcessedTestCase, testConfig: TestConfig): Promise<TestResult> {
+		if (testConfig.replay && !testCase.original_diff_edit_tool_call_message) {
+			return {
+				success: false,
+				error: "missing_original_diff_edit_tool_call_message",
+				errorString: `Test case ${testCase.test_id} is missing 'original_diff_edit_tool_call_message' for replay.`,
+			}
+		}
+
+		const customSystemPrompt = this.constructSystemPrompt(testCase.system_prompt_details, testConfig.system_prompt_name)
+
+		// messages don't include system prompt and are everything up to the first replace_in_file tool call which results in a diff edit error
+		const input: TestInput = {
+			apiKey: this.apiKey,
+			systemPrompt: customSystemPrompt,
+			messages: testCase.messages,
+			modelId: testConfig.model_id,
+			originalFile: testCase.file_contents,
+			originalFilePath: testCase.file_path,
+			parsingFunction: testConfig.parsing_function,
+			diffEditFunction: testConfig.diff_edit_function,
+			thinkingBudgetTokens: testConfig.thinking_tokens_budget,
+			originalDiffEditToolCallMessage: testConfig.replay ? testCase.original_diff_edit_tool_call_message : undefined,
+		}
+
+		return await runSingleEvaluation(input)
+	}
+
+	/**
+	 * Runs all the text examples synchonously
+	 */
+	async runAllTests(testCases: ProcessedTestCase[], testConfig: TestConfig, isVerbose: boolean): Promise<TestResultSet> {
+		const results: TestResultSet = {}
+
+		for (const testCase of testCases) {
+			results[testCase.test_id] = []
+
+			log(isVerbose, `-Running test: ${testCase.test_id}`)
+			for (let i = 0; i < testConfig.number_of_runs; i++) {
+				const result = await this.runSingleTest(testCase, testConfig)
+				results[testCase.test_id].push(result)
+			}
+		}
+		return results
+	}
+
+	/**
+	 * Runs all of the text examples asynchronously, with concurrency limit
+	 */
+	async runAllTestsParallel(
+		testCases: ProcessedTestCase[],
+		testConfig: TestConfig,
+		isVerbose: boolean,
+		maxConcurrency: number = 20,
+	): Promise<TestResultSet> {
+		const results: TestResultSet = {}
+		testCases.forEach((tc) => {
+			results[tc.test_id] = []
+		})
+
+		// Create a flat list of all individual runs we need to execute
+		const allRuns = testCases.flatMap((testCase) =>
+			Array(testConfig.number_of_runs)
+				.fill(null)
+				.map(() => testCase),
+		)
+
+		for (let i = 0; i < allRuns.length; i += maxConcurrency) {
+			const batch = allRuns.slice(i, i + maxConcurrency)
+
+			const batchPromises = batch.map((testCase) =>
+				this.runSingleTest(testCase, testConfig).then((result) => ({
+					...result,
+					test_id: testCase.test_id,
+				})),
+			)
+
+			const batchResults = await Promise.all(batchPromises)
+
+			// Calculate the total cost for this batch
+			const batchCost = batchResults.reduce((total, result) => {
+				return total + (result.streamResult?.usage?.totalCost || 0)
+			}, 0)
+
+			// Populate the results dictionary
+			for (const result of batchResults) {
+				if (result.test_id) {
+					results[result.test_id].push(result)
+				}
+			}
+
+			const batchNumber = i / maxConcurrency + 1
+			const totalBatches = Math.ceil(allRuns.length / maxConcurrency)
+			log(isVerbose, `-Completed batch ${batchNumber} of ${totalBatches}... (Batch Cost: $${batchCost.toFixed(6)})`)
+		}
+
+		return results
+	}
+
+	/**
+	 * Print output of the tests
+	 */
+	printSummary(results: TestResultSet, isVerbose: boolean) {
+		let totalRuns = 0
+		let totalPasses = 0
+		let totalInputTokens = 0
+		let totalOutputTokens = 0
+		let totalCost = 0
+		let runsWithUsageData = 0
+		let totalDiffEditSuccesses = 0
+		let totalRunsWithToolCalls = 0
+		const testCaseIds = Object.keys(results)
+
+		log(isVerbose, "\n=== TEST SUMMARY ===")
+
+		for (const testId of testCaseIds) {
+			const testResults = results[testId]
+			const passedCount = testResults.filter((r) => r.success && r.diffEditSuccess).length
+			const runCount = testResults.length
+
+			totalRuns += runCount
+			totalPasses += passedCount
+
+			const runsWithToolCalls = testResults.filter((r) => r.success === true).length
+			const diffEditSuccesses = passedCount
+			totalRunsWithToolCalls += runsWithToolCalls
+			totalDiffEditSuccesses += diffEditSuccesses
+
+			// Accumulate token and cost data
+			for (const result of testResults) {
+				if (result.streamResult?.usage) {
+					totalInputTokens += result.streamResult.usage.inputTokens
+					totalOutputTokens += result.streamResult.usage.outputTokens
+					totalCost += result.streamResult.usage.totalCost
+					runsWithUsageData++
+				}
+			}
+
+			log(isVerbose, `\n--- Test Case: ${testId} ---`)
+			log(isVerbose, `  Runs: ${runCount}`)
+			log(isVerbose, `  Passed: ${passedCount}`)
+			log(isVerbose, `  Success Rate: ${runCount > 0 ? ((passedCount / runCount) * 100).toFixed(1) : "N/A"}%`)
+		}
+
+		log(isVerbose, "\n\n=== OVERALL SUMMARY ===")
+		log(isVerbose, `Total Test Cases: ${testCaseIds.length}`)
+		log(isVerbose, `Total Runs Executed: ${totalRuns}`)
+		log(isVerbose, `Overall Passed: ${totalPasses}`)
+		log(isVerbose, `Overall Failed: ${totalRuns - totalPasses}`)
+		log(isVerbose, `Overall Success Rate: ${totalRuns > 0 ? ((totalPasses / totalRuns) * 100).toFixed(1) : "N/A"}%`)
+
+		log(isVerbose, "\n\n=== OVERALL DIFF EDIT SUCCESS RATE ===")
+		if (totalRunsWithToolCalls > 0) {
+			const diffSuccessRate = (totalDiffEditSuccesses / totalRunsWithToolCalls) * 100
+			log(isVerbose, `Total Runs with Successful Tool Calls: ${totalRunsWithToolCalls}`)
+			log(isVerbose, `Total Runs with Successful Diff Edits: ${totalDiffEditSuccesses}`)
+			log(isVerbose, `Diff Edit Success Rate: ${diffSuccessRate.toFixed(1)}%`)
+		} else {
+			log(isVerbose, "No successful tool calls to analyze for diff edit success.")
+		}
+
+		log(isVerbose, "\n\n=== TOKEN & COST ANALYSIS ===")
+		if (runsWithUsageData > 0) {
+			log(isVerbose, `Total Input Tokens: ${totalInputTokens.toLocaleString()}`)
+			log(isVerbose, `Total Output Tokens: ${totalOutputTokens.toLocaleString()}`)
+			log(isVerbose, `Total Cost: $${totalCost.toFixed(6)}`)
+			log(isVerbose, "---")
+			log(
+				isVerbose,
+				`Avg Input Tokens / Run: ${(totalInputTokens / runsWithUsageData).toLocaleString(undefined, {
+					maximumFractionDigits: 0,
+				})}`,
+			)
+			log(
+				isVerbose,
+				`Avg Output Tokens / Run: ${(totalOutputTokens / runsWithUsageData).toLocaleString(undefined, {
+					maximumFractionDigits: 0,
+				})}`,
+			)
+			log(isVerbose, `Avg Cost / Run: $${(totalCost / runsWithUsageData).toFixed(6)}`)
+		} else {
+			log(isVerbose, "No usage data available to analyze.")
+		}
+	}
+}
+
+async function main() {
+	const program = new Command()
+
+	const defaultTestPath = path.join(__dirname, "test_cases")
+	const defaultOutputPath = path.join(__dirname, "test_outputs")
+
+	program
+		.name("TestRunner")
+		.description("Run evaluation tests for diff editing")
+		.version("1.0.0")
+		.option("--test-path <path>", "Path to the directory containing test case JSON files", defaultTestPath)
+		.option("--output-path <path>", "Path to the directory to save the test output JSON files", defaultOutputPath)
+		.option("--model-id <model_id>", "The model ID to use for the test")
+		.option("--system-prompt-name <name>", "The name of the system prompt to use", "basicSystemPrompt")
+		.option("-n, --number-of-runs <number>", "Number of times to run each test case", "1")
+		.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("--parallel", "Run tests in parallel", false)
+		.option("--replay", "Run evaluation from a pre-recorded LLM output, skipping the API call", false)
+		.option("-v, --verbose", "Enable verbose logging", false)
+
+	program.parse(process.argv)
+
+	const options = program.opts()
+	const isVerbose = options.verbose
+	const testPath = options.testPath
+	const outputPath = options.outputPath
+
+	const testConfig: TestConfig = {
+		model_id: options.modelId,
+		system_prompt_name: options.systemPromptName,
+		number_of_runs: parseInt(options.numberOfRuns, 10),
+		parsing_function: options.parsingFunction,
+		diff_edit_function: options.diffEditFunction,
+		thinking_tokens_budget: parseInt(options.thinkingBudget, 10),
+		replay: options.replay,
+	}
+
+	try {
+		const startTime = Date.now()
+
+		const runner = new NodeTestRunner(testConfig.replay)
+		const testCases = runner.loadTestCases(testPath)
+
+		const processedTestCases: ProcessedTestCase[] = testCases.map((tc) => ({
+			...tc,
+			messages: runner.transformMessages(tc.messages),
+		}))
+
+		log(isVerbose, `-Loaded ${testCases.length} test cases.`)
+		log(isVerbose, `-Executing ${testConfig.number_of_runs} run(s) per test case.`)
+		if (testConfig.replay) {
+			log(isVerbose, `-Running in REPLAY mode. No API calls will be made.`)
+		}
+		log(isVerbose, "Starting tests...\n")
+
+		const results = options.parallel
+			? await runner.runAllTestsParallel(processedTestCases, testConfig, isVerbose)
+			: await runner.runAllTests(processedTestCases, testConfig, isVerbose)
+
+		runner.printSummary(results, isVerbose)
+
+		const endTime = Date.now()
+		const durationSeconds = ((endTime - startTime) / 1000).toFixed(2)
+		log(isVerbose, `\n-Total execution time: ${durationSeconds} seconds`)
+
+		runner.saveTestResults(results, outputPath)
+	} catch (error) {
+		console.error("\nError running tests:", error)
+		process.exit(1)
+	}
+}
+
+if (require.main === module) {
+	main()
+}

evals/diff_editing/diff-apply/diff-06-06-25.ts

@@ -0,0 +1,729 @@
+const SEARCH_BLOCK_START = "------- SEARCH"
+const SEARCH_BLOCK_END = "======="
+const REPLACE_BLOCK_END = "+++++++ REPLACE"
+
+const SEARCH_BLOCK_CHAR = "-"
+const REPLACE_BLOCK_CHAR = "+"
+
+/**
+ * Attempts a line-trimmed fallback match for the given search content in the original content.
+ * It tries to match `searchContent` lines against a block of lines in `originalContent` starting
+ * from `lastProcessedIndex`. Lines are matched by trimming leading/trailing whitespace and ensuring
+ * they are identical afterwards.
+ *
+ * Returns [matchIndexStart, matchIndexEnd] if found, or false if not found.
+ */
+function lineTrimmedFallbackMatch(originalContent: string, searchContent: string, startIndex: number): [number, number] | false {
+	// Split both contents into lines
+	const originalLines = originalContent.split("\n")
+	const searchLines = searchContent.split("\n")
+
+	// Trim trailing empty line if exists (from the trailing \n in searchContent)
+	if (searchLines[searchLines.length - 1] === "") {
+		searchLines.pop()
+	}
+
+	// Find the line number where startIndex falls
+	let startLineNum = 0
+	let currentIndex = 0
+	while (currentIndex < startIndex && startLineNum < originalLines.length) {
+		currentIndex += originalLines[startLineNum].length + 1 // +1 for \n
+		startLineNum++
+	}
+
+	// For each possible starting position in original content
+	for (let i = startLineNum; i <= originalLines.length - searchLines.length; i++) {
+		let matches = true
+
+		// Try to match all search lines from this position
+		for (let j = 0; j < searchLines.length; j++) {
+			const originalTrimmed = originalLines[i + j].trim()
+			const searchTrimmed = searchLines[j].trim()
+
+			if (originalTrimmed !== searchTrimmed) {
+				matches = false
+				break
+			}
+		}
+
+		// If we found a match, calculate the exact character positions
+		if (matches) {
+			// Find start character index
+			let matchStartIndex = 0
+			for (let k = 0; k < i; k++) {
+				matchStartIndex += originalLines[k].length + 1 // +1 for \n
+			}
+
+			// Find end character index
+			let matchEndIndex = matchStartIndex
+			for (let k = 0; k < searchLines.length; k++) {
+				matchEndIndex += originalLines[i + k].length + 1 // +1 for \n
+			}
+
+			return [matchStartIndex, matchEndIndex]
+		}
+	}
+
+	return false
+}
+
+/**
+ * Attempts to match blocks of code by using the first and last lines as anchors.
+ * This is a third-tier fallback strategy that helps match blocks where we can identify
+ * the correct location by matching the beginning and end, even if the exact content
+ * differs slightly.
+ *
+ * The matching strategy:
+ * 1. Only attempts to match blocks of 3 or more lines to avoid false positives
+ * 2. Extracts from the search content:
+ *    - First line as the "start anchor"
+ *    - Last line as the "end anchor"
+ * 3. For each position in the original content:
+ *    - Checks if the next line matches the start anchor
+ *    - If it does, jumps ahead by the search block size
+ *    - Checks if that line matches the end anchor
+ *    - All comparisons are done after trimming whitespace
+ *
+ * This approach is particularly useful for matching blocks of code where:
+ * - The exact content might have minor differences
+ * - The beginning and end of the block are distinctive enough to serve as anchors
+ * - The overall structure (number of lines) remains the same
+ *
+ * @param originalContent - The full content of the original file
+ * @param searchContent - The content we're trying to find in the original file
+ * @param startIndex - The character index in originalContent where to start searching
+ * @returns A tuple of [startIndex, endIndex] if a match is found, false otherwise
+ */
+function blockAnchorFallbackMatch(originalContent: string, searchContent: string, startIndex: number): [number, number] | false {
+	const originalLines = originalContent.split("\n")
+	const searchLines = searchContent.split("\n")
+
+	// Only use this approach for blocks of 3+ lines
+	if (searchLines.length < 3) {
+		return false
+	}
+
+	// Trim trailing empty line if exists
+	if (searchLines[searchLines.length - 1] === "") {
+		searchLines.pop()
+	}
+
+	const firstLineSearch = searchLines[0].trim()
+	const lastLineSearch = searchLines[searchLines.length - 1].trim()
+	const searchBlockSize = searchLines.length
+
+	// Find the line number where startIndex falls
+	let startLineNum = 0
+	let currentIndex = 0
+	while (currentIndex < startIndex && startLineNum < originalLines.length) {
+		currentIndex += originalLines[startLineNum].length + 1
+		startLineNum++
+	}
+
+	// Look for matching start and end anchors
+	for (let i = startLineNum; i <= originalLines.length - searchBlockSize; i++) {
+		// Check if first line matches
+		if (originalLines[i].trim() !== firstLineSearch) {
+			continue
+		}
+
+		// Check if last line matches at the expected position
+		if (originalLines[i + searchBlockSize - 1].trim() !== lastLineSearch) {
+			continue
+		}
+
+		// Calculate exact character positions
+		let matchStartIndex = 0
+		for (let k = 0; k < i; k++) {
+			matchStartIndex += originalLines[k].length + 1
+		}
+
+		let matchEndIndex = matchStartIndex
+		for (let k = 0; k < searchBlockSize; k++) {
+			matchEndIndex += originalLines[i + k].length + 1
+		}
+
+		return [matchStartIndex, matchEndIndex]
+	}
+
+	return false
+}
+
+/**
+ * This function reconstructs the file content by applying a streamed diff (in a
+ * specialized SEARCH/REPLACE block format) to the original file content. It is designed
+ * to handle both incremental updates and the final resulting file after all chunks have
+ * been processed.
+ *
+ * The diff format is a custom structure that uses three markers to define changes:
+ *
+ *   ------- SEARCH
+ *   [Exact content to find in the original file]
+ *   =======
+ *   [Content to replace with]
+ *   +++++++ REPLACE
+ *
+ * Behavior and Assumptions:
+ * 1. The file is processed chunk-by-chunk. Each chunk of `diffContent` may contain
+ *    partial or complete SEARCH/REPLACE blocks. By calling this function with each
+ *    incremental chunk (with `isFinal` indicating the last chunk), the final reconstructed
+ *    file content is produced.
+ *
+ * 2. Matching Strategy (in order of attempt):
+ *    a. Exact Match: First attempts to find the exact SEARCH block text in the original file
+ *    b. Line-Trimmed Match: Falls back to line-by-line comparison ignoring leading/trailing whitespace
+ *    c. Block Anchor Match: For blocks of 3+ lines, tries to match using first/last lines as anchors
+ *    If all matching strategies fail, an error is thrown.
+ *
+ * 3. Empty SEARCH Section:
+ *    - If SEARCH is empty and the original file is empty, this indicates creating a new file
+ *      (pure insertion).
+ *    - If SEARCH is empty and the original file is not empty, this indicates a complete
+ *      file replacement (the entire original content is considered matched and replaced).
+ *
+ * 4. Applying Changes:
+ *    - Before encountering the "=======" marker, lines are accumulated as search content.
+ *    - After "=======" and before ">>>>>>> REPLACE", lines are accumulated as replacement content.
+ *    - Once the block is complete (">>>>>>> REPLACE"), the matched section in the original
+ *      file is replaced with the accumulated replacement lines, and the position in the original
+ *      file is advanced.
+ *
+ * 5. Incremental Output:
+ *    - As soon as the match location is found and we are in the REPLACE section, each new
+ *      replacement line is appended to the result so that partial updates can be viewed
+ *      incrementally.
+ *
+ * 6. Partial Markers:
+ *    - If the final line of the chunk looks like it might be part of a marker but is not one
+ *      of the known markers, it is removed. This prevents incomplete or partial markers
+ *      from corrupting the output.
+ *
+ * 7. Finalization:
+ *    - Once all chunks have been processed (when `isFinal` is true), any remaining original
+ *      content after the last replaced section is appended to the result.
+ *    - Trailing newlines are not forcibly added. The code tries to output exactly what is specified.
+ *
+ * Errors:
+ * - If the search block cannot be matched using any of the available matching strategies,
+ *   an error is thrown.
+ */
+export async function constructNewFileContent(
+	diffContent: string,
+	originalContent: string,
+	isFinal: boolean,
+	version: "v1" | "v2" = "v1",
+): Promise<string> {
+	const constructor = constructNewFileContentVersionMapping[version]
+	if (!constructor) {
+		throw new Error(`Invalid version '${version}' for file content constructor`)
+	}
+	return constructor(diffContent, originalContent, isFinal)
+}
+
+const constructNewFileContentVersionMapping: Record<
+	string,
+	(diffContent: string, originalContent: string, isFinal: boolean) => Promise<string>
+> = {
+	v1: constructNewFileContentV1,
+	v2: constructNewFileContentV2,
+} as const
+
+/**
+ * @deprecated
+ */
+async function constructNewFileContentV1(diffContent: string, originalContent: string, isFinal: boolean): Promise<string> {
+	let result = ""
+	let lastProcessedIndex = 0
+
+	let currentSearchContent = ""
+	let currentReplaceContent = ""
+	let inSearch = false
+	let inReplace = false
+
+	let searchMatchIndex = -1
+	let searchEndIndex = -1
+
+	let lines = diffContent.split("\n")
+
+	// If the last line looks like a partial marker but isn't recognized,
+	// remove it because it might be incomplete.
+	const lastLine = lines[lines.length - 1]
+	if (
+		lines.length > 0 &&
+		(lastLine.startsWith(SEARCH_BLOCK_CHAR) || lastLine.startsWith("=") || lastLine.startsWith(REPLACE_BLOCK_CHAR)) &&
+		lastLine !== SEARCH_BLOCK_START &&
+		lastLine !== SEARCH_BLOCK_END &&
+		lastLine !== REPLACE_BLOCK_END
+	) {
+		lines.pop()
+	}
+
+	for (const line of lines) {
+		if (line === SEARCH_BLOCK_START) {
+			inSearch = true
+			currentSearchContent = ""
+			currentReplaceContent = ""
+			continue
+		}
+
+		if (line === SEARCH_BLOCK_END) {
+			inSearch = false
+			inReplace = true
+
+			// Remove trailing linebreak for adding the === marker
+			// if (currentSearchContent.endsWith("\r\n")) {
+			// 	currentSearchContent = currentSearchContent.slice(0, -2)
+			// } else if (currentSearchContent.endsWith("\n")) {
+			// 	currentSearchContent = currentSearchContent.slice(0, -1)
+			// }
+
+			if (!currentSearchContent) {
+				// Empty search block
+				if (originalContent.length === 0) {
+					// New file scenario: nothing to match, just start inserting
+					searchMatchIndex = 0
+					searchEndIndex = 0
+				} else {
+					// Complete file replacement scenario: treat the entire file as matched
+					searchMatchIndex = 0
+					searchEndIndex = originalContent.length
+				}
+			} else {
+				// Add check for inefficient full-file search
+				// if (currentSearchContent.trim() === originalContent.trim()) {
+				// 	throw new Error(
+				// 		"The SEARCH block contains the entire file content. Please either:\n" +
+				// 			"1. Use an empty SEARCH block to replace the entire file, or\n" +
+				// 			"2. Make focused changes to specific parts of the file that need modification.",
+				// 	)
+				// }
+
+				// Exact search match scenario
+				const exactIndex = originalContent.indexOf(currentSearchContent, lastProcessedIndex)
+				if (exactIndex !== -1) {
+					searchMatchIndex = exactIndex
+					searchEndIndex = exactIndex + currentSearchContent.length
+				} else {
+					// Attempt fallback line-trimmed matching
+					const lineMatch = lineTrimmedFallbackMatch(originalContent, currentSearchContent, lastProcessedIndex)
+					if (lineMatch) {
+						;[searchMatchIndex, searchEndIndex] = lineMatch
+					} else {
+						// Try block anchor fallback for larger blocks
+						const blockMatch = blockAnchorFallbackMatch(originalContent, currentSearchContent, lastProcessedIndex)
+						if (blockMatch) {
+							;[searchMatchIndex, searchEndIndex] = blockMatch
+						} else {
+							throw new Error(
+								`The SEARCH block:\n${currentSearchContent.trimEnd()}\n...does not match anything in the file or was searched out of order in the provided blocks.`,
+							)
+						}
+					}
+				}
+			}
+
+			// Output everything up to the match location
+			result += originalContent.slice(lastProcessedIndex, searchMatchIndex)
+			continue
+		}
+
+		if (line === REPLACE_BLOCK_END) {
+			// Finished one replace block
+
+			// // Remove the artificially added linebreak in the last line of the REPLACE block
+			// if (result.endsWith("\r\n")) {
+			// 	result = result.slice(0, -2)
+			// } else if (result.endsWith("\n")) {
+			// 	result = result.slice(0, -1)
+			// }
+
+			// Advance lastProcessedIndex to after the matched section
+			lastProcessedIndex = searchEndIndex
+
+			// Reset for next block
+			inSearch = false
+			inReplace = false
+			currentSearchContent = ""
+			currentReplaceContent = ""
+			searchMatchIndex = -1
+			searchEndIndex = -1
+			continue
+		}
+
+		// Accumulate content for search or replace
+		// (currentReplaceContent is not being used for anything right now since we directly append to result.)
+		// (We artificially add a linebreak since we split on \n at the beginning. In order to not include a trailing linebreak in the final search/result blocks we need to remove it before using them. This allows for partial line matches to be correctly identified.)
+		// NOTE: search/replace blocks must be arranged in the order they appear in the file due to how we build the content using lastProcessedIndex. We also cannot strip the trailing newline since for non-partial lines it would remove the linebreak from the original content. (If we remove end linebreak from search, then we'd also have to remove it from replace but we can't know if it's a partial line or not since the model may be using the line break to indicate the end of the block rather than as part of the search content.) We require the model to output full lines in order for our fallbacks to work as well.
+		if (inSearch) {
+			currentSearchContent += line + "\n"
+		} else if (inReplace) {
+			currentReplaceContent += line + "\n"
+			// Output replacement lines immediately if we know the insertion point
+			if (searchMatchIndex !== -1) {
+				result += line + "\n"
+			}
+		}
+	}
+
+	// If this is the final chunk, append any remaining original content
+	if (isFinal && lastProcessedIndex < originalContent.length) {
+		result += originalContent.slice(lastProcessedIndex)
+	}
+
+	return result
+}
+
+enum ProcessingState {
+	Idle = 0,
+	StateSearch = 1 << 0,
+	StateReplace = 1 << 1,
+}
+
+class NewFileContentConstructor {
+	private originalContent: string
+	private isFinal: boolean
+	private state: number
+	private pendingNonStandardLines: string[]
+	private result: string
+	private lastProcessedIndex: number
+	private currentSearchContent: string
+	private currentReplaceContent: string
+	private searchMatchIndex: number
+	private searchEndIndex: number
+
+	constructor(originalContent: string, isFinal: boolean) {
+		this.originalContent = originalContent
+		this.isFinal = isFinal
+		this.pendingNonStandardLines = []
+		this.result = ""
+		this.lastProcessedIndex = 0
+		this.state = ProcessingState.Idle
+		this.currentSearchContent = ""
+		this.currentReplaceContent = ""
+		this.searchMatchIndex = -1
+		this.searchEndIndex = -1
+	}
+
+	private resetForNextBlock() {
+		// Reset for next block
+		this.state = ProcessingState.Idle
+		this.currentSearchContent = ""
+		this.currentReplaceContent = ""
+		this.searchMatchIndex = -1
+		this.searchEndIndex = -1
+	}
+
+	private findLastMatchingLineIndex(regx: RegExp, lineLimit: number) {
+		for (let i = lineLimit; i > 0; ) {
+			i--
+			if (this.pendingNonStandardLines[i].match(regx)) {
+				return i
+			}
+		}
+		return -1
+	}
+
+	private updateProcessingState(newState: ProcessingState) {
+		const isValidTransition =
+			(this.state === ProcessingState.Idle && newState === ProcessingState.StateSearch) ||
+			(this.state === ProcessingState.StateSearch && newState === ProcessingState.StateReplace)
+
+		if (!isValidTransition) {
+			throw new Error(
+				`Invalid state transition.\n` +
+					"Valid transitions are:\n" +
+					"- Idle → StateSearch\n" +
+					"- StateSearch → StateReplace",
+			)
+		}
+
+		this.state |= newState
+	}
+
+	private isStateActive(state: ProcessingState): boolean {
+		return (this.state & state) === state
+	}
+
+	private activateReplaceState() {
+		this.updateProcessingState(ProcessingState.StateReplace)
+	}
+
+	private activateSearchState() {
+		this.updateProcessingState(ProcessingState.StateSearch)
+		this.currentSearchContent = ""
+		this.currentReplaceContent = ""
+	}
+
+	private isSearchingActive(): boolean {
+		return this.isStateActive(ProcessingState.StateSearch)
+	}
+
+	private isReplacingActive(): boolean {
+		return this.isStateActive(ProcessingState.StateReplace)
+	}
+
+	private hasPendingNonStandardLines(pendingNonStandardLineLimit: number): boolean {
+		return this.pendingNonStandardLines.length - pendingNonStandardLineLimit < this.pendingNonStandardLines.length
+	}
+
+	public processLine(line: string) {
+		this.internalProcessLine(line, true, this.pendingNonStandardLines.length)
+	}
+
+	public getResult() {
+		// If this is the final chunk, append any remaining original content
+		if (this.isFinal && this.lastProcessedIndex < this.originalContent.length) {
+			this.result += this.originalContent.slice(this.lastProcessedIndex)
+		}
+		if (this.isFinal && this.state !== ProcessingState.Idle) {
+			throw new Error("File processing incomplete - SEARCH/REPLACE operations still active during finalization")
+		}
+		return this.result
+	}
+
+	private internalProcessLine(
+		line: string,
+		canWritependingNonStandardLines: boolean,
+		pendingNonStandardLineLimit: number,
+	): number {
+		let removeLineCount = 0
+		if (line === SEARCH_BLOCK_START) {
+			removeLineCount = this.trimPendingNonStandardTrailingEmptyLines(pendingNonStandardLineLimit)
+			if (removeLineCount > 0) {
+				pendingNonStandardLineLimit = pendingNonStandardLineLimit - removeLineCount
+			}
+			if (this.hasPendingNonStandardLines(pendingNonStandardLineLimit)) {
+				this.tryFixSearchReplaceBlock(pendingNonStandardLineLimit)
+				canWritependingNonStandardLines && (this.pendingNonStandardLines.length = 0)
+			}
+			this.activateSearchState()
+		} else if (line === SEARCH_BLOCK_END) {
+			// 校验非标内容
+			if (!this.isSearchingActive()) {
+				this.tryFixSearchBlock(pendingNonStandardLineLimit)
+				canWritependingNonStandardLines && (this.pendingNonStandardLines.length = 0)
+			}
+			this.activateReplaceState()
+			this.beforeReplace()
+		} else if (line === REPLACE_BLOCK_END) {
+			if (!this.isReplacingActive()) {
+				this.tryFixReplaceBlock(pendingNonStandardLineLimit)
+				canWritependingNonStandardLines && (this.pendingNonStandardLines.length = 0)
+			}
+			this.lastProcessedIndex = this.searchEndIndex
+			this.resetForNextBlock()
+		} else {
+			// Accumulate content for search or replace
+			// (currentReplaceContent is not being used for anything right now since we directly append to result.)
+			// (We artificially add a linebreak since we split on \n at the beginning. In order to not include a trailing linebreak in the final search/result blocks we need to remove it before using them. This allows for partial line matches to be correctly identified.)
+			// NOTE: search/replace blocks must be arranged in the order they appear in the file due to how we build the content using lastProcessedIndex. We also cannot strip the trailing newline since for non-partial lines it would remove the linebreak from the original content. (If we remove end linebreak from search, then we'd also have to remove it from replace but we can't know if it's a partial line or not since the model may be using the line break to indicate the end of the block rather than as part of the search content.) We require the model to output full lines in order for our fallbacks to work as well.
+			if (this.isReplacingActive()) {
+				this.currentReplaceContent += line + "\n"
+				// Output replacement lines immediately if we know the insertion point
+				if (this.searchMatchIndex !== -1) {
+					this.result += line + "\n"
+				}
+			} else if (this.isSearchingActive()) {
+				this.currentSearchContent += line + "\n"
+			} else {
+				let appendToPendingNonStandardLines = canWritependingNonStandardLines
+				if (appendToPendingNonStandardLines) {
+					// 处理非标内容
+					this.pendingNonStandardLines.push(line)
+				}
+			}
+		}
+		return removeLineCount
+	}
+
+	private beforeReplace() {
+		// Remove trailing linebreak for adding the === marker
+		// if (currentSearchContent.endsWith("\r\n")) {
+		// 	currentSearchContent = currentSearchContent.slice(0, -2)
+		// } else if (currentSearchContent.endsWith("\n")) {
+		// 	currentSearchContent = currentSearchContent.slice(0, -1)
+		// }
+
+		if (!this.currentSearchContent) {
+			// Empty search block
+			if (this.originalContent.length === 0) {
+				// New file scenario: nothing to match, just start inserting
+				this.searchMatchIndex = 0
+				this.searchEndIndex = 0
+			} else {
+				// Complete file replacement scenario: treat the entire file as matched
+				this.searchMatchIndex = 0
+				this.searchEndIndex = this.originalContent.length
+			}
+		} else {
+			// Add check for inefficient full-file search
+			// if (currentSearchContent.trim() === originalContent.trim()) {
+			// 	throw new Error(
+			// 		"The SEARCH block contains the entire file content. Please either:\n" +
+			// 			"1. Use an empty SEARCH block to replace the entire file, or\n" +
+			// 			"2. Make focused changes to specific parts of the file that need modification.",
+			// 	)
+			// }
+			// Exact search match scenario
+			const exactIndex = this.originalContent.indexOf(this.currentSearchContent, this.lastProcessedIndex)
+			if (exactIndex !== -1) {
+				this.searchMatchIndex = exactIndex
+				this.searchEndIndex = exactIndex + this.currentSearchContent.length
+			} else {
+				// Attempt fallback line-trimmed matching
+				const lineMatch = lineTrimmedFallbackMatch(
+					this.originalContent,
+					this.currentSearchContent,
+					this.lastProcessedIndex,
+				)
+				if (lineMatch) {
+					;[this.searchMatchIndex, this.searchEndIndex] = lineMatch
+				} else {
+					// Try block anchor fallback for larger blocks
+					const blockMatch = blockAnchorFallbackMatch(
+						this.originalContent,
+						this.currentSearchContent,
+						this.lastProcessedIndex,
+					)
+					if (blockMatch) {
+						;[this.searchMatchIndex, this.searchEndIndex] = blockMatch
+					} else {
+						throw new Error(
+							`The SEARCH block:\n${this.currentSearchContent.trimEnd()}\n...does not match anything in the file.`,
+						)
+					}
+				}
+			}
+		}
+		if (this.searchMatchIndex < this.lastProcessedIndex) {
+			throw new Error(
+				`The SEARCH block:\n${this.currentSearchContent.trimEnd()}\n...matched an incorrect content in the file.`,
+			)
+		}
+		// Output everything up to the match location
+		this.result += this.originalContent.slice(this.lastProcessedIndex, this.searchMatchIndex)
+	}
+
+	private tryFixSearchBlock(lineLimit: number): number {
+		let removeLineCount = 0
+		if (lineLimit < 0) {
+			lineLimit = this.pendingNonStandardLines.length
+		}
+		if (!lineLimit) {
+			throw new Error("Invalid SEARCH/REPLACE block structure - no lines available to process")
+		}
+		let searchTagRegexp = /^[-]{3,} SEARCH$/
+		const searchTagIndex = this.findLastMatchingLineIndex(searchTagRegexp, lineLimit)
+		if (searchTagIndex !== -1) {
+			let fixLines = this.pendingNonStandardLines.slice(searchTagIndex, lineLimit)
+			fixLines[0] = SEARCH_BLOCK_START
+			for (const line of fixLines) {
+				removeLineCount += this.internalProcessLine(line, false, searchTagIndex)
+			}
+		} else {
+			throw new Error(
+				`Invalid REPLACE marker detected - could not find matching SEARCH block starting from line ${searchTagIndex + 1}`,
+			)
+		}
+		return removeLineCount
+	}
+
+	private tryFixReplaceBlock(lineLimit: number): number {
+		let removeLineCount = 0
+		if (lineLimit < 0) {
+			lineLimit = this.pendingNonStandardLines.length
+		}
+		if (!lineLimit) {
+			throw new Error()
+		}
+		let replaceBeginTagRegexp = /^[=]{3,}$/
+		const replaceBeginTagIndex = this.findLastMatchingLineIndex(replaceBeginTagRegexp, lineLimit)
+		if (replaceBeginTagIndex !== -1) {
+			// // 校验非标内容
+			// if (!this.isSearchingActive()) {
+			// 	removeLineCount += this.tryFixSearchBlock(replaceBeginTagIndex)
+			// }
+			let fixLines = this.pendingNonStandardLines.slice(replaceBeginTagIndex - removeLineCount, lineLimit - removeLineCount)
+			fixLines[0] = SEARCH_BLOCK_END
+			for (const line of fixLines) {
+				removeLineCount += this.internalProcessLine(line, false, replaceBeginTagIndex - removeLineCount)
+			}
+		} else {
+			throw new Error(`Malformed REPLACE block - missing valid separator after line ${replaceBeginTagIndex + 1}`)
+		}
+		return removeLineCount
+	}
+
+	private tryFixSearchReplaceBlock(lineLimit: number): number {
+		let removeLineCount = 0
+		if (lineLimit < 0) {
+			lineLimit = this.pendingNonStandardLines.length
+		}
+		if (!lineLimit) {
+			throw new Error()
+		}
+
+		let replaceEndTagRegexp = /^[+]{3,} REPLACE$/
+		const replaceEndTagIndex = this.findLastMatchingLineIndex(replaceEndTagRegexp, lineLimit)
+		const likeReplaceEndTag = replaceEndTagIndex === lineLimit - 1
+		if (likeReplaceEndTag) {
+			// // 校验非标内容
+			// if (!this.isReplacingActive()) {
+			// 	removeLineCount += this.tryFixReplaceBlock(replaceEndTagIndex)
+			// }
+			let fixLines = this.pendingNonStandardLines.slice(replaceEndTagIndex - removeLineCount, lineLimit - removeLineCount)
+			fixLines[fixLines.length - 1] = REPLACE_BLOCK_END
+			for (const line of fixLines) {
+				removeLineCount += this.internalProcessLine(line, false, replaceEndTagIndex - removeLineCount)
+			}
+		} else {
+			throw new Error("Malformed SEARCH/REPLACE block structure: Missing valid closing REPLACE marker")
+		}
+		return removeLineCount
+	}
+
+	/**
+	 * Removes trailing empty lines from the pendingNonStandardLines array
+	 * @param lineLimit - The index to start checking from (exclusive).
+	 *                    Removes empty lines from lineLimit-1 backwards.
+	 * @returns The number of empty lines removed
+	 */
+	private trimPendingNonStandardTrailingEmptyLines(lineLimit: number): number {
+		let removedCount = 0
+		let i = Math.min(lineLimit, this.pendingNonStandardLines.length) - 1
+
+		while (i >= 0 && this.pendingNonStandardLines[i].trim() === "") {
+			this.pendingNonStandardLines.pop()
+			removedCount++
+			i--
+		}
+
+		return removedCount
+	}
+}
+
+export async function constructNewFileContentV2(diffContent: string, originalContent: string, isFinal: boolean): Promise<string> {
+	let newFileContentConstructor = new NewFileContentConstructor(originalContent, isFinal)
+
+	let lines = diffContent.split("\n")
+
+	// If the last line looks like a partial marker but isn't recognized,
+	// remove it because it might be incomplete.
+	const lastLine = lines[lines.length - 1]
+	if (
+		lines.length > 0 &&
+		(lastLine.startsWith(SEARCH_BLOCK_CHAR) || lastLine.startsWith("=") || lastLine.startsWith(REPLACE_BLOCK_CHAR)) &&
+		lastLine !== SEARCH_BLOCK_START &&
+		lastLine !== SEARCH_BLOCK_END &&
+		lastLine !== REPLACE_BLOCK_END
+	) {
+		lines.pop()
+	}
+
+	for (const line of lines) {
+		newFileContentConstructor.processLine(line)
+	}
+
+	let result = newFileContentConstructor.getResult()
+	return result
+}

evals/diff_editing/helpers.ts

@@ -0,0 +1,25 @@
+import { Anthropic } from "@anthropic-ai/sdk"
+
+const formatImagesIntoBlocks = (images?: string[]): Anthropic.ImageBlockParam[] => {
+	return images
+		? images.map((dataUrl) => {
+				// data:image/png;base64,base64string
+				const [rest, base64] = dataUrl.split(",")
+				const mimeType = rest.split(":")[1].split(";")[0]
+				return {
+					type: "image",
+					source: {
+						type: "base64",
+						media_type: mimeType,
+						data: base64,
+					},
+				} as Anthropic.ImageBlockParam
+			})
+		: []
+}
+
+export const formatResponse = {
+	imageBlocks: (images?: string[]): Anthropic.ImageBlockParam[] => {
+		return formatImagesIntoBlocks(images)
+	},
+}

evals/diff_editing/prompts/basicSystemPrompt-06-06-25.ts

@@ -0,0 +1,615 @@
+/**
+ * Use all standard prompt values to construct prompt
+ */
+export const basicSystemPrompt = (
+	cwdFormatted: string,
+	supportsBrowserUse: boolean,
+	browserWidth: number,
+	browserHeight: number,
+	os: string,
+	shell: string,
+	homeFormatted: string,
+	mcpHubString: string,
+	userCustomInstructions: string,
+) => {
+	return `You are Cline, a highly skilled software engineer with extensive knowledge in many programming languages, frameworks, design patterns, and best practices.
+
+====
+
+TOOL USE
+
+You have access to a set of tools that are executed upon the user's approval. You can use one tool per message, and will receive the result of that tool use in the user's response. You use tools step-by-step to accomplish a given task, with each tool use informed by the result of the previous tool use.
+
+# Tool Use Formatting
+
+Tool use is formatted using XML-style tags. The tool name is enclosed in opening and closing tags, and each parameter is similarly enclosed within its own set of tags. Here's the structure:
+
+<tool_name>
+<parameter1_name>value1</parameter1_name>
+<parameter2_name>value2</parameter2_name>
+...
+</tool_name>
+
+For example:
+
+<read_file>
+<path>src/main.js</path>
+</read_file>
+
+Always adhere to this format for the tool use to ensure proper parsing and execution.
+
+# Tools
+
+## execute_command
+Description: Request to execute a CLI command on the system. Use this when you need to perform system operations or run specific commands to accomplish any step in the user's task. You must tailor your command to the user's system and provide a clear explanation of what the command does. For command chaining, use the appropriate chaining syntax for the user's shell. Prefer to execute complex CLI commands over creating executable scripts, as they are more flexible and easier to run. Commands will be executed in the current working directory: ${cwdFormatted}
+Parameters:
+- command: (required) The CLI command to execute. This should be valid for the current operating system. Ensure the command is properly formatted and does not contain any harmful instructions.
+- requires_approval: (required) A boolean indicating whether this command requires explicit user approval before execution in case the user has auto-approve mode enabled. Set to 'true' for potentially impactful operations like installing/uninstalling packages, deleting/overwriting files, system configuration changes, network operations, or any commands that could have unintended side effects. Set to 'false' for safe operations like reading files/directories, running development servers, building projects, and other non-destructive operations.
+Usage:
+<execute_command>
+<command>Your command here</command>
+<requires_approval>true or false</requires_approval>
+</execute_command>
+
+## read_file
+Description: Request to read the contents of a file at the specified path. Use this when you need to examine the contents of an existing file you do not know the contents of, for example to analyze code, review text files, or extract information from configuration files. Automatically extracts raw text from PDF and DOCX files. May not be suitable for other types of binary files, as it returns the raw content as a string.
+Parameters:
+- path: (required) The path of the file to read (relative to the current working directory ${cwdFormatted})
+Usage:
+<read_file>
+<path>File path here</path>
+</read_file>
+
+## write_to_file
+Description: Request to write content to a file at the specified path. If the file exists, it will be overwritten with the provided content. If the file doesn't exist, it will be created. This tool will automatically create any directories needed to write the file.
+Parameters:
+- path: (required) The path of the file to write to (relative to the current working directory ${cwdFormatted})
+- content: (required) The content to write to the file. ALWAYS provide the COMPLETE intended content of the file, without any truncation or omissions. You MUST include ALL parts of the file, even if they haven't been modified.
+Usage:
+<write_to_file>
+<path>File path here</path>
+<content>
+Your file content here
+</content>
+</write_to_file>
+
+## replace_in_file
+Description: Request to replace sections of content in an existing file using SEARCH/REPLACE blocks that define exact changes to specific parts of the file. This tool should be used when you need to make targeted changes to specific parts of a file.
+Parameters:
+- path: (required) The path of the file to modify (relative to the current working directory ${cwdFormatted})
+- diff: (required) One or more SEARCH/REPLACE blocks following this exact format:
+  \`\`\`
+  ------- SEARCH
+  [exact content to find]
+  =======
+  [new content to replace with]
+  +++++++ REPLACE
+  \`\`\`
+  Critical rules:
+  1. SEARCH content must match the associated file section to find EXACTLY:
+     * Match character-for-character including whitespace, indentation, line endings
+     * Include all comments, docstrings, etc.
+  2. SEARCH/REPLACE blocks will ONLY replace the first match occurrence.
+     * Including multiple unique SEARCH/REPLACE blocks if you need to make multiple changes.
+     * Include *just* enough lines in each SEARCH section to uniquely match each set of lines that need to change.
+     * When using multiple SEARCH/REPLACE blocks, list them in the order they appear in the file.
+  3. Keep SEARCH/REPLACE blocks concise:
+     * Break large SEARCH/REPLACE blocks into a series of smaller blocks that each change a small portion of the file.
+     * Include just the changing lines, and a few surrounding lines if needed for uniqueness.
+     * Do not include long runs of unchanging lines in SEARCH/REPLACE blocks.
+     * Each line must be complete. Never truncate lines mid-way through as this can cause matching failures.
+  4. Special operations:
+     * To move code: Use two SEARCH/REPLACE blocks (one to delete from original + one to insert at new location)
+     * To delete code: Use empty REPLACE section
+Usage:
+<replace_in_file>
+<path>File path here</path>
+<diff>
+Search and replace blocks here
+</diff> 
+</replace_in_file>
+
+
+## search_files
+Description: Request to perform a regex search across files in a specified directory, providing context-rich results. This tool searches for patterns or specific content across multiple files, displaying each match with encapsulating context.
+Parameters:
+- path: (required) The path of the directory to search in (relative to the current working directory ${cwdFormatted}). This directory will be recursively searched.
+- regex: (required) The regular expression pattern to search for. Uses Rust regex syntax.
+- file_pattern: (optional) Glob pattern to filter files (e.g., '*.ts' for TypeScript files). If not provided, it will search all files (*).
+Usage:
+<search_files>
+<path>Directory path here</path>
+<regex>Your regex pattern here</regex>
+<file_pattern>file pattern here (optional)</file_pattern>
+</search_files>
+
+## list_files
+Description: Request to list files and directories within the specified directory. If recursive is true, it will list all files and directories recursively. If recursive is false or not provided, it will only list the top-level contents. Do not use this tool to confirm the existence of files you may have created, as the user will let you know if the files were created successfully or not.
+Parameters:
+- path: (required) The path of the directory to list contents for (relative to the current working directory ${cwdFormatted})
+- recursive: (optional) Whether to list files recursively. Use true for recursive listing, false or omit for top-level only.
+Usage:
+<list_files>
+<path>Directory path here</path>
+<recursive>true or false (optional)</recursive>
+</list_files>
+
+## list_code_definition_names
+Description: Request to list definition names (classes, functions, methods, etc.) used in source code files at the top level of the specified directory. This tool provides insights into the codebase structure and important constructs, encapsulating high-level concepts and relationships that are crucial for understanding the overall architecture.
+Parameters:
+- path: (required) The path of the directory (relative to the current working directory ${cwdFormatted}) to list top level source code definitions for.
+Usage:
+<list_code_definition_names>
+<path>Directory path here</path>
+</list_code_definition_names>${
+		supportsBrowserUse
+			? `
+
+## browser_action
+Description: Request to interact with a Puppeteer-controlled browser. Every action, except \`close\`, will be responded to with a screenshot of the browser's current state, along with any new console logs. You may only perform one browser action per message, and wait for the user's response including a screenshot and logs to determine the next action.
+- The sequence of actions **must always start with** launching the browser at a URL, and **must always end with** closing the browser. If you need to visit a new URL that is not possible to navigate to from the current webpage, you must first close the browser, then launch again at the new URL.
+- While the browser is active, only the \`browser_action\` tool can be used. No other tools should be called during this time. You may proceed to use other tools only after closing the browser. For example if you run into an error and need to fix a file, you must close the browser, then use other tools to make the necessary changes, then re-launch the browser to verify the result.
+- The browser window has a resolution of **${browserWidth}x${browserHeight}** pixels. When performing any click actions, ensure the coordinates are within this resolution range.
+- Before clicking on any elements such as icons, links, or buttons, you must consult the provided screenshot of the page to determine the coordinates of the element. The click should be targeted at the **center of the element**, not on its edges.
+Parameters:
+- action: (required) The action to perform. The available actions are:
+    * launch: Launch a new Puppeteer-controlled browser instance at the specified URL. This **must always be the first action**.
+        - Use with the \`url\` parameter to provide the URL.
+        - Ensure the URL is valid and includes the appropriate protocol (e.g. http://localhost:3000/page, file:///path/to/file.html, etc.)
+    * click: Click at a specific x,y coordinate.
+        - Use with the \`coordinate\` parameter to specify the location.
+        - Always click in the center of an element (icon, button, link, etc.) based on coordinates derived from a screenshot.
+    * type: Type a string of text on the keyboard. You might use this after clicking on a text field to input text.
+        - Use with the \`text\` parameter to provide the string to type.
+    * scroll_down: Scroll down the page by one page height.
+    * scroll_up: Scroll up the page by one page height.
+    * close: Close the Puppeteer-controlled browser instance. This **must always be the final browser action**.
+        - Example: \`<action>close</action>\`
+- url: (optional) Use this for providing the URL for the \`launch\` action.
+    * Example: <url>https://example.com</url>
+- coordinate: (optional) The X and Y coordinates for the \`click\` action. Coordinates should be within the **${browserWidth}x${browserHeight}** resolution.
+    * Example: <coordinate>450,300</coordinate>
+- text: (optional) Use this for providing the text for the \`type\` action.
+    * Example: <text>Hello, world!</text>
+Usage:
+<browser_action>
+<action>Action to perform (e.g., launch, click, type, scroll_down, scroll_up, close)</action>
+<url>URL to launch the browser at (optional)</url>
+<coordinate>x,y coordinates (optional)</coordinate>
+<text>Text to type (optional)</text>
+</browser_action>`
+			: ""
+	}
+
+## use_mcp_tool
+Description: Request to use a tool provided by a connected MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.
+Parameters:
+- server_name: (required) The name of the MCP server providing the tool
+- tool_name: (required) The name of the tool to execute
+- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema
+Usage:
+<use_mcp_tool>
+<server_name>server name here</server_name>
+<tool_name>tool name here</tool_name>
+<arguments>
+{
+  "param1": "value1",
+  "param2": "value2"
+}
+</arguments>
+</use_mcp_tool>
+
+## access_mcp_resource
+Description: Request to access a resource provided by a connected MCP server. Resources represent data sources that can be used as context, such as files, API responses, or system information.
+Parameters:
+- server_name: (required) The name of the MCP server providing the resource
+- uri: (required) The URI identifying the specific resource to access
+Usage:
+<access_mcp_resource>
+<server_name>server name here</server_name>
+<uri>resource URI here</uri>
+</access_mcp_resource>
+
+## ask_followup_question
+Description: Ask the user a question to gather additional information needed to complete the task. This tool should be used when you encounter ambiguities, need clarification, or require more details to proceed effectively. It allows for interactive problem-solving by enabling direct communication with the user. Use this tool judiciously to maintain a balance between gathering necessary information and avoiding excessive back-and-forth.
+Parameters:
+- question: (required) The question to ask the user. This should be a clear, specific question that addresses the information you need.
+- options: (optional) An array of 2-5 options for the user to choose from. Each option should be a string describing a possible answer. You may not always need to provide options, but it may be helpful in many cases where it can save the user from having to type out a response manually. IMPORTANT: NEVER include an option to toggle to Act mode, as this would be something you need to direct the user to do manually themselves if needed.
+Usage:
+<ask_followup_question>
+<question>Your question here</question>
+<options>
+Array of options here (optional), e.g. ["Option 1", "Option 2", "Option 3"]
+</options>
+</ask_followup_question>
+
+## attempt_completion
+Description: After each tool use, the user will respond with the result of that tool use, i.e. if it succeeded or failed, along with any reasons for failure. Once you've received the results of tool uses and can confirm that the task is complete, use this tool to present the result of your work to the user. Optionally you may provide a CLI command to showcase the result of your work. The user may respond with feedback if they are not satisfied with the result, which you can use to make improvements and try again.
+IMPORTANT NOTE: This tool CANNOT be used until you've confirmed from the user that any previous tool uses were successful. Failure to do so will result in code corruption and system failure. Before using this tool, you must ask yourself in <thinking></thinking> tags if you've confirmed from the user that any previous tool uses were successful. If not, then DO NOT use this tool.
+Parameters:
+- result: (required) The result of the task. Formulate this result in a way that is final and does not require further input from the user. Don't end your result with questions or offers for further assistance.
+- command: (optional) A CLI command to execute to show a live demo of the result to the user. For example, use \`open index.html\` to display a created html website, or \`open localhost:3000\` to display a locally running development server. But DO NOT use commands like \`echo\` or \`cat\` that merely print text. This command should be valid for the current operating system. Ensure the command is properly formatted and does not contain any harmful instructions.
+Usage:
+<attempt_completion>
+<result>
+Your final result description here
+</result>
+<command>Command to demonstrate result (optional)</command>
+</attempt_completion>
+
+## new_task
+Description: Request to create a new task with preloaded context covering the conversation with the user up to this point and key information for continuing with the new task. With this tool, you will create a detailed summary of the conversation so far, paying close attention to the user's explicit requests and your previous actions, with a focus on the most relevant information required for the new task.
+Among other important areas of focus, this summary should be thorough in capturing technical details, code patterns, and architectural decisions that would be essential for continuing with the new task. The user will be presented with a preview of your generated context and can choose to create a new task or keep chatting in the current conversation. The user may choose to start a new task at any point.
+Parameters:
+- Context: (required) The context to preload the new task with. If applicable based on the current task, this should include:
+  1. Current Work: Describe in detail what was being worked on prior to this request to create a new task. Pay special attention to the more recent messages / conversation.
+  2. Key Technical Concepts: List all important technical concepts, technologies, coding conventions, and frameworks discussed, which might be relevant for the new task.
+  3. Relevant Files and Code: If applicable, enumerate specific files and code sections examined, modified, or created for the task continuation. Pay special attention to the most recent messages and changes.
+  4. Problem Solving: Document problems solved thus far and any ongoing troubleshooting efforts.
+  5. Pending Tasks and Next Steps: Outline all pending tasks that you have explicitly been asked to work on, as well as list the next steps you will take for all outstanding work, if applicable. Include code snippets where they add clarity. For any next steps, include direct quotes from the most recent conversation showing exactly what task you were working on and where you left off. This should be verbatim to ensure there's no information loss in context between tasks. It's important to be detailed here.
+Usage:
+<new_task>
+<context>context to preload new task with</context>
+</new_task>
+
+## plan_mode_respond
+Description: Respond to the user's inquiry in an effort to plan a solution to the user's task. This tool should be used when you need to provide a response to a question or statement from the user about how you plan to accomplish the task. This tool is only available in PLAN MODE. The environment_details will specify the current mode, if it is not PLAN MODE then you should not use this tool. Depending on the user's message, you may ask questions to get clarification about the user's request, architect a solution to the task, and to brainstorm ideas with the user. For example, if the user's task is to create a website, you may start by asking some clarifying questions, then present a detailed plan for how you will accomplish the task given the context, and perhaps engage in a back and forth to finalize the details before the user switches you to ACT MODE to implement the solution.
+Parameters:
+- response: (required) The response to provide to the user. Do not try to use tools in this parameter, this is simply a chat response. (You MUST use the response parameter, do not simply place the response text directly within <plan_mode_respond> tags.)
+Usage:
+<plan_mode_respond>
+<response>Your response here</response>
+</plan_mode_respond>
+
+## load_mcp_documentation
+Description: Load documentation about creating MCP servers. This tool should be used when the user requests to create or install an MCP server (the user may ask you something along the lines of "add a tool" that does some function, in other words to create an MCP server that provides tools and resources that may connect to external APIs for example. You have the ability to create an MCP server and add it to a configuration file that will then expose the tools and resources for you to use with \`use_mcp_tool\` and \`access_mcp_resource\`). The documentation provides detailed information about the MCP server creation process, including setup instructions, best practices, and examples.
+Parameters: None
+Usage:
+<load_mcp_documentation>
+</load_mcp_documentation>
+
+# Tool Use Examples
+
+## Example 1: Requesting to execute a command
+
+<execute_command>
+<command>npm run dev</command>
+<requires_approval>false</requires_approval>
+</execute_command>
+
+## Example 2: Requesting to create a new file
+
+<write_to_file>
+<path>src/frontend-config.json</path>
+<content>
+{
+  "apiEndpoint": "https://api.example.com",
+  "theme": {
+    "primaryColor": "#007bff",
+    "secondaryColor": "#6c757d",
+    "fontFamily": "Arial, sans-serif"
+  },
+  "features": {
+    "darkMode": true,
+    "notifications": true,
+    "analytics": false
+  },
+  "version": "1.0.0"
+}
+</content>
+</write_to_file>
+
+## Example 3: Creating a new task
+
+<new_task>
+<context>
+1. Current Work:
+   [Detailed description]
+
+2. Key Technical Concepts:
+   - [Concept 1]
+   - [Concept 2]
+   - [...]
+
+3. Relevant Files and Code:
+   - [File Name 1]
+      - [Summary of why this file is important]
+      - [Summary of the changes made to this file, if any]
+      - [Important Code Snippet]
+   - [File Name 2]
+      - [Important Code Snippet]
+   - [...]
+
+4. Problem Solving:
+   [Detailed description]
+
+5. Pending Tasks and Next Steps:
+   - [Task 1 details & next steps]
+   - [Task 2 details & next steps]
+   - [...]
+</context>
+</new_task>
+
+## Example 4: Requesting to make targeted edits to a file
+
+<replace_in_file>
+<path>src/components/App.tsx</path>
+<diff>
+------- SEARCH
+import React from 'react';
+=======
+import React, { useState } from 'react';
++++++++ REPLACE
+
+------- SEARCH
+function handleSubmit() {
+  saveData();
+  setLoading(false);
+}
+
+=======
++++++++ REPLACE
+
+------- SEARCH
+return (
+  <div>
+=======
+function handleSubmit() {
+  saveData();
+  setLoading(false);
+}
+
+return (
+  <div>
++++++++ REPLACE
+</diff>
+</replace_in_file>
+
+
+## Example 5: Requesting to use an MCP tool
+
+<use_mcp_tool>
+<server_name>weather-server</server_name>
+<tool_name>get_forecast</tool_name>
+<arguments>
+{
+  "city": "San Francisco",
+  "days": 5
+}
+</arguments>
+</use_mcp_tool>
+
+## Example 6: Another example of using an MCP tool (where the server name is a unique identifier such as a URL)
+
+<use_mcp_tool>
+<server_name>github.com/modelcontextprotocol/servers/tree/main/src/github</server_name>
+<tool_name>create_issue</tool_name>
+<arguments>
+{
+  "owner": "octocat",
+  "repo": "hello-world",
+  "title": "Found a bug",
+  "body": "I'm having a problem with this.",
+  "labels": ["bug", "help wanted"],
+  "assignees": ["octocat"]
+}
+</arguments>
+</use_mcp_tool>
+
+# Tool Use Guidelines
+
+1. In <thinking> tags, assess what information you already have and what information you need to proceed with the task.
+2. Choose the most appropriate tool based on the task and the tool descriptions provided. Assess if you need additional information to proceed, and which of the available tools would be most effective for gathering this information. For example using the list_files tool is more effective than running a command like \`ls\` in the terminal. It's critical that you think about each available tool and use the one that best fits the current step in the task.
+3. If multiple actions are needed, use one tool at a time per message to accomplish the task iteratively, with each tool use being informed by the result of the previous tool use. Do not assume the outcome of any tool use. Each step must be informed by the previous step's result.
+4. Formulate your tool use using the XML format specified for each tool.
+5. After each tool use, the user will respond with the result of that tool use. This result will provide you with the necessary information to continue your task or make further decisions. This response may include:
+  - Information about whether the tool succeeded or failed, along with any reasons for failure.
+  - Linter errors that may have arisen due to the changes you made, which you'll need to address.
+  - New terminal output in reaction to the changes, which you may need to consider or act upon.
+  - Any other relevant feedback or information related to the tool use.
+6. ALWAYS wait for user confirmation after each tool use before proceeding. Never assume the success of a tool use without explicit confirmation of the result from the user.
+
+It is crucial to proceed step-by-step, waiting for the user's message after each tool use before moving forward with the task. This approach allows you to:
+1. Confirm the success of each step before proceeding.
+2. Address any issues or errors that arise immediately.
+3. Adapt your approach based on new information or unexpected results.
+4. Ensure that each action builds correctly on the previous ones.
+
+By waiting for and carefully considering the user's response after each tool use, you can react accordingly and make informed decisions about how to proceed with the task. This iterative process helps ensure the overall success and accuracy of your work.
+
+====
+
+MCP SERVERS
+
+The Model Context Protocol (MCP) enables communication between the system and locally running MCP servers that provide additional tools and resources to extend your capabilities.
+
+# Connected MCP Servers
+
+When a server is connected, you can use the server's tools via the \`use_mcp_tool\` tool, and access the server's resources via the \`access_mcp_resource\` tool.
+
+${mcpHubString}
+
+====
+
+EDITING FILES
+
+You have access to two tools for working with files: **write_to_file** and **replace_in_file**. Understanding their roles and selecting the right one for the job will help ensure efficient and accurate modifications.
+
+# write_to_file
+
+## Purpose
+
+- Create a new file, or overwrite the entire contents of an existing file.
+
+## When to Use
+
+- Initial file creation, such as when scaffolding a new project.  
+- Overwriting large boilerplate files where you want to replace the entire content at once.
+- When the complexity or number of changes would make replace_in_file unwieldy or error-prone.
+- When you need to completely restructure a file's content or change its fundamental organization.
+
+## Important Considerations
+
+- Using write_to_file requires providing the file's complete final content.  
+- If you only need to make small changes to an existing file, consider using replace_in_file instead to avoid unnecessarily rewriting the entire file.
+- While write_to_file should not be your default choice, don't hesitate to use it when the situation truly calls for it.
+
+# replace_in_file
+
+## Purpose
+
+- Make targeted edits to specific parts of an existing file without overwriting the entire file.
+
+## When to Use
+
+- Small, localized changes like updating a few lines, function implementations, changing variable names, modifying a section of text, etc.
+- Targeted improvements where only specific portions of the file's content needs to be altered.
+- Especially useful for long files where much of the file will remain unchanged.
+
+## Advantages
+
+- More efficient for minor edits, since you don't need to supply the entire file content.  
+- Reduces the chance of errors that can occur when overwriting large files.
+
+# Choosing the Appropriate Tool
+
+- **Default to replace_in_file** for most changes. It's the safer, more precise option that minimizes potential issues.
+- **Use write_to_file** when:
+  - Creating new files
+  - The changes are so extensive that using replace_in_file would be more complex or risky
+  - You need to completely reorganize or restructure a file
+  - The file is relatively small and the changes affect most of its content
+  - You're generating boilerplate or template files
+
+# Auto-formatting Considerations
+
+- After using either write_to_file or replace_in_file, the user's editor may automatically format the file
+- This auto-formatting may modify the file contents, for example:
+  - Breaking single lines into multiple lines
+  - Adjusting indentation to match project style (e.g. 2 spaces vs 4 spaces vs tabs)
+  - Converting single quotes to double quotes (or vice versa based on project preferences)
+  - Organizing imports (e.g. sorting, grouping by type)
+  - Adding/removing trailing commas in objects and arrays
+  - Enforcing consistent brace style (e.g. same-line vs new-line)
+  - Standardizing semicolon usage (adding or removing based on style)
+- The write_to_file and replace_in_file tool responses will include the final state of the file after any auto-formatting
+- Use this final state as your reference point for any subsequent edits. This is ESPECIALLY important when crafting SEARCH blocks for replace_in_file which require the content to match what's in the file exactly.
+
+# Workflow Tips
+
+1. Before editing, assess the scope of your changes and decide which tool to use.
+2. For targeted edits, apply replace_in_file with carefully crafted SEARCH/REPLACE blocks. If you need multiple changes, you can stack multiple SEARCH/REPLACE blocks within a single replace_in_file call.
+3. For major overhauls or initial file creation, rely on write_to_file.
+4. Once the file has been edited with either write_to_file or replace_in_file, the system will provide you with the final state of the modified file. Use this updated content as the reference point for any subsequent SEARCH/REPLACE operations, since it reflects any auto-formatting or user-applied changes.
+By thoughtfully selecting between write_to_file and replace_in_file, you can make your file editing process smoother, safer, and more efficient.
+
+====
+ 
+ACT MODE V.S. PLAN MODE
+
+In each user message, the environment_details will specify the current mode. There are two modes:
+
+- ACT MODE: In this mode, you have access to all tools EXCEPT the plan_mode_respond tool.
+ - In ACT MODE, you use tools to accomplish the user's task. Once you've completed the user's task, you use the attempt_completion tool to present the result of the task to the user.
+- PLAN MODE: In this special mode, you have access to the plan_mode_respond tool.
+ - In PLAN MODE, the goal is to gather information and get context to create a detailed plan for accomplishing the task, which the user will review and approve before they switch you to ACT MODE to implement the solution.
+ - In PLAN MODE, when you need to converse with the user or present a plan, you should use the plan_mode_respond tool to deliver your response directly, rather than using <thinking> tags to analyze when to respond. Do not talk about using plan_mode_respond - just use it directly to share your thoughts and provide helpful answers.
+
+## What is PLAN MODE?
+
+- While you are usually in ACT MODE, the user may switch to PLAN MODE in order to have a back and forth with you to plan how to best accomplish the task. 
+- When starting in PLAN MODE, depending on the user's request, you may need to do some information gathering e.g. using read_file or search_files to get more context about the task. You may also ask the user clarifying questions to get a better understanding of the task. You may return mermaid diagrams to visually display your understanding.
+- Once you've gained more context about the user's request, you should architect a detailed plan for how you will accomplish the task. Returning mermaid diagrams may be helpful here as well.
+- Then you might ask the user if they are pleased with this plan, or if they would like to make any changes. Think of this as a brainstorming session where you can discuss the task and plan the best way to accomplish it.
+- If at any point a mermaid diagram would make your plan clearer to help the user quickly see the structure, you are encouraged to include a Mermaid code block in the response. (Note: if you use colors in your mermaid diagrams, be sure to use high contrast colors so the text is readable.)
+- Finally once it seems like you've reached a good plan, ask the user to switch you back to ACT MODE to implement the solution.
+
+====
+ 
+CAPABILITIES
+
+- You have access to tools that let you execute CLI commands on the user's computer, list files, view source code definitions, regex search${
+		supportsBrowserUse ? ", use the browser" : ""
+	}, read and edit files, and ask follow-up questions. These tools help you effectively accomplish a wide range of tasks, such as writing code, making edits or improvements to existing files, understanding the current state of a project, performing system operations, and much more.
+- When the user initially gives you a task, a recursive list of all filepaths in the current working directory ('${cwdFormatted}') will be included in environment_details. This provides an overview of the project's file structure, offering key insights into the project from directory/file names (how developers conceptualize and organize their code) and file extensions (the language used). This can also guide decision-making on which files to explore further. If you need to further explore directories such as outside the current working directory, you can use the list_files tool. If you pass 'true' for the recursive parameter, it will list files recursively. Otherwise, it will list files at the top level, which is better suited for generic directories where you don't necessarily need the nested structure, like the Desktop.
+- You can use search_files to perform regex searches across files in a specified directory, outputting context-rich results that include surrounding lines. This is particularly useful for understanding code patterns, finding specific implementations, or identifying areas that need refactoring.
+- You can use the list_code_definition_names tool to get an overview of source code definitions for all files at the top level of a specified directory. This can be particularly useful when you need to understand the broader context and relationships between certain parts of the code. You may need to call this tool multiple times to understand various parts of the codebase related to the task.
+	- For example, when asked to make edits or improvements you might analyze the file structure in the initial environment_details to get an overview of the project, then use list_code_definition_names to get further insight using source code definitions for files located in relevant directories, then read_file to examine the contents of relevant files, analyze the code and suggest improvements or make necessary edits, then use the replace_in_file tool to implement changes. If you refactored code that could affect other parts of the codebase, you could use search_files to ensure you update other files as needed.
+- You can use the execute_command tool to run commands on the user's computer whenever you feel it can help accomplish the user's task. When you need to execute a CLI command, you must provide a clear explanation of what the command does. Prefer to execute complex CLI commands over creating executable scripts, since they are more flexible and easier to run. Interactive and long-running commands are allowed, since the commands are run in the user's VSCode terminal. The user may keep commands running in the background and you will be kept updated on their status along the way. Each command you execute is run in a new terminal instance.${
+		supportsBrowserUse
+			? "\n- You can use the browser_action tool to interact with websites (including html files and locally running development servers) through a Puppeteer-controlled browser when you feel it is necessary in accomplishing the user's task. This tool is particularly useful for web development tasks as it allows you to launch a browser, navigate to pages, interact with elements through clicks and keyboard input, and capture the results through screenshots and console logs. This tool may be useful at key stages of web development tasks-such as after implementing new features, making substantial changes, when troubleshooting issues, or to verify the result of your work. You can analyze the provided screenshots to ensure correct rendering or identify errors, and review console logs for runtime issues.\n	- For example, if asked to add a component to a react website, you might create the necessary files, use execute_command to run the site locally, then use browser_action to launch the browser, navigate to the local server, and verify the component renders & functions correctly before closing the browser."
+			: ""
+	}
+- You have access to MCP servers that may provide additional tools and resources. Each server may provide different capabilities that you can use to accomplish tasks more effectively.
+- You can use LaTeX syntax in your responses to render mathematical expressions
+
+====
+
+RULES
+
+- Your current working directory is: ${cwdFormatted}
+- You cannot \`cd\` into a different directory to complete a task. You are stuck operating from '${cwdFormatted}', so be sure to pass in the correct 'path' parameter when using tools that require a path.
+- Do not use the ~ character or $HOME to refer to the home directory.
+- Before using the execute_command tool, you must first think about the SYSTEM INFORMATION context provided to understand the user's environment and tailor your commands to ensure they are compatible with their system. You must also consider if the command you need to run should be executed in a specific directory outside of the current working directory '${cwdFormatted}', and if so prepend with \`cd\`'ing into that directory && then executing the command (as one command since you are stuck operating from '${cwdFormatted}'). For example, if you needed to run \`npm install\` in a project outside of '${cwdFormatted}', you would need to prepend with a \`cd\` i.e. pseudocode for this would be \`cd (path to project) && (command, in this case npm install)\`.
+- When using the search_files tool, craft your regex patterns carefully to balance specificity and flexibility. Based on the user's task you may use it to find code patterns, TODO comments, function definitions, or any text-based information across the project. The results include context, so analyze the surrounding code to better understand the matches. Leverage the search_files tool in combination with other tools for more comprehensive analysis. For example, use it to find specific code patterns, then use read_file to examine the full context of interesting matches before using replace_in_file to make informed changes.
+- When creating a new project (such as an app, website, or any software project), organize all new files within a dedicated project directory unless the user specifies otherwise. Use appropriate file paths when creating files, as the write_to_file tool will automatically create any necessary directories. Structure the project logically, adhering to best practices for the specific type of project being created. Unless otherwise specified, new projects should be easily run without additional setup, for example most projects can be built in HTML, CSS, and JavaScript - which you can open in a browser.
+- Be sure to consider the type of project (e.g. Python, JavaScript, web application) when determining the appropriate structure and files to include. Also consider what files may be most relevant to accomplishing the task, for example looking at a project's manifest file would help you understand the project's dependencies, which you could incorporate into any code you write.
+- When making changes to code, always consider the context in which the code is being used. Ensure that your changes are compatible with the existing codebase and that they follow the project's coding standards and best practices.
+- When you want to modify a file, use the replace_in_file or write_to_file tool directly with the desired changes. You do not need to display the changes before using the tool.
+- Do not ask for more information than necessary. Use the tools provided to accomplish the user's request efficiently and effectively. When you've completed your task, you must use the attempt_completion tool to present the result to the user. The user may provide feedback, which you can use to make improvements and try again.
+- You are only allowed to ask the user questions using the ask_followup_question tool. Use this tool only when you need additional details to complete a task, and be sure to use a clear and concise question that will help you move forward with the task. However if you can use the available tools to avoid having to ask the user questions, you should do so. For example, if the user mentions a file that may be in an outside directory like the Desktop, you should use the list_files tool to list the files in the Desktop and check if the file they are talking about is there, rather than asking the user to provide the file path themselves.
+- When executing commands, if you don't see the expected output, assume the terminal executed the command successfully and proceed with the task. The user's terminal may be unable to stream the output back properly. If you absolutely need to see the actual terminal output, use the ask_followup_question tool to request the user to copy and paste it back to you.
+- The user may provide a file's contents directly in their message, in which case you shouldn't use the read_file tool to get the file contents again since you already have it.
+- Your goal is to try to accomplish the user's task, NOT engage in a back and forth conversation.${
+		supportsBrowserUse
+			? `\n- The user may ask generic non-development tasks, such as "what\'s the latest news" or "look up the weather in San Diego", in which case you might use the browser_action tool to complete the task if it makes sense to do so, rather than trying to create a website or using curl to answer the question. However, if an available MCP server tool or resource can be used instead, you should prefer to use it over browser_action.`
+			: ""
+	}
+- NEVER end attempt_completion result with a question or request to engage in further conversation! Formulate the end of your result in a way that is final and does not require further input from the user.
+- You are STRICTLY FORBIDDEN from starting your messages with "Great", "Certainly", "Okay", "Sure". You should NOT be conversational in your responses, but rather direct and to the point. For example you should NOT say "Great, I've updated the CSS" but instead something like "I've updated the CSS". It is important you be clear and technical in your messages.
+- When presented with images, utilize your vision capabilities to thoroughly examine them and extract meaningful information. Incorporate these insights into your thought process as you accomplish the user's task.
+- At the end of each user message, you will automatically receive environment_details. This information is not written by the user themselves, but is auto-generated to provide potentially relevant context about the project structure and environment. While this information can be valuable for understanding the project context, do not treat it as a direct part of the user's request or response. Use it to inform your actions and decisions, but don't assume the user is explicitly asking about or referring to this information unless they clearly do so in their message. When using environment_details, explain your actions clearly to ensure the user understands, as they may not be aware of these details.
+- Before executing commands, check the "Actively Running Terminals" section in environment_details. If present, consider how these active processes might impact your task. For example, if a local development server is already running, you wouldn't need to start it again. If no active terminals are listed, proceed with command execution as normal.
+- When using the replace_in_file tool, you must include complete lines in your SEARCH blocks, not partial lines. The system requires exact line matches and cannot match partial lines. For example, if you want to match a line containing "const x = 5;", your SEARCH block must include the entire line, not just "x = 5" or other fragments.
+- When using the replace_in_file tool, if you use multiple SEARCH/REPLACE blocks, list them in the order they appear in the file. For example if you need to make changes to both line 10 and line 50, first include the SEARCH/REPLACE block for line 10, followed by the SEARCH/REPLACE block for line 50.
+- When using the replace_in_file tool, Do NOT add extra characters to the markers (e.g., ------- SEARCH> is INVALID). Do NOT forget to use the closing +++++++ REPLACE marker. Do NOT modify the marker format in any way. Malformed XML will cause complete tool failure and break the entire editing process.
+- It is critical you wait for the user's response after each tool use, in order to confirm the success of the tool use. For example, if asked to make a todo app, you would create a file, wait for the user's response it was created successfully, then create another file if needed, wait for the user's response it was created successfully, etc.${
+		supportsBrowserUse
+			? " Then if you want to test your work, you might use browser_action to launch the site, wait for the user's response confirming the site was launched along with a screenshot, then perhaps e.g., click a button to test functionality if needed, wait for the user's response confirming the button was clicked along with a screenshot of the new state, before finally closing the browser."
+			: ""
+	}
+- MCP operations should be used one at a time, similar to other tool usage. Wait for confirmation of success before proceeding with additional operations.
+
+====
+
+SYSTEM INFORMATION
+
+Operating System: ${os}
+Default Shell: ${shell}
+Home Directory: ${homeFormatted}
+Current Working Directory: ${cwdFormatted}
+
+====
+
+OBJECTIVE
+
+You accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.
+
+1. Analyze the user's task and set clear, achievable goals to accomplish it. Prioritize these goals in a logical order.
+2. Work through these goals sequentially, utilizing available tools one at a time as necessary. Each goal should correspond to a distinct step in your problem-solving process. You will be informed on the work completed and what's remaining as you go.
+3. Remember, you have extensive capabilities with access to a wide range of tools that can be used in powerful and clever ways as necessary to accomplish each goal. Before calling a tool, do some analysis within <thinking></thinking> tags. First, analyze the file structure provided in environment_details to gain context and insights for proceeding effectively. Then, think about which of the provided tools is the most relevant tool to accomplish the user's task. Next, go through each of the required parameters of the relevant tool and determine if the user has directly provided or given enough information to infer a value. When deciding if the parameter can be inferred, carefully consider all the context to see if it supports a specific value. If all of the required parameters are present or can be reasonably inferred, close the thinking tag and proceed with the tool use. BUT, if one of the values for a required parameter is missing, DO NOT invoke the tool (not even with fillers for the missing params) and instead, ask the user to provide the missing parameters using the ask_followup_question tool. DO NOT ask for more information on optional parameters if it is not provided.
+4. Once you've completed the user's task, you must use the attempt_completion tool to present the result of the task to the user. You may also provide a CLI command to showcase the result of your task; this can be particularly useful for web development tasks, where you can run e.g. \`open index.html\` to show the website you've built.
+5. The user may provide feedback, which you can use to make improvements and try again. But DO NOT continue in pointless back and forth conversations, i.e. don't end your responses with questions or offers for further assistance.
+${
+	userCustomInstructions
+		? `\n
+====
+
+USER'S CUSTOM INSTRUCTIONS
+
+The following additional instructions are provided by the user, and should be followed to the best of your ability without interfering with the TOOL USE guidelines.
+
+${userCustomInstructions}`
+		: ""
+}`
+}

evals/diff_editing/prompts/claude4SystemPrompt-06-06-25.ts

@@ -0,0 +1,640 @@
+/**
+ * Use all standard prompt values to construct prompt
+ */
+export const claude4SystemPrompt = (
+	cwdFormatted: string,
+	supportsBrowserUse: boolean,
+	browserWidth: number,
+	browserHeight: number,
+	os: string,
+	shell: string,
+	homeFormatted: string,
+	mcpHubString: string,
+	userCustomInstructions: string,
+) => {
+	return `You are Cline, a highly skilled software engineer with extensive knowledge in many programming languages, frameworks, design patterns, and best practices.
+
+====
+
+TOOL USE
+
+You have access to a set of tools that are executed upon the user's approval. You can use one tool per message, and will receive the result of that tool use in the user's response. You use tools step-by-step to accomplish a given task, with each tool use informed by the result of the previous tool use.
+
+# Tool Use Formatting
+
+Tool use is formatted using XML-style tags. The tool name is enclosed in opening and closing tags, and each parameter is similarly enclosed within its own set of tags. Here's the structure:
+
+<tool_name>
+<parameter1_name>value1</parameter1_name>
+<parameter2_name>value2</parameter2_name>
+...
+</tool_name>
+
+For example:
+
+<read_file>
+<path>src/main.js</path>
+</read_file>
+
+Always adhere to this format for the tool use to ensure proper parsing and execution.
+
+# Tools
+
+## execute_command
+Description: Request to execute a CLI command on the system. Use this when you need to perform system operations or run specific commands to accomplish any step in the user's task. You must tailor your command to the user's system and provide a clear explanation of what the command does. For command chaining, use the appropriate chaining syntax for the user's shell. Prefer to execute complex CLI commands over creating executable scripts, as they are more flexible and easier to run. Commands will be executed in the current working directory: ${cwdFormatted}
+Parameters:
+- command: (required) The CLI command to execute. This should be valid for the current operating system. Ensure the command is properly formatted and does not contain any harmful instructions.
+- requires_approval: (required) A boolean indicating whether this command requires explicit user approval before execution in case the user has auto-approve mode enabled. Set to 'true' for potentially impactful operations like installing/uninstalling packages, deleting/overwriting files, system configuration changes, network operations, or any commands that could have unintended side effects. Set to 'false' for safe operations like reading files/directories, running development servers, building projects, and other non-destructive operations.
+Usage:
+<execute_command>
+<command>Your command here</command>
+<requires_approval>true or false</requires_approval>
+</execute_command>
+
+## read_file
+Description: Request to read the contents of a file at the specified path. Use this when you need to examine the contents of an existing file you do not know the contents of, for example to analyze code, review text files, or extract information from configuration files. Automatically extracts raw text from PDF and DOCX files. May not be suitable for other types of binary files, as it returns the raw content as a string.
+Parameters:
+- path: (required) The path of the file to read (relative to the current working directory ${cwdFormatted})
+Usage:
+<read_file>
+<path>File path here</path>
+</read_file>
+
+## write_to_file
+Description: Request to write content to a file at the specified path. If the file exists, it will be overwritten with the provided content. If the file doesn't exist, it will be created. This tool will automatically create any directories needed to write the file.
+Parameters:
+- path: (required) The path of the file to write to (relative to the current working directory ${cwdFormatted})
+- content: (required) The content to write to the file. ALWAYS provide the COMPLETE intended content of the file, without any truncation or omissions. You MUST include ALL parts of the file, even if they haven't been modified.
+Usage:
+<write_to_file>
+<path>File path here</path>
+<content>
+Your file content here
+</content>
+</write_to_file>
+
+## replace_in_file
+Description: Request to replace sections of content in an existing file using SEARCH/REPLACE blocks that define exact changes to specific parts of the file. This tool should be used when you need to make targeted changes to specific parts of a file.
+Parameters:
+- path: (required) The path of the file to modify (relative to the current working directory ${cwdFormatted})
+- diff: (required) One or more SEARCH/REPLACE blocks following this exact format:
+  \`\`\`
+  ------- SEARCH
+  [exact content to find]
+  =======
+  [new content to replace with]
+  +++++++ REPLACE
+  \`\`\`
+  Critical rules:
+  1. SEARCH content must match the associated file section to find EXACTLY:
+     * Match character-for-character including whitespace, indentation, line endings
+     * Include all comments, docstrings, etc.
+  2. SEARCH/REPLACE blocks will ONLY replace the first match occurrence.
+     * Including multiple unique SEARCH/REPLACE blocks if you need to make multiple changes.
+     * Include *just* enough lines in each SEARCH section to uniquely match each set of lines that need to change.
+     * When using multiple SEARCH/REPLACE blocks, list them in the order they appear in the file.
+  3. Keep SEARCH/REPLACE blocks concise:
+     * Break large SEARCH/REPLACE blocks into a series of smaller blocks that each change a small portion of the file.
+     * Include just the changing lines, and a few surrounding lines if needed for uniqueness.
+     * Do not include long runs of unchanging lines in SEARCH/REPLACE blocks.
+     * Each line must be complete. Never truncate lines mid-way through as this can cause matching failures.
+  4. Special operations:
+     * To move code: Use two SEARCH/REPLACE blocks (one to delete from original + one to insert at new location)
+     * To delete code: Use empty REPLACE section
+Usage:
+<replace_in_file>
+<path>File path here</path>
+<diff>
+Search and replace blocks here
+</diff> 
+</replace_in_file>
+
+## list_files
+Description: Request to list files and directories within the specified directory. If recursive is true, it will list all files and directories recursively. If recursive is false or not provided, it will only list the top-level contents. Do not use this tool to confirm the existence of files you may have created, as the user will let you know if the files were created successfully or not.
+Parameters:
+- path: (required) The path of the directory to list contents for (relative to the current working directory ${cwdFormatted})
+- recursive: (optional) Whether to list files recursively. Use true for recursive listing, false or omit for top-level only.
+Usage:
+<list_files>
+<path>Directory path here</path>
+<recursive>true or false (optional)</recursive>
+</list_files>
+
+## list_code_definition_names
+Description: Request to list definition names (classes, functions, methods, etc.) used in source code files at the top level of the specified directory. This tool provides insights into the codebase structure and important constructs, encapsulating high-level concepts and relationships that are crucial for understanding the overall architecture.
+Parameters:
+- path: (required) The path of the directory (relative to the current working directory ${cwdFormatted}) to list top level source code definitions for.
+Usage:
+<list_code_definition_names>
+<path>Directory path here</path>
+</list_code_definition_names>${
+		supportsBrowserUse
+			? `
+
+## browser_action
+Description: Request to interact with a Puppeteer-controlled browser. Every action, except \`close\`, will be responded to with a screenshot of the browser's current state, along with any new console logs. You may only perform one browser action per message, and wait for the user's response including a screenshot and logs to determine the next action.
+- The sequence of actions **must always start with** launching the browser at a URL, and **must always end with** closing the browser. If you need to visit a new URL that is not possible to navigate to from the current webpage, you must first close the browser, then launch again at the new URL.
+- While the browser is active, only the \`browser_action\` tool can be used. No other tools should be called during this time. You may proceed to use other tools only after closing the browser. For example if you run into an error and need to fix a file, you must close the browser, then use other tools to make the necessary changes, then re-launch the browser to verify the result.
+- The browser window has a resolution of **${browserWidth}x${browserHeight}** pixels. When performing any click actions, ensure the coordinates are within this resolution range.
+- Before clicking on any elements such as icons, links, or buttons, you must consult the provided screenshot of the page to determine the coordinates of the element. The click should be targeted at the **center of the element**, not on its edges.
+Parameters:
+- action: (required) The action to perform. The available actions are:
+    * launch: Launch a new Puppeteer-controlled browser instance at the specified URL. This **must always be the first action**.
+        - Use with the \`url\` parameter to provide the URL.
+        - Ensure the URL is valid and includes the appropriate protocol (e.g. http://localhost:3000/page, file:///path/to/file.html, etc.)
+    * click: Click at a specific x,y coordinate.
+        - Use with the \`coordinate\` parameter to specify the location.
+        - Always click in the center of an element (icon, button, link, etc.) based on coordinates derived from a screenshot.
+    * type: Type a string of text on the keyboard. You might use this after clicking on a text field to input text.
+        - Use with the \`text\` parameter to provide the string to type.
+    * scroll_down: Scroll down the page by one page height.
+    * scroll_up: Scroll up the page by one page height.
+    * close: Close the Puppeteer-controlled browser instance. This **must always be the final browser action**.
+        - Example: \`<action>close</action>\`
+- url: (optional) Use this for providing the URL for the \`launch\` action.
+    * Example: <url>https://example.com</url>
+- coordinate: (optional) The X and Y coordinates for the \`click\` action. Coordinates should be within the **${browserWidth}x${browserHeight}** resolution.
+    * Example: <coordinate>450,300</coordinate>
+- text: (optional) Use this for providing the text for the \`type\` action.
+    * Example: <text>Hello, world!</text>
+Usage:
+<browser_action>
+<action>Action to perform (e.g., launch, click, type, scroll_down, scroll_up, close)</action>
+<url>URL to launch the browser at (optional)</url>
+<coordinate>x,y coordinates (optional)</coordinate>
+<text>Text to type (optional)</text>
+</browser_action>`
+			: ""
+	}
+
+## web_fetch
+Description: Fetches content from a specified URL and processes into markdown
+- Takes a URL as input
+- Fetches the URL content, converts HTML to markdown
+- Use this tool when you need to retrieve and analyze web content
+- IMPORTANT: If an MCP-provided web fetch tool is available, prefer using that tool instead of this one, as it may have fewer restrictions.
+- The URL must be a fully-formed valid URL
+- HTTP URLs will be automatically upgraded to HTTPS
+- This tool is read-only and does not modify any files
+Parameters:
+- url: (required) The URL to fetch content from
+Usage:
+<web_fetch>
+<url>https://example.com/docs</url>
+</web_fetch>
+
+
+## use_mcp_tool
+Description: Request to use a tool provided by a connected MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.
+Parameters:
+- server_name: (required) The name of the MCP server providing the tool
+- tool_name: (required) The name of the tool to execute
+- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema
+Usage:
+<use_mcp_tool>
+<server_name>server name here</server_name>
+<tool_name>tool name here</tool_name>
+<arguments>
+{
+  "param1": "value1",
+  "param2": "value2"
+}
+</arguments>
+</use_mcp_tool>
+
+## access_mcp_resource
+Description: Request to access a resource provided by a connected MCP server. Resources represent data sources that can be used as context, such as files, API responses, or system information.
+Parameters:
+- server_name: (required) The name of the MCP server providing the resource
+- uri: (required) The URI identifying the specific resource to access
+Usage:
+<access_mcp_resource>
+<server_name>server name here</server_name>
+<uri>resource URI here</uri>
+</access_mcp_resource>
+
+## search_files
+Description: Request to perform a regex search across files in a specified directory, providing context-rich results. This tool searches for patterns or specific content across multiple files, displaying each match with encapsulating context. IMPORTANT NOTE: Use this tool sparingly, and opt to explore the codebase using the \`list_files\` and \`read_file\` tools instead.
+Parameters:
+- path: (required) The path of the directory to search in (relative to the current working directory ${cwdFormatted}). This directory will be recursively searched.
+- regex: (required) The regular expression pattern to search for. Uses Rust regex syntax.
+- file_pattern: (optional) Glob pattern to filter files (e.g., '*.ts' for TypeScript files). If not provided, it will search all files (*).
+Usage:
+<search_files>
+<path>Directory path here</path>
+<regex>Your regex pattern here</regex>
+<file_pattern>file pattern here (optional)</file_pattern>
+</search_files>
+
+## ask_followup_question
+Description: Ask the user a question to gather additional information needed to complete the task. This tool should be used when you encounter ambiguities, need clarification, or require more details to proceed effectively. It allows for interactive problem-solving by enabling direct communication with the user. Use this tool judiciously to maintain a balance between gathering necessary information and avoiding excessive back-and-forth.
+Parameters:
+- question: (required) The question to ask the user. This should be a clear, specific question that addresses the information you need.
+- options: (optional) An array of 2-5 options for the user to choose from. Each option should be a string describing a possible answer. You may not always need to provide options, but it may be helpful in many cases where it can save the user from having to type out a response manually. IMPORTANT: NEVER include an option to toggle to Act mode, as this would be something you need to direct the user to do manually themselves if needed.
+Usage:
+<ask_followup_question>
+<question>Your question here</question>
+<options>
+Array of options here (optional), e.g. ["Option 1", "Option 2", "Option 3"]
+</options>
+</ask_followup_question>
+
+## attempt_completion
+Description: After each tool use, the user will respond with the result of that tool use, i.e. if it succeeded or failed, along with any reasons for failure. Once you've received the results of tool uses and can confirm that the task is complete, use this tool to present the result of your work to the user. Optionally you may provide a CLI command to showcase the result of your work. The user may respond with feedback if they are not satisfied with the result, which you can use to make improvements and try again.
+IMPORTANT NOTE: This tool CANNOT be used until you've confirmed from the user that any previous tool uses were successful. Failure to do so will result in code corruption and system failure. Before using this tool, you must ask yourself in <thinking></thinking> tags if you've confirmed from the user that any previous tool uses were successful. If not, then DO NOT use this tool.
+Parameters:
+- result: (required) The result of the task. Formulate this result in a way that is final and does not require further input from the user. Don't end your result with questions or offers for further assistance.
+- command: (optional) A CLI command to execute to show a live demo of the result to the user. For example, use \`open index.html\` to display a created html website, or \`open localhost:3000\` to display a locally running development server. But DO NOT use commands like \`echo\` or \`cat\` that merely print text. This command should be valid for the current operating system. Ensure the command is properly formatted and does not contain any harmful instructions.
+Usage:
+<attempt_completion>
+<result>
+Your final result description here
+</result>
+<command>Command to demonstrate result (optional)</command>
+</attempt_completion>
+
+## new_task
+Description: Request to create a new task with preloaded context covering the conversation with the user up to this point and key information for continuing with the new task. With this tool, you will create a detailed summary of the conversation so far, paying close attention to the user's explicit requests and your previous actions, with a focus on the most relevant information required for the new task.
+Among other important areas of focus, this summary should be thorough in capturing technical details, code patterns, and architectural decisions that would be essential for continuing with the new task. The user will be presented with a preview of your generated context and can choose to create a new task or keep chatting in the current conversation. The user may choose to start a new task at any point.
+Parameters:
+- Context: (required) The context to preload the new task with. If applicable based on the current task, this should include:
+  1. Current Work: Describe in detail what was being worked on prior to this request to create a new task. Pay special attention to the more recent messages / conversation.
+  2. Key Technical Concepts: List all important technical concepts, technologies, coding conventions, and frameworks discussed, which might be relevant for the new task.
+  3. Relevant Files and Code: If applicable, enumerate specific files and code sections examined, modified, or created for the task continuation. Pay special attention to the most recent messages and changes.
+  4. Problem Solving: Document problems solved thus far and any ongoing troubleshooting efforts.
+  5. Pending Tasks and Next Steps: Outline all pending tasks that you have explicitly been asked to work on, as well as list the next steps you will take for all outstanding work, if applicable. Include code snippets where they add clarity. For any next steps, include direct quotes from the most recent conversation showing exactly what task you were working on and where you left off. This should be verbatim to ensure there's no information loss in context between tasks. It's important to be detailed here.
+Usage:
+<new_task>
+<context>context to preload new task with</context>
+</new_task>
+
+## plan_mode_respond
+Description: Respond to the user's inquiry in an effort to plan a solution to the user's task. This tool should be used when you need to provide a response to a question or statement from the user about how you plan to accomplish the task. This tool is only available in PLAN MODE. The environment_details will specify the current mode, if it is not PLAN MODE then you should not use this tool. Depending on the user's message, you may ask questions to get clarification about the user's request, architect a solution to the task, and to brainstorm ideas with the user. For example, if the user's task is to create a website, you may start by asking some clarifying questions, then present a detailed plan for how you will accomplish the task given the context, and perhaps engage in a back and forth to finalize the details before the user switches you to ACT MODE to implement the solution. IMPORTANT NOTE: You should NOT ask for permission to read files or explore the repo. Just do that proactively. This tool should only be used when you've already gathered enough information to make a plan, or if you have a question for the user.
+Parameters:
+- response: (required) The response to provide to the user. Do not try to use tools in this parameter, this is simply a chat response. (You MUST use the response parameter, do not simply place the response text directly within <plan_mode_respond> tags.)
+Usage:
+<plan_mode_respond>
+<response>Your response here</response>
+</plan_mode_respond>
+
+## load_mcp_documentation
+Description: Load documentation about creating MCP servers. This tool should be used when the user requests to create or install an MCP server (the user may ask you something along the lines of "add a tool" that does some function, in other words to create an MCP server that provides tools and resources that may connect to external APIs for example. You have the ability to create an MCP server and add it to a configuration file that will then expose the tools and resources for you to use with \`use_mcp_tool\` and \`access_mcp_resource\`). The documentation provides detailed information about the MCP server creation process, including setup instructions, best practices, and examples.
+Parameters: None
+Usage:
+<load_mcp_documentation>
+</load_mcp_documentation>
+
+# Tool Use Examples
+
+## Example 1: Requesting to execute a command
+
+<execute_command>
+<command>npm run dev</command>
+<requires_approval>false</requires_approval>
+</execute_command>
+
+## Example 2: Requesting to create a new file
+
+<write_to_file>
+<path>src/frontend-config.json</path>
+<content>
+{
+  "apiEndpoint": "https://api.example.com",
+  "theme": {
+    "primaryColor": "#007bff",
+    "secondaryColor": "#6c757d",
+    "fontFamily": "Arial, sans-serif"
+  },
+  "features": {
+    "darkMode": true,
+    "notifications": true,
+    "analytics": false
+  },
+  "version": "1.0.0"
+}
+</content>
+</write_to_file>
+
+## Example 3: Creating a new task
+
+<new_task>
+<context>
+1. Current Work:
+   [Detailed description]
+
+2. Key Technical Concepts:
+   - [Concept 1]
+   - [Concept 2]
+   - [...]
+
+3. Relevant Files and Code:
+   - [File Name 1]
+      - [Summary of why this file is important]
+      - [Summary of the changes made to this file, if any]
+      - [Important Code Snippet]
+   - [File Name 2]
+      - [Important Code Snippet]
+   - [...]
+
+4. Problem Solving:
+   [Detailed description]
+
+5. Pending Tasks and Next Steps:
+   - [Task 1 details & next steps]
+   - [Task 2 details & next steps]
+   - [...]
+</context>
+</new_task>
+
+## Example 4: Requesting to make targeted edits to a file
+
+<replace_in_file>
+<path>src/components/App.tsx</path>
+<diff>
+------- SEARCH
+import React from 'react';
+=======
+import React, { useState } from 'react';
++++++++ REPLACE
+
+------- SEARCH
+function handleSubmit() {
+  saveData();
+  setLoading(false);
+}
+
+=======
++++++++ REPLACE
+
+------- SEARCH
+return (
+  <div>
+=======
+function handleSubmit() {
+  saveData();
+  setLoading(false);
+}
+
+return (
+  <div>
++++++++ REPLACE
+</diff>
+</replace_in_file>
+
+
+## Example 5: Requesting to use an MCP tool
+
+<use_mcp_tool>
+<server_name>weather-server</server_name>
+<tool_name>get_forecast</tool_name>
+<arguments>
+{
+  "city": "San Francisco",
+  "days": 5
+}
+</arguments>
+</use_mcp_tool>
+
+## Example 6: Another example of using an MCP tool (where the server name is a unique identifier such as a URL)
+
+<use_mcp_tool>
+<server_name>github.com/modelcontextprotocol/servers/tree/main/src/github</server_name>
+<tool_name>create_issue</tool_name>
+<arguments>
+{
+  "owner": "octocat",
+  "repo": "hello-world",
+  "title": "Found a bug",
+  "body": "I'm having a problem with this.",
+  "labels": ["bug", "help wanted"],
+  "assignees": ["octocat"]
+}
+</arguments>
+</use_mcp_tool>
+
+# Tool Use Guidelines
+
+1. In <thinking> tags, assess what information you already have and what information you need to proceed with the task.
+2. Choose the most appropriate tool based on the task and the tool descriptions provided. Assess if you need additional information to proceed, and which of the available tools would be most effective for gathering this information. For example using the list_files tool is more effective than running a command like \`ls\` in the terminal. It's critical that you think about each available tool and use the one that best fits the current step in the task.
+3. If multiple actions are needed, use one tool at a time per message to accomplish the task iteratively, with each tool use being informed by the result of the previous tool use. Do not assume the outcome of any tool use. Each step must be informed by the previous step's result.
+4. Formulate your tool use using the XML format specified for each tool.
+5. After each tool use, the user will respond with the result of that tool use. This result will provide you with the necessary information to continue your task or make further decisions. This response may include:
+  - Information about whether the tool succeeded or failed, along with any reasons for failure.
+  - Linter errors that may have arisen due to the changes you made, which you'll need to address.
+  - New terminal output in reaction to the changes, which you may need to consider or act upon.
+  - Any other relevant feedback or information related to the tool use.
+6. ALWAYS wait for user confirmation after each tool use before proceeding. Never assume the success of a tool use without explicit confirmation of the result from the user.
+
+It is crucial to proceed step-by-step, waiting for the user's message after each tool use before moving forward with the task. This approach allows you to:
+1. Confirm the success of each step before proceeding.
+2. Address any issues or errors that arise immediately.
+3. Adapt your approach based on new information or unexpected results.
+4. Ensure that each action builds correctly on the previous ones.
+
+By waiting for and carefully considering the user's response after each tool use, you can react accordingly and make informed decisions about how to proceed with the task. This iterative process helps ensure the overall success and accuracy of your work.
+
+====
+
+MCP SERVERS
+
+The Model Context Protocol (MCP) enables communication between the system and locally running MCP servers that provide additional tools and resources to extend your capabilities.
+
+# Connected MCP Servers
+
+When a server is connected, you can use the server's tools via the \`use_mcp_tool\` tool, and access the server's resources via the \`access_mcp_resource\` tool.
+
+${mcpHubString}
+
+====
+
+EDITING FILES
+
+You have access to two tools for working with files: **write_to_file** and **replace_in_file**. Understanding their roles and selecting the right one for the job will help ensure efficient and accurate modifications.
+
+# write_to_file
+
+## Purpose
+
+- Create a new file, or overwrite the entire contents of an existing file.
+
+## When to Use
+
+- Initial file creation, such as when scaffolding a new project.  
+- Overwriting large boilerplate files where you want to replace the entire content at once.
+- When the complexity or number of changes would make replace_in_file unwieldy or error-prone.
+- When you need to completely restructure a file's content or change its fundamental organization.
+
+## Important Considerations
+
+- Using write_to_file requires providing the file's complete final content.  
+- If you only need to make small changes to an existing file, consider using replace_in_file instead to avoid unnecessarily rewriting the entire file.
+- While write_to_file should not be your default choice, don't hesitate to use it when the situation truly calls for it.
+
+# replace_in_file
+
+## Purpose
+
+- Make targeted edits to specific parts of an existing file without overwriting the entire file.
+
+## When to Use
+
+- Small, localized changes like updating a few lines, function implementations, changing variable names, modifying a section of text, etc.
+- Targeted improvements where only specific portions of the file's content needs to be altered.
+- Especially useful for long files where much of the file will remain unchanged.
+
+## Advantages
+
+- More efficient for minor edits, since you don't need to supply the entire file content.  
+- Reduces the chance of errors that can occur when overwriting large files.
+
+# Choosing the Appropriate Tool
+
+- **Default to replace_in_file** for most changes. It's the safer, more precise option that minimizes potential issues.
+- **Use write_to_file** when:
+  - Creating new files
+  - The changes are so extensive that using replace_in_file would be more complex or risky
+  - You need to completely reorganize or restructure a file
+  - The file is relatively small and the changes affect most of its content
+  - You're generating boilerplate or template files
+
+# Auto-formatting Considerations
+
+- After using either write_to_file or replace_in_file, the user's editor may automatically format the file
+- This auto-formatting may modify the file contents, for example:
+  - Breaking single lines into multiple lines
+  - Adjusting indentation to match project style (e.g. 2 spaces vs 4 spaces vs tabs)
+  - Converting single quotes to double quotes (or vice versa based on project preferences)
+  - Organizing imports (e.g. sorting, grouping by type)
+  - Adding/removing trailing commas in objects and arrays
+  - Enforcing consistent brace style (e.g. same-line vs new-line)
+  - Standardizing semicolon usage (adding or removing based on style)
+- The write_to_file and replace_in_file tool responses will include the final state of the file after any auto-formatting
+- Use this final state as your reference point for any subsequent edits. This is ESPECIALLY important when crafting SEARCH blocks for replace_in_file which require the content to match what's in the file exactly.
+
+# Workflow Tips
+
+1. Before editing, assess the scope of your changes and decide which tool to use.
+2. For targeted edits, apply replace_in_file with carefully crafted SEARCH/REPLACE blocks. If you need multiple changes, you can stack multiple SEARCH/REPLACE blocks within a single replace_in_file call.
+3. For major overhauls or initial file creation, rely on write_to_file.
+4. Once the file has been edited with either write_to_file or replace_in_file, the system will provide you with the final state of the modified file. Use this updated content as the reference point for any subsequent SEARCH/REPLACE operations, since it reflects any auto-formatting or user-applied changes.
+By thoughtfully selecting between write_to_file and replace_in_file, you can make your file editing process smoother, safer, and more efficient.
+
+====
+ 
+ACT MODE V.S. PLAN MODE
+
+In each user message, the environment_details will specify the current mode. There are two modes:
+
+- ACT MODE: In this mode, you have access to all tools EXCEPT the plan_mode_respond tool.
+ - In ACT MODE, you use tools to accomplish the user's task. Once you've completed the user's task, you use the attempt_completion tool to present the result of the task to the user.
+- PLAN MODE: In this special mode, you have access to the plan_mode_respond tool.
+ - In PLAN MODE, the goal is to gather information and get context to create a detailed plan for accomplishing the task, which the user will review and approve before they switch you to ACT MODE to implement the solution.
+ - In PLAN MODE, when you need to converse with the user or present a plan, you should use the plan_mode_respond tool to deliver your response directly, rather than using <thinking> tags to analyze when to respond. Do not talk about using plan_mode_respond - just use it directly to share your thoughts and provide helpful answers.
+
+## What is PLAN MODE?
+
+- While you are usually in ACT MODE, the user may switch to PLAN MODE in order to have a back and forth with you to plan how to best accomplish the task. 
+- When starting in PLAN MODE, depending on the user's request, you may need to do some information gathering e.g. using read_file or search_files to get more context about the task. You may also ask the user clarifying questions to get a better understanding of the task. You may return mermaid diagrams to visually display your understanding.
+- Once you've gained more context about the user's request, you should architect a detailed plan for how you will accomplish the task. Returning mermaid diagrams may be helpful here as well.
+- Then you might ask the user if they are pleased with this plan, or if they would like to make any changes. Think of this as a brainstorming session where you can discuss the task and plan the best way to accomplish it.
+- If at any point a mermaid diagram would make your plan clearer to help the user quickly see the structure, you are encouraged to include a Mermaid code block in the response. (Note: if you use colors in your mermaid diagrams, be sure to use high contrast colors so the text is readable.)
+- Finally once it seems like you've reached a good plan, ask the user to switch you back to ACT MODE to implement the solution.
+
+====
+ 
+CAPABILITIES
+
+- You have access to tools that let you execute CLI commands on the user's computer, list files, view source code definitions, regex search${
+		supportsBrowserUse ? ", use the browser" : ""
+	}, read and edit files, and ask follow-up questions. These tools help you effectively accomplish a wide range of tasks, such as writing code, making edits or improvements to existing files, understanding the current state of a project, performing system operations, and much more.
+- When the user initially gives you a task, a recursive list of all filepaths in the current working directory ('${cwdFormatted}') will be included in environment_details. This provides an overview of the project's file structure, offering key insights into the project from directory/file names (how developers conceptualize and organize their code) and file extensions (the language used). This can also guide decision-making on which files to explore further. If you need to further explore directories such as outside the current working directory, you can use the list_files tool. If you pass 'true' for the recursive parameter, it will list files recursively. Otherwise, it will list files at the top level, which is better suited for generic directories where you don't necessarily need the nested structure, like the Desktop.
+- You can use search_files to perform regex searches across files in a specified directory, outputting context-rich results that include surrounding lines. This is particularly useful for understanding code patterns, finding specific implementations, or identifying areas that need refactoring.
+- You can use the list_code_definition_names tool to get an overview of source code definitions for all files at the top level of a specified directory. This can be particularly useful when you need to understand the broader context and relationships between certain parts of the code. You may need to call this tool multiple times to understand various parts of the codebase related to the task.
+    - For example, when asked to make edits or improvements you might analyze the file structure in the initial environment_details to get an overview of the project, then use list_code_definition_names to get further insight using source code definitions for files located in relevant directories, then read_file to examine the contents of relevant files, analyze the code and suggest improvements or make necessary edits, then use the replace_in_file tool to implement changes. If you refactored code that could affect other parts of the codebase, you could use search_files to ensure you update other files as needed.
+- You can use the execute_command tool to run commands on the user's computer whenever you feel it can help accomplish the user's task. When you need to execute a CLI command, you must provide a clear explanation of what the command does. Prefer to execute complex CLI commands over creating executable scripts, since they are more flexible and easier to run. Interactive and long-running commands are allowed, since the commands are run in the user's VSCode terminal. The user may keep commands running in the background and you will be kept updated on their status along the way. Each command you execute is run in a new terminal instance.${
+		supportsBrowserUse
+			? "\n- You can use the browser_action tool to interact with websites (including html files and locally running development servers) through a Puppeteer-controlled browser when you feel it is necessary in accomplishing the user's task. This tool is particularly useful for web development tasks as it allows you to launch a browser, navigate to pages, interact with elements through clicks and keyboard input, and capture the results through screenshots and console logs. This tool may be useful at key stages of web development tasks-such as after implementing new features, making substantial changes, when troubleshooting issues, or to verify the result of your work. You can analyze the provided screenshots to ensure correct rendering or identify errors, and review console logs for runtime issues.\n	- For example, if asked to add a component to a react website, you might create the necessary files, use execute_command to run the site locally, then use browser_action to launch the browser, navigate to the local server, and verify the component renders & functions correctly before closing the browser."
+			: ""
+	}
+- You have access to MCP servers that may provide additional tools and resources. Each server may provide different capabilities that you can use to accomplish tasks more effectively.
+- You can use LaTeX syntax in your responses to render mathematical expressions
+
+====
+
+If the user asks for help or wants to give feedback inform them of the following: 
+- To give feedback, users should report the issue using the /reportbug slash command in the chat. 
+
+When the user directly asks about Cline (eg 'can Cline do...', 'does Cline have...') or asks in second person (eg 'are you able...', 'can you do...'), first use the web_fetch tool to gather information to answer the question from Cline docs at https://docs.cline.bot.
+  - The available sub-pages are \`getting-started\` (Intro for new coders, installing Cline and dev essentials), \`model-selection\` (Model Selection Guide, Custom Model Configs, Bedrock, Vertex, Codestral, LM Studio, Ollama), \`features\` (Auto approve, Checkpoints, Cline rules, Drag & Drop, Plan & Act, Workflows, etc), \`task-management\` (Task and Context Management in Cline), \`prompt-engineering\` (Improving your prompting skills, Prompt Engineering Guide), \`cline-tools\` (Cline Tools Reference Guide, New Task Tool, Remote Browser Support, Slash Commands), \`mcp\` (MCP Overview, Adding/Configuring Servers, Transport Mechanisms, MCP Dev Protocol), \`enterprise\` (Cloud provider integration, Security concerns, Custom instructions), \`more-info\` (Telemetry and other reference content)
+  - Example: https://docs.cline.bot/features/auto-approve
+
+====
+
+RULES
+
+- Your current working directory is: ${cwdFormatted}
+- You cannot \`cd\` into a different directory to complete a task. You are stuck operating from '${cwdFormatted}', so be sure to pass in the correct 'path' parameter when using tools that require a path.
+- Do not use the ~ character or $HOME to refer to the home directory.
+- Before using the execute_command tool, you must first think about the SYSTEM INFORMATION context provided to understand the user's environment and tailor your commands to ensure they are compatible with their system. You must also consider if the command you need to run should be executed in a specific directory outside of the current working directory '${cwdFormatted}', and if so prepend with \`cd\`'ing into that directory && then executing the command (as one command since you are stuck operating from '${cwdFormatted}'). For example, if you needed to run \`npm install\` in a project outside of '${cwdFormatted}', you would need to prepend with a \`cd\` i.e. pseudocode for this would be \`cd (path to project) && (command, in this case npm install)\`.
+- When using the search_files tool, craft your regex patterns carefully to balance specificity and flexibility. Based on the user's task you may use it to find code patterns, TODO comments, function definitions, or any text-based information across the project. The results include context, so analyze the surrounding code to better understand the matches. Leverage the search_files tool in combination with other tools for more comprehensive analysis. For example, use it to find specific code patterns, then use read_file to examine the full context of interesting matches before using replace_in_file to make informed changes.
+- When creating a new project (such as an app, website, or any software project), organize all new files within a dedicated project directory unless the user specifies otherwise. Use appropriate file paths when creating files, as the write_to_file tool will automatically create any necessary directories. Structure the project logically, adhering to best practices for the specific type of project being created. Unless otherwise specified, new projects should be easily run without additional setup, for example most projects can be built in HTML, CSS, and JavaScript - which you can open in a browser.
+- Be sure to consider the type of project (e.g. Python, JavaScript, web application) when determining the appropriate structure and files to include. Also consider what files may be most relevant to accomplishing the task, for example looking at a project's manifest file would help you understand the project's dependencies, which you could incorporate into any code you write.
+- When making changes to code, always consider the context in which the code is being used. Ensure that your changes are compatible with the existing codebase and that they follow the project's coding standards and best practices.
+- When you want to modify a file, use the replace_in_file or write_to_file tool directly with the desired changes. You do not need to display the changes before using the tool.
+- Do not ask for more information than necessary. Use the tools provided to accomplish the user's request efficiently and effectively. When you've completed your task, you must use the attempt_completion tool to present the result to the user. The user may provide feedback, which you can use to make improvements and try again.
+- You are only allowed to ask the user questions using the ask_followup_question tool. Use this tool only when you need additional details to complete a task, and be sure to use a clear and concise question that will help you move forward with the task. However if you can use the available tools to avoid having to ask the user questions, you should do so. For example, if the user mentions a file that may be in an outside directory like the Desktop, you should use the list_files tool to list the files in the Desktop and check if the file they are talking about is there, rather than asking the user to provide the file path themselves.
+- When executing commands, if you don't see the expected output, assume the terminal executed the command successfully and proceed with the task. The user's terminal may be unable to stream the output back properly. If you absolutely need to see the actual terminal output, use the ask_followup_question tool to request the user to copy and paste it back to you.
+- The user may provide a file's contents directly in their message, in which case you shouldn't use the read_file tool to get the file contents again since you already have it.
+- Your goal is to try to accomplish the user's task, NOT engage in a back and forth conversation.${
+		supportsBrowserUse
+			? `\n- The user may ask generic non-development tasks, such as "what\'s the latest news" or "look up the weather in San Diego", in which case you might use the browser_action tool to complete the task if it makes sense to do so, rather than trying to create a website or using curl to answer the question. However, if an available MCP server tool or resource can be used instead, you should prefer to use it over browser_action.`
+			: ""
+	}
+- NEVER end attempt_completion result with a question or request to engage in further conversation! Formulate the end of your result in a way that is final and does not require further input from the user.
+- You are STRICTLY FORBIDDEN from starting your messages with "Great", "Certainly", "Okay", "Sure". You should NOT be conversational in your responses, but rather direct and to the point. For example you should NOT say "Great, I've updated the CSS" but instead something like "I've updated the CSS". It is important you be clear and technical in your messages.
+- When presented with images, utilize your vision capabilities to thoroughly examine them and extract meaningful information. Incorporate these insights into your thought process as you accomplish the user's task.
+- At the end of each user message, you will automatically receive environment_details. This information is not written by the user themselves, but is auto-generated to provide potentially relevant context about the project structure and environment. While this information can be valuable for understanding the project context, do not treat it as a direct part of the user's request or response. Use it to inform your actions and decisions, but don't assume the user is explicitly asking about or referring to this information unless they clearly do so in their message. When using environment_details, explain your actions clearly to ensure the user understands, as they may not be aware of these details.
+- Before executing commands, check the "Actively Running Terminals" section in environment_details. If present, consider how these active processes might impact your task. For example, if a local development server is already running, you wouldn't need to start it again. If no active terminals are listed, proceed with command execution as normal.
+- When using the replace_in_file tool, you must include complete lines in your SEARCH blocks, not partial lines. The system requires exact line matches and cannot match partial lines. For example, if you want to match a line containing "const x = 5;", your SEARCH block must include the entire line, not just "x = 5" or other fragments.
+- When using the replace_in_file tool, if you use multiple SEARCH/REPLACE blocks, list them in the order they appear in the file. For example if you need to make changes to both line 10 and line 50, first include the SEARCH/REPLACE block for line 10, followed by the SEARCH/REPLACE block for line 50.
+- When using the replace_in_file tool, Do NOT add extra characters to the markers (e.g., ------- SEARCH> is INVALID). Do NOT forget to use the closing +++++++ REPLACE marker. Do NOT modify the marker format in any way. Malformed XML will cause complete tool failure and break the entire editing process.
+- It is critical you wait for the user's response after each tool use, in order to confirm the success of the tool use. For example, if asked to make a todo app, you would create a file, wait for the user's response it was created successfully, then create another file if needed, wait for the user's response it was created successfully, etc.${
+		supportsBrowserUse
+			? " Then if you want to test your work, you might use browser_action to launch the site, wait for the user's response confirming the site was launched along with a screenshot, then perhaps e.g., click a button to test functionality if needed, wait for the user's response confirming the button was clicked along with a screenshot of the new state, before finally closing the browser."
+			: ""
+	}
+- MCP operations should be used one at a time, similar to other tool usage. Wait for confirmation of success before proceeding with additional operations.
+
+====
+
+SYSTEM INFORMATION
+
+Operating System: ${os}
+Default Shell: ${shell}
+Home Directory: ${homeFormatted}
+Current Working Directory: ${cwdFormatted}
+
+====
+
+OBJECTIVE
+
+You accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.
+
+1. Analyze the user's task and set clear, achievable goals to accomplish it. Prioritize these goals in a logical order.
+2. Work through these goals sequentially, utilizing available tools one at a time as necessary. Each goal should correspond to a distinct step in your problem-solving process. You will be informed on the work completed and what's remaining as you go.
+3. Remember, you have extensive capabilities with access to a wide range of tools that can be used in powerful and clever ways as necessary to accomplish each goal. Before calling a tool, do some analysis within <thinking></thinking> tags. First, analyze the file structure provided in environment_details to gain context and insights for proceeding effectively. Then, think about which of the provided tools is the most relevant tool to accomplish the user's task. Next, go through each of the required parameters of the relevant tool and determine if the user has directly provided or given enough information to infer a value. When deciding if the parameter can be inferred, carefully consider all the context to see if it supports a specific value. If all of the required parameters are present or can be reasonably inferred, close the thinking tag and proceed with the tool use. BUT, if one of the values for a required parameter is missing, DO NOT invoke the tool (not even with fillers for the missing params) and instead, ask the user to provide the missing parameters using the ask_followup_question tool. DO NOT ask for more information on optional parameters if it is not provided.
+4. Once you've completed the user's task, you must use the attempt_completion tool to present the result of the task to the user. You may also provide a CLI command to showcase the result of your task; this can be particularly useful for web development tasks, where you can run e.g. \`open index.html\` to show the website you've built.
+5. The user may provide feedback, which you can use to make improvements and try again. But DO NOT continue in pointless back and forth conversations, i.e. don't end your responses with questions or offers for further assistance.
+${
+	userCustomInstructions
+		? `\n
+====
+
+USER'S CUSTOM INSTRUCTIONS
+
+The following additional instructions are provided by the user, and should be followed to the best of your ability without interfering with the TOOL USE guidelines.
+
+${userCustomInstructions}`
+		: ""
+}`
+}

evals/diff_editing/types.ts

@@ -0,0 +1,88 @@
+import { Anthropic } from "@anthropic-ai/sdk"
+import { ToolUseName, ToolParamName } from "../../src/core/assistant-message"
+
+export interface InputMessage {
+	role: "user" | "assistant"
+	text: string
+	images?: string[]
+}
+
+export interface ProcessedTestCase {
+	test_id: string
+	messages: Anthropic.Messages.MessageParam[]
+	file_contents: string
+	file_path: string
+	system_prompt_details: SystemPromptDetails
+	original_diff_edit_tool_call_message: string
+}
+
+export interface TestCase {
+	test_id: string
+	messages: InputMessage[]
+	file_contents: string
+	file_path: string
+	system_prompt_details: SystemPromptDetails
+	original_diff_edit_tool_call_message: string
+}
+
+export interface TestConfig {
+	model_id: string
+	system_prompt_name: string
+	number_of_runs: number
+	parsing_function: string
+	diff_edit_function: string
+	thinking_tokens_budget: number
+	replay: boolean
+}
+
+export interface SystemPromptDetails {
+	mcp_string: string
+	cwd_value: string
+	browser_use: boolean
+	width: number
+	height: number
+	os_value: string
+	shell_value: string
+	home_value: string
+	user_custom_instructions: string
+}
+
+export type ConstructSystemPromptFn = (
+	cwdFormatted: string,
+	supportsBrowserUse: boolean,
+	browserWidth: number,
+	browserHeight: number,
+	os: string,
+	shell: string,
+	homeFormatted: string,
+	mcpHubString: string,
+	userCustomInstructions: string,
+) => string
+
+export interface TestResult {
+	success: boolean
+	streamResult?: any
+	diffEdit?: string
+	toolCalls?: ExtractedToolCall[]
+	diffEditSuccess?: boolean
+	error?: string
+	errorString?: string
+}
+
+export interface ExtractedToolCall {
+	name: ToolUseName
+	input: Partial<Record<ToolParamName, string>>
+}
+
+export interface TestInput {
+	apiKey?: string
+	systemPrompt: string
+	messages: Anthropic.Messages.MessageParam[]
+	modelId: string
+	originalFile: string
+	originalFilePath: string
+	parsingFunction: string
+	diffEditFunction: string
+	thinkingBudgetTokens: number
+	originalDiffEditToolCallMessage?: string
+}

package.json

@@ -2,7 +2,7 @@
 	"name": "claude-dev",
 	"displayName": "Cline",
 	"description": "Autonomous coding agent right in your IDE, capable of creating/editing files, running commands, using the browser, and more with your permission every step of the way.",
-	"version": "3.17.9",
+	"version": "3.17.14",
 	"icon": "assets/icons/icon.png",
 	"engines": {
 		"vscode": "^1.84.0"
@@ -46,6 +46,58 @@
 	],
 	"main": "./dist/extension.js",
 	"contributes": {
+		"walkthroughs": [
+			{
+				"id": "ClineWalkthrough",
+				"title": "Meet Cline, your new coding partner",
+				"description": "Cline codes like a developer because it thinks like one. Here are 5 ways to put it to work:",
+				"steps": [
+					{
+						"id": "welcome",
+						"title": "Start with a Goal, Not Just a Prompt",
+						"description": "Tell Cline what you want to achieve. It plans, asks, and then codes, like a true partner.",
+						"media": {
+							"markdown": "walkthrough/step1.md"
+						}
+					},
+					{
+						"id": "learn",
+						"title": "Let Cline Learn Your Codebase",
+						"description": "Point Cline to your project. It builds understanding to make smart, context-aware changes.",
+						"media": {
+							"markdown": "walkthrough/step2.md"
+						}
+					},
+					{
+						"id": "advanced-features",
+						"title": "Always Use the Best AI Models",
+						"description": "Cline empowers you with State-of-the-Art AI, connecting to top models (Anthropic, Gemini, OpenAI & more).",
+						"media": {
+							"markdown": "walkthrough/step3.md"
+						}
+					},
+					{
+						"id": "mcp",
+						"title": "Extend with Powerful Tools (MCP)",
+						"description": "Connect to databases, APIs, or discover new capabilities in the MCP Marketplace.",
+						"media": {
+							"markdown": "walkthrough/step4.md"
+						}
+					},
+					{
+						"id": "getting-started",
+						"title": "You're Always in Control",
+						"description": "Review Cline's plans and diffs. Approve changes before they happen. No surprises.",
+						"media": {
+							"markdown": "walkthrough/step5.md"
+						},
+						"content": {
+							"path": "walkthrough/step5.md"
+						}
+					}
+				]
+			}
+		],
 		"viewsContainers": {
 			"activitybar": [
 				{
@@ -143,6 +195,11 @@
 				"command": "cline.improveCode",
 				"title": "Improve with Cline",
 				"category": "Cline"
+			},
+			{
+				"command": "cline.openWalkthrough",
+				"title": "Open Walkthrough",
+				"category": "Cline"
 			}
 		],
 		"keybindings": [
@@ -273,13 +330,13 @@
 		"watch:esbuild": "node esbuild.js --watch",
 		"watch:tsc": "tsc --noEmit --watch --project tsconfig.json",
 		"package": "npm run check-types && npm run build:webview && npm run lint && node esbuild.js --production",
-		"protos": "node proto/build-proto.js && node scripts/generate-server-setup.mjs",
-		"postprotos": "prettier src/shared/proto src/core/controller webview-ui/src/services src/standalone/server-setup.ts --write --log-level silent",
+		"protos": "node proto/build-proto.js && node scripts/generate-server-setup.mjs && node scripts/generate-host-bridge-client.mjs",
+		"postprotos": "prettier src/shared/proto src/core/controller src/hosts/ webview-ui/src/services src/generated --write --log-level warn",
 		"compile-tests": "node ./scripts/build-tests.js",
 		"watch-tests": "tsc -p . -w --outDir out",
 		"pretest": "npm run compile-tests && npm run compile && npm run compile-standalone && npm run lint",
 		"check-types": "npm run protos && tsc --noEmit",
-		"lint": "eslint src --ext ts && eslint webview-ui/src --ext ts && cd webview-ui && npm run lint",
+		"lint": "eslint src --ext ts && eslint webview-ui/src --ext ts && buf lint && cd webview-ui && npm run lint",
 		"format": "prettier . --check",
 		"format:fix": "prettier . --write",
 		"test": "npm-run-all test:unit test:integration",
@@ -301,7 +358,13 @@
 		"docs:rename-file": "cd docs && mintlify rename",
 		"report-issue": "node scripts/report-issue.js"
 	},
+	"lint-staged": {
+		"*": [
+			"prettier --write --ignore-unknown --log-level=log"
+		]
+	},
 	"devDependencies": {
+		"@bufbuild/buf": "^1.54.0",
 		"@changesets/cli": "^2.27.12",
 		"@types/chai": "^5.0.1",
 		"@types/clone-deep": "^4.0.4",
@@ -327,6 +390,7 @@
 		"eslint-plugin-eslint-rules": "file:eslint-rules",
 		"grpc-tools": "^1.13.0",
 		"husky": "^9.1.7",
+		"lint-staged": "^16.1.0",
 		"mintlify": "^4.0.515",
 		"npm-run-all": "^4.1.5",
 		"prettier": "^3.3.3",
@@ -340,14 +404,14 @@
 		"typescript": "^5.4.5"
 	},
 	"dependencies": {
-		"@anthropic-ai/bedrock-sdk": "^0.12.4",
 		"@anthropic-ai/sdk": "^0.37.0",
 		"@anthropic-ai/vertex-sdk": "^0.6.4",
-		"@aws-sdk/client-bedrock-runtime": "^3.821.0",
+		"@aws-sdk/client-bedrock-runtime": "^3.826.0",
+		"@aws-sdk/credential-providers": "^3.826.0",
 		"@bufbuild/protobuf": "^2.2.5",
 		"@cerebras/cerebras_cloud_sdk": "^1.35.0",
 		"@google-cloud/vertexai": "^1.9.3",
-		"@google/genai": "^0.13.0",
+		"@google/genai": "1.0.0",
 		"@grpc/grpc-js": "^1.9.15",
 		"@grpc/reflection": "^1.0.4",
 		"@mistralai/mistralai": "^1.5.0",
@@ -384,6 +448,7 @@
 		"jschardet": "^3.1.4",
 		"mammoth": "^1.8.0",
 		"monaco-vscode-textmate-theme-converter": "^0.1.7",
+		"nice-grpc": "^2.1.12",
 		"ollama": "^0.5.13",
 		"open": "^10.1.2",
 		"open-graph-scraper": "^6.9.0",

proto/build-proto.js

@@ -10,49 +10,30 @@ import os from "os"
 
 import { createRequire } from "module"
 const require = createRequire(import.meta.url)
-const protoc = path.join(require.resolve("grpc-tools"), "../bin/protoc")
+const PROTOC = path.join(require.resolve("grpc-tools"), "../bin/protoc")
 
-// Check for Apple Silicon compatibility
-function checkAppleSiliconCompatibility() {
-	// Only run check on macOS
-	if (process.platform !== "darwin") {
-		return
-	}
-
-	// Check if running on Apple Silicon
-	const cpuArchitecture = os.arch()
-	if (cpuArchitecture === "arm64") {
-		try {
-			// Check if Rosetta is installed
-			const rosettaCheck = execSync('/usr/bin/pgrep oahd || echo "NOT_INSTALLED"').toString().trim()
-
-			if (rosettaCheck === "NOT_INSTALLED") {
-				console.log(chalk.yellow("Detected Apple Silicon (ARM64) architecture."))
-				console.log(
-					chalk.red("Rosetta 2 is NOT installed. The npm version of protoc is not compatible with Apple Silicon."),
-				)
-				console.log(chalk.cyan("Please install Rosetta 2 using the following command:"))
-				console.log(chalk.cyan("  softwareupdate --install-rosetta --agree-to-license"))
-				console.log(chalk.red("Aborting build process."))
-				process.exit(1)
-			} else {
-				console.log(chalk.green("Rosetta 2 is installed. Continuing with build."))
-			}
-		} catch (error) {
-			console.log(chalk.yellow("Could not determine Rosetta installation status. Proceeding anyway."))
-		}
-	}
-}
-
-const __filename = fileURLToPath(import.meta.url)
-const SCRIPT_DIR = path.dirname(__filename)
+const SCRIPT_DIR = path.dirname(fileURLToPath(import.meta.url))
 const ROOT_DIR = path.resolve(SCRIPT_DIR, "..")
 
+const TS_OUT_DIR = path.join(ROOT_DIR, "src/shared/proto")
+const GRPC_JS_OUT_DIR = path.join(ROOT_DIR, "src/generated/grpc-js")
+const NICE_JS_OUT_DIR = path.join(ROOT_DIR, "src/generated/nice-grpc")
+const DESCRIPTOR_OUT_DIR = path.join(ROOT_DIR, "dist-standalone/proto")
+
 const isWindows = process.platform === "win32"
-const tsProtoPlugin = isWindows
+const TS_PROTO_PLUGIN = isWindows
 	? path.join(ROOT_DIR, "node_modules", ".bin", "protoc-gen-ts_proto.cmd") // Use the .bin directory path for Windows
 	: require.resolve("ts-proto/protoc-gen-ts_proto")
 
+const TS_PROTO_OPTIONS = [
+	"env=node",
+	"esModuleInterop=true",
+	"outputServices=generic-definitions", // output generic ServiceDefinitions
+	"outputIndex=true", // output an index file for each package which exports all protos in the package.
+	"useOptionals=messages", // Message fields are optional, scalars are not.
+	"useDate=false", // Timestamp fields will not be automatically converted to Date.
+]
+
 // List of gRPC services
 // To add a new service, simply add it to this map and run this script
 // The service handler will be automatically discovered and used by grpc-handler.ts
@@ -70,7 +51,16 @@ const serviceNameMap = {
 	ui: "cline.UiService",
 	// Add new services here - no other code changes needed!
 }
-const serviceDirs = Object.keys(serviceNameMap).map((serviceKey) => path.join(ROOT_DIR, "src", "core", "controller", serviceKey))
+const serviceDirs = Object.keys(serviceNameMap).map((serviceKey) => path.join(ROOT_DIR, "src/core/controller", serviceKey))
+
+// List of host gRPC services (IDE API bridge)
+// These services are implemented in the IDE extension and called by the standalone Cline Core
+const hostServiceNameMap = {
+	uri: "host.UriService",
+	watch: "host.WatchService",
+	// Add new host services here
+}
+const hostServiceDirs = Object.keys(hostServiceNameMap).map((serviceKey) => path.join(ROOT_DIR, "src/hosts/vscode", serviceKey))
 
 async function main() {
 	console.log(chalk.bold.blue("Starting Protocol Buffer code generation..."))
@@ -78,68 +68,73 @@ async function main() {
 	// Check for Apple Silicon compatibility before proceeding
 	checkAppleSiliconCompatibility()
 
-	// Define output directories
-	const TS_OUT_DIR = path.join(ROOT_DIR, "src", "shared", "proto")
-
-	// Create output directory if it doesn't exist
-	await fs.mkdir(TS_OUT_DIR, { recursive: true })
-
-	// Clean up existing generated files
-	console.log(chalk.cyan("Cleaning up existing generated TypeScript files..."))
-	const existingFiles = await globby("**/*.ts", { cwd: TS_OUT_DIR })
-	for (const file of existingFiles) {
-		await fs.unlink(path.join(TS_OUT_DIR, file))
+	// Create output directories if they don't exist
+	for (const dir of [TS_OUT_DIR, GRPC_JS_OUT_DIR, NICE_JS_OUT_DIR, DESCRIPTOR_OUT_DIR]) {
+		await fs.mkdir(dir, { recursive: true })
 	}
 
+	await cleanup()
+
 	// Check for missing proto files for services in serviceNameMap
 	await ensureProtoFilesExist()
 
 	// Process all proto files
-	console.log(chalk.cyan("Processing proto files from"), SCRIPT_DIR)
-	const protoFiles = await globby("*.proto", { cwd: SCRIPT_DIR, realpath: true })
+	const protoFiles = await globby("**/*.proto", { cwd: SCRIPT_DIR, realpath: true })
+	console.log(chalk.cyan(`Processing ${protoFiles.length} proto files from`), SCRIPT_DIR)
 
-	// Build the protoc command with proper path handling for cross-platform
-	const tsProtocCommand = [
-		protoc,
-		`--proto_path="${SCRIPT_DIR}"`,
-		`--plugin=protoc-gen-ts_proto="${tsProtoPlugin}"`,
-		`--ts_proto_out="${TS_OUT_DIR}"`,
-		"--ts_proto_opt=outputServices=generic-definitions,env=node,esModuleInterop=true,useDate=false,useOptionals=messages",
-		...protoFiles,
-	].join(" ")
-	try {
-		console.log(chalk.cyan(`Generating TypeScript code for:\n${protoFiles.join("\n")}...`))
-		execSync(tsProtocCommand, { stdio: "inherit" })
-	} catch (error) {
-		console.error(chalk.red("Error generating TypeScript for proto files:"), error)
-		process.exit(1)
-	}
+	tsProtoc(TS_OUT_DIR, protoFiles, TS_PROTO_OPTIONS)
+	// grpc-js is used to generate service impls for the ProtoBus service.
+	tsProtoc(GRPC_JS_OUT_DIR, protoFiles, ["outputServices=grpc-js,outputClientImpl=false", ...TS_PROTO_OPTIONS])
+	// nice-js is used for the Host Bridge client impls because it uses promises.
+	tsProtoc(NICE_JS_OUT_DIR, protoFiles, ["outputServices=nice-grpc,useExactTypes=false", ...TS_PROTO_OPTIONS])
 
-	const descriptorOutDir = path.join(ROOT_DIR, "dist-standalone", "proto")
-	await fs.mkdir(descriptorOutDir, { recursive: true })
-
-	const descriptorFile = path.join(descriptorOutDir, "descriptor_set.pb")
+	const descriptorFile = path.join(DESCRIPTOR_OUT_DIR, "descriptor_set.pb")
 	const descriptorProtocCommand = [
-		protoc,
+		PROTOC,
 		`--proto_path="${SCRIPT_DIR}"`,
 		`--descriptor_set_out="${descriptorFile}"`,
 		"--include_imports",
 		...protoFiles,
 	].join(" ")
 	try {
-		console.log(chalk.cyan("Generating descriptor set..."))
+		log_verbose(chalk.cyan("Generating descriptor set..."))
 		execSync(descriptorProtocCommand, { stdio: "inherit" })
 	} catch (error) {
 		console.error(chalk.red("Error generating descriptor set for proto file:"), error)
 		process.exit(1)
 	}
 
-	console.log(chalk.green("Protocol Buffer code generation completed successfully."))
-	console.log(chalk.green(`TypeScript files generated in: ${TS_OUT_DIR}`))
+	log_verbose(chalk.green("Protocol Buffer code generation completed successfully."))
+	log_verbose(chalk.green(`TypeScript files generated in: ${TS_OUT_DIR}`))
 
 	await generateMethodRegistrations()
+	await generateHostMethodRegistrations()
 	await generateServiceConfig()
+	await generateHostServiceConfig()
 	await generateGrpcClientConfig()
+	await generateHostGrpcClientConfig()
+
+	console.log(chalk.bold.blue("Finished Protocol Buffer code generation."))
+}
+
+async function tsProtoc(outDir, protoFiles, protoOptions) {
+	// Build the protoc command with proper path handling for cross-platform
+	const command = [
+		PROTOC,
+		`--proto_path="${SCRIPT_DIR}"`,
+		`--plugin=protoc-gen-ts_proto="${TS_PROTO_PLUGIN}"`,
+		`--ts_proto_out="${outDir}"`,
+		`--ts_proto_opt=${protoOptions.join(",")} `,
+		...protoFiles.map((s) => `"${s}"`),
+	].join(" ")
+	try {
+		log_verbose(chalk.cyan(`Generating TypeScript code in ${outDir} for:\n${protoFiles.join("\n")}...`))
+		log_verbose(command)
+		execSync(command, { stdio: "inherit" })
+	} catch (error) {
+		console.error(chalk.red("Error generating TypeScript for proto files:"), error)
+		process.exit(1)
+	}
 }
 
 /**
@@ -147,14 +142,14 @@ async function main() {
  * This eliminates the need for manual imports and client creation in grpc-client.ts
  */
 async function generateGrpcClientConfig() {
-	console.log(chalk.cyan("Generating gRPC client configuration..."))
+	log_verbose(chalk.cyan("Generating gRPC client configuration..."))
 
 	const serviceImports = []
 	const serviceClientCreations = []
 	const serviceExports = []
 
 	// Process each service in the serviceNameMap
-	for (const [dirName, fullServiceName] of Object.entries(serviceNameMap)) {
+	for (const [dirName, _fullServiceName] of Object.entries(serviceNameMap)) {
 		const capitalizedName = dirName.charAt(0).toUpperCase() + dirName.slice(1)
 
 		// Add import statement
@@ -182,9 +177,9 @@ export {
 	${serviceExports.join(",\n\t")}
 }`
 
-	const configPath = path.join(ROOT_DIR, "webview-ui", "src", "services", "grpc-client.ts")
-	await fs.writeFile(configPath, content)
-	console.log(chalk.green(`Generated gRPC client at ${configPath}`))
+	const filePath = path.join(ROOT_DIR, "webview-ui/src/services/grpc-client.ts")
+	await writeFileWithMkdirs(filePath, content)
+	log_verbose(chalk.green(`Generated gRPC client at ${filePath}`))
 }
 
 /**
@@ -194,7 +189,7 @@ export {
  * @returns Map of service names to their streaming methods
  */
 async function parseProtoForStreamingMethods(protoFiles, scriptDir) {
-	console.log(chalk.cyan("Parsing proto files for streaming methods..."))
+	log_verbose(chalk.cyan("Parsing proto files for streaming methods..."))
 
 	// Map of service name to array of streaming method names
 	const streamingMethodsMap = new Map()
@@ -246,28 +241,18 @@ async function parseProtoForStreamingMethods(protoFiles, scriptDir) {
 }
 
 async function generateMethodRegistrations() {
-	console.log(chalk.cyan("Generating method registration files..."))
+	log_verbose(chalk.cyan("Generating method registration files..."))
 
 	// Parse proto files for streaming methods
 	const protoFiles = await globby("*.proto", { cwd: SCRIPT_DIR })
 	const streamingMethodsMap = await parseProtoForStreamingMethods(protoFiles, SCRIPT_DIR)
 
 	for (const serviceDir of serviceDirs) {
-		try {
-			await fs.access(serviceDir)
-		} catch (error) {
-			console.log(chalk.cyan(`Creating directory ${serviceDir} for new service`))
-			await fs.mkdir(serviceDir, { recursive: true })
-		}
-
 		const serviceName = path.basename(serviceDir)
-		const registryFile = path.join(serviceDir, "methods.ts")
-		const indexFile = path.join(serviceDir, "index.ts")
-
 		const fullServiceName = serviceNameMap[serviceName]
 		const streamingMethods = streamingMethodsMap.get(fullServiceName) || []
 
-		console.log(chalk.cyan(`Generating method registrations for ${serviceName}...`))
+		log_verbose(chalk.cyan(`Generating method registrations for ${serviceName}...`))
 
 		// Get all TypeScript files in the service directory
 		const files = await globby("*.ts", { cwd: serviceDir })
@@ -282,7 +267,7 @@ async function generateMethodRegistrations() {
 // Import all method implementations
 import { registerMethod } from "./index"\n`
 
-		// Add imports for all implementation files
+		// Import implementations directly
 		for (const file of implementationFiles) {
 			const baseName = path.basename(file, ".ts")
 			methodsContent += `import { ${baseName} } from "./${baseName}"\n`
@@ -319,8 +304,9 @@ export function registerAllMethods(): void {
 		methodsContent += `}`
 
 		// Write the methods.ts file
-		await fs.writeFile(registryFile, methodsContent)
-		console.log(chalk.green(`Generated ${registryFile}`))
+		const registryFile = path.join(serviceDir, "methods.ts")
+		await writeFileWithMkdirs(registryFile, methodsContent)
+		log_verbose(chalk.green(`Generated ${registryFile}`))
 
 		// Generate index.ts file
 		const capitalizedServiceName = serviceName.charAt(0).toUpperCase() + serviceName.slice(1)
@@ -348,11 +334,12 @@ export const isStreamingMethod = ${serviceName}Service.isStreamingMethod
 registerAllMethods()`
 
 		// Write the index.ts file
-		await fs.writeFile(indexFile, indexContent)
-		console.log(chalk.green(`Generated ${indexFile}`))
+		const indexFile = path.join(serviceDir, "index.ts")
+		await writeFileWithMkdirs(indexFile, indexContent)
+		log_verbose(chalk.green(`Generated ${indexFile}`))
 	}
 
-	console.log(chalk.green("Method registration files generated successfully."))
+	log_verbose(chalk.green("Method registration files generated successfully."))
 }
 
 /**
@@ -360,7 +347,7 @@ registerAllMethods()`
  * This eliminates the need for manual switch/case statements in grpc-handler.ts
  */
 async function generateServiceConfig() {
-	console.log(chalk.cyan("Generating service configuration file..."))
+	log_verbose(chalk.cyan("Generating service configuration file..."))
 
 	const serviceImports = []
 	const serviceConfigs = []
@@ -399,9 +386,9 @@ export interface ServiceHandlerConfig {
 export const serviceHandlers: Record<string, ServiceHandlerConfig> = {${serviceConfigs.join(",")}
 };`
 
-	const configPath = path.join(ROOT_DIR, "src", "core", "controller", "grpc-service-config.ts")
-	await fs.writeFile(configPath, content)
-	console.log(chalk.green(`Generated service configuration at ${configPath}`))
+	const configPath = path.join(ROOT_DIR, "src/core/controller/grpc-service-config.ts")
+	await writeFileWithMkdirs(configPath, content)
+	log_verbose(chalk.green(`Generated service configuration at ${configPath}`))
 }
 
 /**
@@ -409,7 +396,7 @@ export const serviceHandlers: Record<string, ServiceHandlerConfig> = {${serviceC
  * If a .proto file doesn't exist, create a template file
  */
 async function ensureProtoFilesExist() {
-	console.log(chalk.cyan("Checking for missing proto files..."))
+	log_verbose(chalk.cyan("Checking for missing proto files..."))
 
 	// Get existing proto files
 	const existingProtoFiles = await globby("*.proto", { cwd: SCRIPT_DIR })
@@ -418,7 +405,7 @@ async function ensureProtoFilesExist() {
 	// Check each service in serviceNameMap
 	for (const [serviceName, fullServiceName] of Object.entries(serviceNameMap)) {
 		if (!existingProtoServices.includes(serviceName)) {
-			console.log(chalk.yellow(`Creating template proto file for ${serviceName}...`))
+			log_verbose(chalk.yellow(`Creating template proto file for ${serviceName}...`))
 
 			// Extract service class name from full name (e.g., "cline.ModelsService" -> "ModelsService")
 			const serviceClassName = fullServiceName.split(".").pop()
@@ -451,11 +438,271 @@ service ${serviceClassName} {
 			// Write the template proto file
 			const protoFilePath = path.join(SCRIPT_DIR, `${serviceName}.proto`)
 			await fs.writeFile(protoFilePath, protoContent)
-			console.log(chalk.green(`Created template proto file at ${protoFilePath}`))
+			log_verbose(chalk.green(`Created template proto file at ${protoFilePath}`))
 		}
 	}
 }
 
+/**
+ * Generate method registration files for host services
+ */
+async function generateHostMethodRegistrations() {
+	log_verbose(chalk.cyan("Generating host method registration files..."))
+
+	// Parse proto files for streaming methods
+	const hostProtoFiles = await globby("*.proto", { cwd: path.join(SCRIPT_DIR, "host") })
+	const streamingMethodsMap = await parseProtoForStreamingMethods(hostProtoFiles, path.join(SCRIPT_DIR, "host"))
+
+	for (const serviceDir of hostServiceDirs) {
+		const serviceName = path.basename(serviceDir)
+		const fullServiceName = hostServiceNameMap[serviceName]
+		const streamingMethods = streamingMethodsMap.get(fullServiceName) || []
+
+		log_verbose(chalk.cyan(`Generating method registrations for host ${serviceName}...`))
+
+		// Get all TypeScript files in the service directory
+		const files = await globby("*.ts", { cwd: serviceDir })
+
+		// Filter out index.ts and methods.ts
+		const implementationFiles = files.filter((file) => file !== "index.ts" && file !== "methods.ts")
+
+		// Create the methods.ts file with header
+		let methodsContent = `// AUTO-GENERATED FILE - DO NOT MODIFY DIRECTLY
+// Generated by proto/build-proto.js
+
+// Import all method implementations
+import { registerMethod } from "./index"\n`
+
+		// Import implementations directly
+		for (const file of implementationFiles) {
+			const baseName = path.basename(file, ".ts")
+			methodsContent += `import { ${baseName} } from "./${baseName}"\n`
+		}
+
+		// Add streaming methods information
+		if (streamingMethods.length > 0) {
+			methodsContent += `\n// Streaming methods for this service
+export const streamingMethods = ${JSON.stringify(
+				streamingMethods.map((m) => m.name),
+				null,
+				2,
+			)}\n`
+		}
+
+		// Add registration function
+		methodsContent += `\n// Register all ${serviceName} service methods
+export function registerAllMethods(): void {
+\t// Register each method with the registry\n`
+
+		// Add registration statements
+		for (const file of implementationFiles) {
+			const baseName = path.basename(file, ".ts")
+			const isStreaming = streamingMethods.some((m) => m.name === baseName)
+
+			if (isStreaming) {
+				methodsContent += `\tregisterMethod("${baseName}", ${baseName}, { isStreaming: true })\n`
+			} else {
+				methodsContent += `\tregisterMethod("${baseName}", ${baseName})\n`
+			}
+		}
+
+		// Close the function
+		methodsContent += `}`
+
+		// Write the methods.ts file
+		const registryFile = path.join(serviceDir, "methods.ts")
+		await writeFileWithMkdirs(registryFile, methodsContent)
+		log_verbose(chalk.green(`Generated ${registryFile}`))
+
+		// Generate index.ts file
+		const capitalizedServiceName = serviceName.charAt(0).toUpperCase() + serviceName.slice(1)
+		const indexContent = `// AUTO-GENERATED FILE - DO NOT MODIFY DIRECTLY
+// Generated by proto/build-proto.js
+
+import { createServiceRegistry, ServiceMethodHandler, StreamingMethodHandler } from "../host-grpc-service"
+import { StreamingResponseHandler } from "../host-grpc-handler"
+import { registerAllMethods } from "./methods"
+
+// Create ${serviceName} service registry
+const ${serviceName}Service = createServiceRegistry("${serviceName}")
+
+// Export the method handler types and registration function
+export type ${capitalizedServiceName}MethodHandler = ServiceMethodHandler
+export type ${capitalizedServiceName}StreamingMethodHandler = StreamingMethodHandler
+export const registerMethod = ${serviceName}Service.registerMethod
+
+// Export the request handlers
+export const handle${capitalizedServiceName}ServiceRequest = ${serviceName}Service.handleRequest
+export const handle${capitalizedServiceName}ServiceStreamingRequest = ${serviceName}Service.handleStreamingRequest
+export const isStreamingMethod = ${serviceName}Service.isStreamingMethod
+
+// Register all ${serviceName} methods
+registerAllMethods()`
+
+		// Write the index.ts file
+		const indexFile = path.join(serviceDir, "index.ts")
+		await writeFileWithMkdirs(indexFile, indexContent)
+		log_verbose(chalk.green(`Generated ${indexFile}`))
+	}
+
+	log_verbose(chalk.green("Host method registration files generated successfully."))
+}
+
+/**
+ * Generate a service configuration file for host services
+ */
+async function generateHostServiceConfig() {
+	log_verbose(chalk.cyan("Generating host service configuration file..."))
+
+	const serviceImports = []
+	const serviceConfigs = []
+
+	// Add all services from the hostServiceNameMap
+	for (const [dirName, fullServiceName] of Object.entries(hostServiceNameMap)) {
+		const capitalizedName = dirName.charAt(0).toUpperCase() + dirName.slice(1)
+		serviceImports.push(
+			`import { handle${capitalizedName}ServiceRequest, handle${capitalizedName}ServiceStreamingRequest } from "./${dirName}/index"`,
+		)
+		serviceConfigs.push(`
+  "${fullServiceName}": {
+    requestHandler: handle${capitalizedName}ServiceRequest,
+    streamingHandler: handle${capitalizedName}ServiceStreamingRequest
+  }`)
+	}
+
+	const content = `// AUTO-GENERATED FILE - DO NOT MODIFY DIRECTLY
+// Generated by proto/build-proto.js
+
+import { StreamingResponseHandler } from "./host-grpc-handler"
+${serviceImports.join("\n")}
+
+/**
+ * Configuration for a host service handler
+ */
+export interface HostServiceHandlerConfig {
+  requestHandler: (method: string, message: any) => Promise<any>;
+  streamingHandler: (method: string, message: any, responseStream: StreamingResponseHandler, requestId?: string) => Promise<void>;
+}
+
+/**
+ * Map of host service names to their handler configurations
+ */
+export const hostServiceHandlers: Record<string, HostServiceHandlerConfig> = {${serviceConfigs.join(",")}
+};`
+
+	const filePath = path.join(ROOT_DIR, "src/hosts/vscode/host-grpc-service-config.ts")
+	await writeFileWithMkdirs(filePath, content)
+	log_verbose(chalk.green(`Generated host service configuration at ${filePath}`))
+}
+
+/**
+ * Generate a gRPC client configuration file for host services
+ */
+async function generateHostGrpcClientConfig() {
+	log_verbose(chalk.cyan("Generating host gRPC client configuration..."))
+	const clients = []
+
+	// Process each service in the hostServiceNameMap
+	for (const [_dirName, fullServiceName] of Object.entries(hostServiceNameMap)) {
+		const serviceName = fullServiceName.replace(/.*\./, "")
+		clients.push(`${serviceName}Client: createGrpcClient(${fullServiceName}Definition)`)
+	}
+
+	// Generate the file content
+	const content = `// AUTO-GENERATED FILE - DO NOT MODIFY DIRECTLY
+// Generated by proto/build-proto.js
+
+import { createGrpcClient } from "@hosts/vscode/client/host-grpc-client-base"
+import { HostBridgeClientProvider } from "@/hosts/host-bridge-client"
+import * as host from "@shared/proto/index.host"
+
+export const vscodeHostBridgeClient: HostBridgeClientProvider = {
+	${clients.join(",\n\t")}
+}`
+
+	const filePath = path.join(ROOT_DIR, "src/generated/hosts/vscode/client/host-grpc-client.ts")
+	await writeFileWithMkdirs(filePath, content)
+	log_verbose(chalk.green(`Generated host gRPC client at ${filePath}`))
+}
+
+async function cleanup() {
+	// Clean up existing generated files
+	log_verbose(chalk.cyan("Cleaning up existing generated TypeScript files..."))
+	const existingFiles = await globby("**/*.ts", { cwd: TS_OUT_DIR })
+	for (const file of existingFiles) {
+		await fs.unlink(path.join(TS_OUT_DIR, file))
+	}
+	await rmdir(path.join(ROOT_DIR, "src/generated"))
+
+	// Clean up generated files that were moved.
+	await fs.rm(path.join(ROOT_DIR, "src/hosts/vscode/client/host-grpc-client.ts"), { force: true })
+	await fs.rm(path.join(ROOT_DIR, "src/standalone/services/host-grpc-client.ts"), { force: true })
+	await rmdir(path.join(ROOT_DIR, "src/standalone/services"))
+
+	await fs.rm(path.join(ROOT_DIR, "hosts/vscode"), { force: true, recursive: true })
+	await rmdir(path.join(ROOT_DIR, "hosts"))
+
+	await fs.rm(path.join(ROOT_DIR, "src/standalone/server-setup.ts"), { force: true })
+}
+
+/**
+ * Write `contents` to `filePath`, creating any necessary directories in `filePath`.
+ */
+async function writeFileWithMkdirs(filePath, content) {
+	await fs.mkdir(path.dirname(filePath), { recursive: true })
+	await fs.writeFile(filePath, content)
+}
+
+/**
+ * Remove an empty dir, do nothing if the directory doesn't exist or is not empty.
+ */
+async function rmdir(path) {
+	try {
+		await fs.rmdir(path)
+	} catch (error) {
+		if (error.code !== "ENOTEMPTY" && error.code !== "ENOENT") {
+			// Only re-throw if it's not "not empty" or "doesn't exist"
+			throw error
+		}
+	}
+}
+
+// Check for Apple Silicon compatibility
+function checkAppleSiliconCompatibility() {
+	// Only run check on macOS
+	if (process.platform !== "darwin") {
+		return
+	}
+
+	// Check if running on Apple Silicon
+	const cpuArchitecture = os.arch()
+	if (cpuArchitecture === "arm64") {
+		try {
+			// Check if Rosetta is installed
+			const rosettaCheck = execSync('/usr/bin/pgrep oahd || echo "NOT_INSTALLED"').toString().trim()
+
+			if (rosettaCheck === "NOT_INSTALLED") {
+				console.log(chalk.yellow("Detected Apple Silicon (ARM64) architecture."))
+				console.log(
+					chalk.red("Rosetta 2 is NOT installed. The npm version of protoc is not compatible with Apple Silicon."),
+				)
+				console.log(chalk.cyan("Please install Rosetta 2 using the following command:"))
+				console.log(chalk.cyan("  softwareupdate --install-rosetta --agree-to-license"))
+				console.log(chalk.red("Aborting build process."))
+				process.exit(1)
+			}
+		} catch (error) {
+			console.log(chalk.yellow("Could not determine Rosetta installation status. Proceeding anyway."))
+		}
+	}
+}
+
+function log_verbose(s) {
+	if (process.argv.includes("-v") || process.argv.includes("--verbose")) {
+		console.log(s)
+	}
+}
+
 // Run the main function
 main().catch((error) => {
 	console.error(chalk.red("Error:"), error)

proto/common.proto

@@ -63,3 +63,8 @@ message StringArrays {
   repeated string values1 = 1;
   repeated string values2 = 2;
 }
+
+message KeyValuePair {
+  string key = 1;
+  string value = 2;
+}

proto/file.proto

@@ -55,6 +55,12 @@ service FileService {
 
   // Opens a task's conversation history file on disk
   rpc openTaskHistory(StringRequest) returns (Empty);
+  
+  // Toggles a workflow on or off
+  rpc toggleWorkflow(ToggleWorkflowRequest) returns (ClineRulesToggles);
+
+  // Subscribe to workspace file updates
+  rpc subscribeToWorkspaceUpdates(EmptyRequest) returns (stream StringArray);
 }
 
 // Response for refreshRules operation
@@ -161,3 +167,11 @@ message ToggleCursorRuleRequest {
   string rule_path = 2;   // Path to the rule file
   bool enabled = 3;       // Whether to enable or disable the rule
 }
+
+// Request to toggle a workflow on or off
+message ToggleWorkflowRequest {
+  Metadata metadata = 1;
+  string workflow_path = 2;
+  bool enabled = 3;
+  bool is_global = 4;
+}

proto/host/uri.proto

@@ -0,0 +1,36 @@
+syntax = "proto3";
+
+package host;
+option java_package = "bot.cline.host.proto";
+option java_multiple_files = true;
+
+import "common.proto";
+
+// UriService provides methods for working with URIs in the IDE
+service UriService {
+  // Create a new file URI from a file path
+  rpc file(cline.StringRequest) returns (Uri);
+  
+  // Join a URI with additional path segments
+  rpc joinPath(JoinPathRequest) returns (Uri);
+  
+  // Parse a string URI into a Uri object
+  rpc parse(cline.StringRequest) returns (Uri);
+}
+
+// Uri represents a URI in the IDE
+message Uri {
+  string scheme = 1;
+  string authority = 2;
+  string path = 3;
+  string query = 4;
+  string fragment = 5;
+  string fs_path = 6;
+}
+
+// Request for joining path segments to a URI
+message JoinPathRequest {
+  cline.Metadata metadata = 1;
+  Uri base = 2;
+  repeated string path_segments = 3;
+}

proto/host/watch.proto

@@ -0,0 +1,32 @@
+syntax = "proto3";
+
+package host;
+option java_package = "bot.cline.host.proto";
+option java_multiple_files = true;
+
+import "common.proto";
+
+// WatchService provides methods for watching files in the IDE
+service WatchService {
+  // Subscribe to file changes
+  rpc subscribeToFile(SubscribeToFileRequest) returns (stream FileChangeEvent);
+}
+
+// Request to subscribe to file changes
+message SubscribeToFileRequest {
+  cline.Metadata metadata = 1;
+  string path = 2;
+}
+
+// Event representing a file change
+message FileChangeEvent {
+  enum ChangeType {
+    CREATED = 0;
+    CHANGED = 1;
+    DELETED = 2;
+  }
+  
+  string path = 1;
+  ChangeType type = 2;
+  string content = 3; // Optional content of the file after change
+}

proto/mcp.proto

@@ -16,6 +16,13 @@ service McpService {
   rpc toggleToolAutoApprove(ToggleToolAutoApproveRequest) returns (McpServers);
   rpc refreshMcpMarketplace(EmptyRequest) returns (McpMarketplaceCatalog);
   rpc openMcpSettings(EmptyRequest) returns (Empty);
+  
+  // Subscribe to MCP marketplace catalog updates
+  rpc subscribeToMcpMarketplaceCatalog(EmptyRequest) returns (stream McpMarketplaceCatalog);
+  rpc getLatestMcpServers(Empty) returns (McpServers);
+  
+  // Subscribe to MCP server updates
+  rpc subscribeToMcpServers(EmptyRequest) returns (stream McpServers);
 }
 
 message ToggleMcpServerRequest {

proto/models.proto

@@ -20,32 +20,61 @@ service ModelsService {
   rpc refreshOpenAiModels(OpenAiModelsRequest) returns (StringArray);
   // Refreshes and returns Requesty models
   rpc refreshRequestyModels(EmptyRequest) returns (OpenRouterCompatibleModelInfo);
+  // Subscribe to OpenRouter models updates
+  rpc subscribeToOpenRouterModels(EmptyRequest) returns (stream OpenRouterCompatibleModelInfo);
+  // Updates API configuration
+  rpc updateApiConfigurationProto(UpdateApiConfigurationRequest) returns (Empty);
 }
 
 // List of VS Code LM models
 message VsCodeLmModelsArray {
-  repeated VsCodeLmModel models = 1;
+  repeated LanguageModelChatSelector models = 1;
 }
 
-// Structure representing a VS Code LM model
-message VsCodeLmModel {
-  string vendor = 1;
-  string family = 2;
-  string version = 3;
-  string id = 4;
+// Structure representing a language model chat selector
+message LanguageModelChatSelector {
+  optional string vendor = 1;
+  optional string family = 2;
+  optional string version = 3;
+  optional string id = 4;
+}
+
+// Price tier for tiered pricing models
+message PriceTier {
+  int32 token_limit = 1;  // Upper limit (inclusive) of input tokens for this price
+  double price = 2;       // Price per million tokens for this tier
+}
+
+// Thinking configuration for models that support thinking/reasoning
+message ThinkingConfig {
+  optional int32 max_budget = 1;                    // Max allowed thinking budget tokens
+  optional double output_price = 2;                 // Output price per million tokens when budget > 0
+  repeated PriceTier output_price_tiers = 3;        // Optional: Tiered output price when budget > 0
+}
+
+// Model tier for tiered pricing structures
+message ModelTier {
+  int32 context_window = 1;
+  optional double input_price = 2;
+  optional double output_price = 3;
+  optional double cache_writes_price = 4;
+  optional double cache_reads_price = 5;
 }
 
 // For OpenRouterCompatibleModelInfo structure in OpenRouterModels
 message OpenRouterModelInfo {
-  int32 max_tokens = 1;
-  int32 context_window = 2;
-  bool supports_images = 3;
+  optional int32 max_tokens = 1;
+  optional int32 context_window = 2;
+  optional bool supports_images = 3;
   bool supports_prompt_cache = 4;
-  double input_price = 5;
-  double output_price = 6;
-  double cache_writes_price = 7;
-  double cache_reads_price = 8;
-  string description = 9;
+  optional double input_price = 5;
+  optional double output_price = 6;
+  optional double cache_writes_price = 7;
+  optional double cache_reads_price = 8;
+  optional string description = 9;
+  optional ThinkingConfig thinking_config = 10;
+  optional bool supports_global_endpoint = 11;
+  repeated ModelTier tiers = 12;
 }
 
 // Shared response message for model information
@@ -56,6 +85,155 @@ message OpenRouterCompatibleModelInfo {
 // Request for fetching OpenAI models
 message OpenAiModelsRequest {
   Metadata metadata = 1;
-  string baseUrl = 2;
-  string apiKey = 3;
+  string base_url = 2;
+  string api_key = 3;
 }
+
+// Request for updating API configuration
+message UpdateApiConfigurationRequest {
+  Metadata metadata = 1;
+  ModelsApiConfiguration api_configuration = 2;
+}
+
+// API Provider enumeration
+enum ApiProvider {
+  ANTHROPIC = 0;
+  OPENROUTER = 1;
+  BEDROCK = 2;
+  VERTEX = 3;
+  OPENAI = 4;
+  OLLAMA = 5;
+  LMSTUDIO = 6;
+  GEMINI = 7;
+  OPENAI_NATIVE = 8;
+  REQUESTY = 9;
+  TOGETHER = 10;
+  DEEPSEEK = 11;
+  QWEN = 12;
+  DOUBAO = 13;
+  MISTRAL = 14;
+  VSCODE_LM = 15;
+  CLINE = 16;
+  LITELLM = 17;
+  NEBIUS = 18;
+  FIREWORKS = 19;
+  ASKSAGE = 20;
+  XAI = 21;
+  SAMBANOVA = 22;
+  CEREBRAS = 23;
+  SAPAICORE = 24;
+  CLAUDE_CODE = 25;
+}
+
+// Model info for OpenAI-compatible models
+message OpenAiCompatibleModelInfo {
+  optional int32 max_tokens = 1;
+  optional int32 context_window = 2;
+  optional bool supports_images = 3;
+  bool supports_prompt_cache = 4;
+  optional double input_price = 5;
+  optional double output_price = 6;
+  optional ThinkingConfig thinking_config = 7;
+  optional bool supports_global_endpoint = 8;
+  optional double cache_writes_price = 9;
+  optional double cache_reads_price = 10;
+  optional string description = 11;
+  repeated ModelTier tiers = 12;
+  optional double temperature = 13;
+  optional bool is_r1_format_required = 14;
+}
+
+// Model info for LiteLLM models
+message LiteLLMModelInfo {
+  optional int32 max_tokens = 1;
+  optional int32 context_window = 2;
+  optional bool supports_images = 3;
+  bool supports_prompt_cache = 4;
+  optional double input_price = 5;
+  optional double output_price = 6;
+  optional ThinkingConfig thinking_config = 7;
+  optional bool supports_global_endpoint = 8;
+  optional double cache_writes_price = 9;
+  optional double cache_reads_price = 10;
+  optional string description = 11;
+  repeated ModelTier tiers = 12;
+  optional double temperature = 13;
+}
+
+// Main ApiConfiguration message
+message ModelsApiConfiguration {
+  // From ApiHandlerOptions (excluding onRetryAttempt function)
+  optional string api_model_id = 1;
+  optional string api_key = 2;
+  optional string cline_api_key = 3;
+  optional string task_id = 4;
+  optional string lite_llm_base_url = 5;
+  optional string lite_llm_model_id = 6;
+  optional string lite_llm_api_key = 7;
+  optional bool lite_llm_use_prompt_cache = 8;
+  map<string, string> open_ai_headers = 9;
+  optional LiteLLMModelInfo lite_llm_model_info = 10;
+  optional string anthropic_base_url = 11;
+  optional string open_router_api_key = 12;
+  optional string open_router_model_id = 13;
+  optional OpenRouterModelInfo open_router_model_info = 14;
+  optional string open_router_provider_sorting = 15;
+  optional string aws_access_key = 16;
+  optional string aws_secret_key = 17;
+  optional string aws_session_token = 18;
+  optional string aws_region = 19;
+  optional bool aws_use_cross_region_inference = 20;
+  optional bool aws_bedrock_use_prompt_cache = 21;
+  optional bool aws_use_profile = 22;
+  optional string aws_profile = 23;
+  optional string aws_bedrock_endpoint = 24;
+  optional bool aws_bedrock_custom_selected = 25;
+  optional string aws_bedrock_custom_model_base_id = 26;
+  optional string vertex_project_id = 27;
+  optional string vertex_region = 28;
+  optional string open_ai_base_url = 29;
+  optional string open_ai_api_key = 30;
+  optional string open_ai_model_id = 31;
+  optional OpenAiCompatibleModelInfo open_ai_model_info = 32;
+  optional string ollama_model_id = 33;
+  optional string ollama_base_url = 34;
+  optional string ollama_api_options_ctx_num = 35;
+  optional string lm_studio_model_id = 36;
+  optional string lm_studio_base_url = 37;
+  optional string gemini_api_key = 38;
+  optional string gemini_base_url = 39;
+  optional string open_ai_native_api_key = 40;
+  optional string deep_seek_api_key = 41;
+  optional string requesty_api_key = 42;
+  optional string requesty_model_id = 43;
+  optional OpenRouterModelInfo requesty_model_info = 44;
+  optional string together_api_key = 45;
+  optional string together_model_id = 46;
+  optional string fireworks_api_key = 47;
+  optional string fireworks_model_id = 48;
+  optional int32 fireworks_model_max_completion_tokens = 49;
+  optional int32 fireworks_model_max_tokens = 50;
+  optional string qwen_api_key = 51;
+  optional string doubao_api_key = 52;
+  optional string mistral_api_key = 53;
+  optional string azure_api_version = 54;
+  optional LanguageModelChatSelector vs_code_lm_model_selector = 55;
+  optional string qwen_api_line = 56;
+  optional string nebius_api_key = 57;
+  optional string asksage_api_url = 58;
+  optional string asksage_api_key = 59;
+  optional string xai_api_key = 60;
+  optional int32 thinking_budget_tokens = 61;
+  optional string reasoning_effort = 62;
+  optional string sambanova_api_key = 63;
+  optional string cerebras_api_key = 64;
+  optional int32 request_timeout_ms = 65;
+  optional ApiProvider api_provider = 66;
+  repeated string favorited_model_ids = 67;
+  optional string sap_ai_core_client_id = 68;
+  optional string sap_ai_core_client_secret = 69;
+  optional string sap_ai_resource_group = 70;
+  optional string sap_ai_core_token_url = 71;
+  optional string sap_ai_core_base_url = 72;
+  optional string claude_code_path = 73;
+}

proto/state.proto

@@ -7,18 +7,39 @@ import "common.proto";
 
 service StateService {
   rpc getLatestState(EmptyRequest) returns (State);
+  rpc updateTerminalConnectionTimeout(Int64Request) returns (Int64);
+  rpc updateTerminalReuseEnabled(BooleanRequest) returns (Empty);
+  rpc updateDefaultTerminalProfile(StringRequest) returns (TerminalProfileUpdateResponse);
+  rpc getAvailableTerminalProfiles(EmptyRequest) returns (TerminalProfiles);
   rpc subscribeToState(EmptyRequest) returns (stream State);
   rpc toggleFavoriteModel(StringRequest) returns (Empty);
-  rpc resetState(EmptyRequest) returns (Empty);
+  rpc resetState(ResetStateRequest) returns (Empty);
   rpc togglePlanActMode(TogglePlanActModeRequest) returns (Empty);
-  rpc updateTerminalConnectionTimeout(Int64Request) returns (Int64);
   rpc updateAutoApprovalSettings(AutoApprovalSettingsRequest) returns (Empty);
+  rpc updateSettings(UpdateSettingsRequest) returns (Empty);
 }
 
 message State {
   string state_json = 1;
 }
 
+message TerminalProfiles {
+  repeated TerminalProfile profiles = 1;
+}
+
+message TerminalProfile {
+  string id = 1;
+  string name = 2;
+  optional string path = 3;
+  optional string description = 4;
+}
+
+message TerminalProfileUpdateResponse {
+  int32 closed_count = 1;
+  int32 busy_terminals_count = 2;
+  bool has_busy_terminals = 3;
+}
+
 message TogglePlanActModeRequest {
   Metadata metadata = 1;
   ChatSettings chat_settings = 2;
@@ -42,10 +63,13 @@ message ChatContent {
   repeated string files = 3;
 }
 
-// Message for auto approval settings
+message ResetStateRequest {
+  Metadata metadata = 1;
+  optional bool global = 2;
+}
+
 message AutoApprovalSettingsRequest {
   Metadata metadata = 1;
-
   message Actions {
     bool read_files = 1;
     bool read_files_externally = 2;
@@ -56,7 +80,6 @@ message AutoApprovalSettingsRequest {
     bool use_browser = 7;
     bool use_mcp = 8;
   }
-
   int32 version = 2;
   bool enabled = 3;
   Actions actions = 4;
@@ -64,3 +87,137 @@ message AutoApprovalSettingsRequest {
   bool enable_notifications = 6;
   repeated string favorites = 7;
 }
+
+// Message for updating settings
+message UpdateSettingsRequest {
+  Metadata metadata = 1;
+  optional ApiConfiguration api_configuration = 2;
+  optional string telemetry_setting = 3;
+  optional bool plan_act_separate_models_setting = 4;
+  optional bool enable_checkpoints_setting = 5;
+  optional bool mcp_marketplace_enabled = 6;
+  optional ChatSettings chat_settings = 7;
+  optional int64 shell_integration_timeout = 8;
+  optional bool terminal_reuse_enabled = 9;
+  optional bool mcp_responses_collapsed = 10;
+  optional bool mcp_rich_display_enabled = 11;
+  optional int64 terminal_output_line_limit = 12;
+}
+
+// Complete API Configuration message
+message ApiConfiguration {
+  // Core API fields
+  optional string api_provider = 1;
+  optional string api_model_id = 2;
+  optional string api_key = 3; // anthropic
+  optional string api_base_url = 4;
+  
+  // Provider-specific API keys
+  optional string cline_api_key = 5;
+  optional string openrouter_api_key = 6;
+  optional string anthropic_base_url = 7;
+  optional string openai_api_key = 8;
+  optional string openai_native_api_key = 9;
+  optional string gemini_api_key = 10;
+  optional string deepseek_api_key = 11;
+  optional string requesty_api_key = 12;
+  optional string together_api_key = 13;
+  optional string fireworks_api_key = 14;
+  optional string qwen_api_key = 15;
+  optional string doubao_api_key = 16;
+  optional string mistral_api_key = 17;
+  optional string nebius_api_key = 18;
+  optional string asksage_api_key = 19;
+  optional string xai_api_key = 20;
+  optional string sambanova_api_key = 21;
+  optional string cerebras_api_key = 22;
+  
+  // Model IDs
+  optional string openrouter_model_id = 23;
+  optional string openai_model_id = 24;
+  optional string anthropic_model_id = 25;
+  optional string bedrock_model_id = 26;
+  optional string vertex_model_id = 27;
+  optional string gemini_model_id = 28;
+  optional string ollama_model_id = 29;
+  optional string lm_studio_model_id = 30;
+  optional string litellm_model_id = 31;
+  optional string requesty_model_id = 32;
+  optional string together_model_id = 33;
+  optional string fireworks_model_id = 34;
+  
+  // AWS Bedrock fields
+  optional bool aws_bedrock_custom_selected = 35;
+  optional string aws_bedrock_custom_model_base_id = 36;
+  optional string aws_access_key = 37;
+  optional string aws_secret_key = 38;
+  optional string aws_session_token = 39;
+  optional string aws_region = 40;
+  optional bool aws_use_cross_region_inference = 41;
+  optional bool aws_bedrock_use_prompt_cache = 42;
+  optional bool aws_use_profile = 43;
+  optional string aws_profile = 44;
+  optional string aws_bedrock_endpoint = 45;
+  
+  // Vertex AI fields
+  optional string vertex_project_id = 46;
+  optional string vertex_region = 47;
+  
+  // Base URLs and endpoints
+  optional string openai_base_url = 48;
+  optional string ollama_base_url = 49;
+  optional string lm_studio_base_url = 50;
+  optional string gemini_base_url = 51;
+  optional string litellm_base_url = 52;
+  optional string asksage_api_url = 53;
+  
+  // LiteLLM specific fields
+  optional string litellm_api_key = 54;
+  optional bool litellm_use_prompt_cache = 55;
+  
+  // Model configuration
+  optional int64 thinking_budget_tokens = 56;
+  optional string reasoning_effort = 57;
+  optional int64 request_timeout_ms = 58;
+  
+  // Fireworks specific
+  optional int64 fireworks_model_max_completion_tokens = 59;
+  optional int64 fireworks_model_max_tokens = 60;
+  
+  // Azure specific
+  optional string azure_api_version = 61;
+  
+  // Ollama specific
+  optional string ollama_api_options_ctx_num = 62;
+  
+  // Qwen specific
+  optional string qwen_api_line = 63;
+  
+  // OpenRouter specific
+  optional string openrouter_provider_sorting = 64;
+  
+  // VSCode LM (stored as JSON string due to complex type)
+  optional string vscode_lm_model_selector = 65;
+  
+  // Model info objects (stored as JSON strings)
+  optional string openrouter_model_info = 66;
+  optional string openai_model_info = 67;
+  optional string requesty_model_info = 68;
+  optional string litellm_model_info = 69;
+  
+  // OpenAI headers (stored as JSON string)
+  optional string openai_headers = 70;
+  
+  // Favorited model IDs
+  repeated string favorited_model_ids = 71;
+
+  // SAP AI Core specific
+  optional string sap_ai_core_client_id = 72;
+  optional string sap_ai_core_client_secret = 73;
+  optional string sap_ai_core_base_url = 74;
+  optional string sap_ai_core_token_url = 75;
+  optional string sap_ai_resource_group = 76;
+
+  // Claude Code specific
+  optional string claude_code_path = 77;
+}

proto/task.proto

@@ -33,6 +33,8 @@ service TaskService {
   rpc taskFeedback(StringRequest) returns (Empty);
   // Shows task completion changes diff in a view
   rpc taskCompletionViewChanges(Int64Request) returns (Empty);
+  // Executes a quick win task with command and title
+  rpc executeQuickWin(ExecuteQuickWinRequest) returns (Empty);
 }
 
 // Request message for creating a new task
@@ -107,3 +109,10 @@ message AskResponseRequest {
   repeated string images = 4;
   repeated string files = 5;
 }
+
+// Request for executing a quick win task
+message ExecuteQuickWinRequest {
+  Metadata metadata = 1;
+  string command = 2;
+  string title = 3;
+}

proto/ui.proto

@@ -15,13 +15,214 @@ enum WebviewProviderType {
 // Define a new message type for webview provider info
 message WebviewProviderTypeRequest {
   Metadata metadata = 1;
-  WebviewProviderType providerType = 2;
+  WebviewProviderType provider_type = 2;
+}
+
+// Enum for ClineMessage type
+enum ClineMessageType {
+  ASK = 0;
+  SAY = 1;
+}
+
+// Enum for ClineAsk types
+enum ClineAsk {
+  FOLLOWUP = 0;
+  PLAN_MODE_RESPOND = 1;
+  COMMAND = 2;
+  COMMAND_OUTPUT = 3;
+  COMPLETION_RESULT = 4;
+  TOOL = 5;
+  API_REQ_FAILED = 6;
+  RESUME_TASK = 7;
+  RESUME_COMPLETED_TASK = 8;
+  MISTAKE_LIMIT_REACHED = 9;
+  AUTO_APPROVAL_MAX_REQ_REACHED = 10;
+  BROWSER_ACTION_LAUNCH = 11;
+  USE_MCP_SERVER = 12;
+  NEW_TASK = 13;
+  CONDENSE = 14;
+  REPORT_BUG = 15;
+}
+
+// Enum for ClineSay types
+enum ClineSay {
+  TASK = 0;
+  ERROR = 1;
+  API_REQ_STARTED = 2;
+  API_REQ_FINISHED = 3;
+  TEXT = 4;
+  REASONING = 5;
+  COMPLETION_RESULT_SAY = 6;
+  USER_FEEDBACK = 7;
+  USER_FEEDBACK_DIFF = 8;
+  API_REQ_RETRIED = 9;
+  COMMAND_SAY = 10;
+  COMMAND_OUTPUT_SAY = 11;
+  TOOL_SAY = 12;
+  SHELL_INTEGRATION_WARNING = 13;
+  BROWSER_ACTION_LAUNCH_SAY = 14;
+  BROWSER_ACTION = 15;
+  BROWSER_ACTION_RESULT = 16;
+  MCP_SERVER_REQUEST_STARTED = 17;
+  MCP_SERVER_RESPONSE = 18;
+  MCP_NOTIFICATION = 19;
+  USE_MCP_SERVER_SAY = 20;
+  DIFF_ERROR = 21;
+  DELETED_API_REQS = 22;
+  CLINEIGNORE_ERROR = 23;
+  CHECKPOINT_CREATED = 24;
+  LOAD_MCP_DOCUMENTATION = 25;
+  INFO = 26;
+}
+
+// Enum for ClineSayTool tool types
+enum ClineSayToolType {
+  EDITED_EXISTING_FILE = 0;
+  NEW_FILE_CREATED = 1;
+  READ_FILE = 2;
+  LIST_FILES_TOP_LEVEL = 3;
+  LIST_FILES_RECURSIVE = 4;
+  LIST_CODE_DEFINITION_NAMES = 5;
+  SEARCH_FILES = 6;
+  WEB_FETCH = 7;
+}
+
+// Enum for browser actions
+enum BrowserAction {
+  LAUNCH = 0;
+  CLICK = 1;
+  TYPE = 2;
+  SCROLL_DOWN = 3;
+  SCROLL_UP = 4;
+  CLOSE = 5;
+}
+
+// Enum for MCP server request types
+enum McpServerRequestType {
+  USE_MCP_TOOL = 0;
+  ACCESS_MCP_RESOURCE = 1;
+}
+
+// Enum for API request cancel reasons
+enum ClineApiReqCancelReason {
+  STREAMING_FAILED = 0;
+  USER_CANCELLED = 1;
+  RETRIES_EXHAUSTED = 2;
+}
+
+// Message for conversation history deleted range
+message ConversationHistoryDeletedRange {
+  int32 start_index = 1;
+  int32 end_index = 2;
+}
+
+// Message for ClineSayTool
+message ClineSayTool {
+  ClineSayToolType tool = 1;
+  string path = 2;
+  string diff = 3;
+  string content = 4;
+  string regex = 5;
+  string file_pattern = 6;
+  bool operation_is_located_in_workspace = 7;
+}
+
+// Message for ClineSayBrowserAction
+message ClineSayBrowserAction {
+  BrowserAction action = 1;
+  string coordinate = 2;
+  string text = 3;
+}
+
+// Message for BrowserActionResult
+message BrowserActionResult {
+  string screenshot = 1;
+  string logs = 2;
+  string current_url = 3;
+  string current_mouse_position = 4;
+}
+
+// Message for ClineAskUseMcpServer
+message ClineAskUseMcpServer {
+  string server_name = 1;
+  McpServerRequestType type = 2;
+  string tool_name = 3;
+  string arguments = 4;
+  string uri = 5;
+}
+
+// Message for ClinePlanModeResponse
+message ClinePlanModeResponse {
+  string response = 1;
+  repeated string options = 2;
+  string selected = 3;
+}
+
+// Message for ClineAskQuestion
+message ClineAskQuestion {
+  string question = 1;
+  repeated string options = 2;
+  string selected = 3;
+}
+
+// Message for ClineAskNewTask
+message ClineAskNewTask {
+  string context = 1;
+}
+
+// Message for API request retry status
+message ApiReqRetryStatus {
+  int32 attempt = 1;
+  int32 max_attempts = 2;
+  int32 delay_sec = 3;
+  string error_snippet = 4;
+}
+
+// Message for ClineApiReqInfo
+message ClineApiReqInfo {
+  string request = 1;
+  int32 tokens_in = 2;
+  int32 tokens_out = 3;
+  int32 cache_writes = 4;
+  int32 cache_reads = 5;
+  double cost = 6;
+  ClineApiReqCancelReason cancel_reason = 7;
+  string streaming_failed_message = 8;
+  ApiReqRetryStatus retry_status = 9;
+}
+
+// Main ClineMessage type
+message ClineMessage {
+  int64 ts = 1;
+  ClineMessageType type = 2;
+  ClineAsk ask = 3;
+  ClineSay say = 4;
+  string text = 5;
+  string reasoning = 6;
+  repeated string images = 7;
+  repeated string files = 8;
+  bool partial = 9;
+  string last_checkpoint_hash = 10;
+  bool is_checkpoint_checked_out = 11;
+  bool is_operation_outside_workspace = 12;
+  int32 conversation_history_index = 13;
+  ConversationHistoryDeletedRange conversation_history_deleted_range = 14;
+  
+  // Additional fields for specific ask/say types
+  ClineSayTool say_tool = 15;
+  ClineSayBrowserAction say_browser_action = 16;
+  BrowserActionResult browser_action_result = 17;
+  ClineAskUseMcpServer ask_use_mcp_server = 18;
+  ClinePlanModeResponse plan_mode_response = 19;
+  ClineAskQuestion ask_question = 20;
+  ClineAskNewTask ask_new_task = 21;
+  ClineApiReqInfo api_req_info = 22;
 }
 
 // UiService provides methods for managing UI interactions
 service UiService {
   // Scrolls to a specific settings section in the settings view
-  rpc scrollToSettings(StringRequest) returns (Empty);
+  rpc scrollToSettings(StringRequest) returns (KeyValuePair);
   
   // Marks the current announcement as shown and returns whether an announcement should still be shown
   rpc onDidShowAnnouncement(EmptyRequest) returns (Boolean);
@@ -34,4 +235,31 @@ service UiService {
   
   // Subscribe to history button click events
   rpc subscribeToHistoryButtonClicked(WebviewProviderTypeRequest) returns (stream Empty);
+  
+  // Subscribe to chat button clicked events (when the chat button is clicked in VSCode)
+  rpc subscribeToChatButtonClicked(EmptyRequest) returns (stream Empty);
+  
+  // Subscribe to account button click events
+  rpc subscribeToAccountButtonClicked(EmptyRequest) returns (stream Empty);
+  
+  // Subscribe to settings button clicked events
+  rpc subscribeToSettingsButtonClicked(WebviewProviderTypeRequest) returns (stream Empty);
+  
+  // Subscribe to partial message updates (streaming Cline messages as they're built)
+  rpc subscribeToPartialMessage(EmptyRequest) returns (stream ClineMessage);
+  
+  // Subscribe to theme change events
+  rpc subscribeToTheme(EmptyRequest) returns (stream String);
+  
+  // Initialize webview when it launches
+  rpc initializeWebview(EmptyRequest) returns (Empty);
+  
+  // Subscribe to relinquish control events
+  rpc subscribeToRelinquishControl(EmptyRequest) returns (stream Empty);
+  
+  // Subscribe to focus chat input events with client ID
+  rpc subscribeToFocusChatInput(StringRequest) returns (stream Empty);
+  
+  // Subscribe to webview visibility change events
+  rpc subscribeToDidBecomeVisible(EmptyRequest) returns (stream Empty);
 }

scripts/build-tests.js

@@ -34,7 +34,6 @@ const srcConfig = {
 	format: "cjs",
 	platform: "node",
 	define: {
-		"process.env.IS_DEV": "true",
 		"process.env.IS_TEST": "true",
 	},
 	external: ["vscode"],

scripts/generate-host-bridge-client.mjs

@@ -0,0 +1,192 @@
+#!/usr/bin/env node
+
+import * as fs from "fs/promises"
+import * as path from "path"
+import * as grpc from "@grpc/grpc-js"
+import * as protoLoader from "@grpc/proto-loader"
+import chalk from "chalk"
+
+const IMPL_FILE = path.resolve("src/generated/standalone/host-bridge-clients.ts")
+const INTERFACE_FILE = path.resolve("src/generated/hosts/host-bridge-client-types.ts")
+const DESCRIPTOR_SET = path.resolve("dist-standalone/proto/descriptor_set.pb")
+
+const typeNameToFQN = new Map()
+
+function addTypeNameToFqn(name, fqn) {
+	if (typeNameToFQN.has(name)) {
+		throw new Error(`Proto type ${name} redefined (${fqn}).`)
+	}
+	typeNameToFQN.set(name, fqn)
+}
+function getFqn(name) {
+	if (!typeNameToFQN.has(name)) {
+		throw Error(`No FQN for ${name}`)
+	}
+	return typeNameToFQN.get(name)
+}
+/**
+ * Main function to generate the host bridge client
+ */
+async function main() {
+	// Load service definitions from descriptor set
+	const descriptorBuffer = await fs.readFile(DESCRIPTOR_SET)
+	const packageDefinition = protoLoader.loadFileDescriptorSetFromBuffer(descriptorBuffer)
+	const proto = grpc.loadPackageDefinition(packageDefinition)
+
+	// Extract host services and proto messages from the proto definition
+	const hostServices = {}
+	for (const [name, def] of Object.entries(proto.host)) {
+		if (def && "service" in def) {
+			hostServices[name] = def
+		} else {
+			addTypeNameToFqn(name, `proto.host.${name}`)
+		}
+	}
+	for (const [name, def] of Object.entries(proto.cline)) {
+		if (def && !("service" in def)) {
+			addTypeNameToFqn(name, `proto.cline.${name}`)
+		}
+	}
+
+	// Generate interfaces file
+	await generateInterfacesFile(hostServices)
+
+	// // Generate implementation file
+	await generateImplementationFile(hostServices)
+
+	console.log(`Generated host bridge client files at:`)
+	console.log(`- ${INTERFACE_FILE}`)
+	console.log(`- ${IMPL_FILE}`)
+}
+
+/**
+ * Generate the client interfaces file.
+ */
+async function generateInterfacesFile(hostServices) {
+	const clientInterfaces = []
+	for (const [name, def] of Object.entries(hostServices)) {
+		const clientInterface = generateClientInterface(name, def)
+		clientInterfaces.push(clientInterface)
+	}
+	const content = `// GENERATED CODE -- DO NOT EDIT!
+// Generated by scripts/generate-host-bridge-client.mjs
+import * as proto from "@shared/proto/index"
+import { StreamingCallbacks } from "@hosts/host-bridge-client"
+
+${clientInterfaces.join("\n\n")}
+`
+	// Write output file
+	await fs.mkdir(path.dirname(INTERFACE_FILE), { recursive: true })
+	await fs.writeFile(INTERFACE_FILE, content)
+}
+
+/**
+ * Generate a client interface for a service.
+ */
+function generateClientInterface(serviceName, serviceDefinition) {
+	// Get the methods from the service definition
+	const methods = Object.entries(serviceDefinition.service)
+		.map(([methodName, methodDef]) => {
+			const requestType = getFqn(methodDef.requestType.type.name)
+			const responseType = getFqn(methodDef.responseType.type.name)
+
+			if (!methodDef.responseStream) {
+				// Generate unary method signature.
+				return `	${methodName}(request: ${requestType}): Promise<${responseType}>;`
+			}
+			// Generate streaming method signature.
+			return `	${methodName}(request: ${requestType}, callbacks: StreamingCallbacks<${responseType}>): () => void;`
+		})
+		.join("\n\n")
+
+	// Generate the interface
+	return `/**
+ * Interface for ${serviceName} client.
+ */
+export interface ${serviceName}ClientInterface {
+
+${methods}
+}`
+}
+
+/**
+ * Generate the client implementations file.
+ */
+async function generateImplementationFile(hostServices) {
+	// Generate imports
+	const imports = []
+	// Add imports for the interfaces
+	for (const [name, _def] of Object.entries(hostServices)) {
+		imports.push(`import { ${name}ClientInterface } from "@generated/hosts/host-bridge-client-types"`)
+	}
+	const clientImplementations = []
+	for (const [name, def] of Object.entries(hostServices)) {
+		clientImplementations.push(generateClientImplementation(name, def))
+	}
+
+	const content = `// GENERATED CODE -- DO NOT EDIT!
+// Generated by scripts/generate-host-bridge-client.mjs
+import { asyncIteratorToCallbacks } from "@/standalone/utils"
+import * as niceGrpc from "@generated/nice-grpc/index"
+import { StreamingCallbacks } from "@hosts/host-bridge-client"
+import * as proto from "@shared/proto/index"
+import { Channel, createClient } from "nice-grpc"
+
+${imports.join("\n")}
+
+${clientImplementations.join("\n\n")}
+`
+
+	// Write output file
+	await fs.mkdir(path.dirname(IMPL_FILE), { recursive: true })
+	await fs.writeFile(IMPL_FILE, content)
+}
+
+/**
+ * Generate a client implementation class for a service
+ */
+function generateClientImplementation(serviceName, serviceDefinition) {
+	// Get the methods from the service definition
+	const methods = Object.entries(serviceDefinition.service)
+		.map(([methodName, methodDef]) => {
+			// Get fully qualified type names
+			const requestType = getFqn(methodDef.requestType.type.name)
+			const responseType = getFqn(methodDef.responseType.type.name)
+			const isStreamingResponse = methodDef.responseStream
+
+			if (!isStreamingResponse) {
+				return `  ${methodName}(request: ${requestType}): Promise<${responseType}> {
+    return this.client.${methodName}(request)
+  }`
+			} else {
+				// Generate streaming method
+				return `  ${methodName}(request: ${requestType}, callbacks: StreamingCallbacks<${responseType}>): () => void {
+	const abortController = new AbortController()
+	const stream: AsyncIterable<${responseType}> = this.client.${methodName}(request, {signal: abortController.signal})
+    asyncIteratorToCallbacks(stream, callbacks)
+	return () => {abortController.abort()}
+  }`
+			}
+		})
+		.join("\n\n")
+
+	// Generate the class
+	return `/**
+ * Type-safe client implementation for ${serviceName}.
+ */
+export class ${serviceName}ClientImpl implements ${serviceName}ClientInterface {
+  private client: niceGrpc.host.${serviceName}Client 
+
+  constructor(channel: Channel) {
+    this.client = createClient(niceGrpc.host.${serviceName}Definition, channel)
+  }
+
+${methods}
+}`
+}
+
+// Run the main function
+main().catch((error) => {
+	console.error(chalk.red("Error:"), error)
+	process.exit(1)
+})

scripts/generate-server-setup.mjs

@@ -1,11 +1,11 @@
-import * as fs from "fs"
 import * as grpc from "@grpc/grpc-js"
 import * as protoLoader from "@grpc/proto-loader"
+import * as fs from "fs"
 import * as health from "grpc-health-check"
+import path, { basename, dirname } from "path"
 import { fileURLToPath } from "url"
-import path from "path"
 
-const OUT_FILE = path.resolve("src/standalone/server-setup.ts")
+const OUT_FILE = path.resolve("src/generated/standalone/server-setup.ts")
 const DESCRIPTOR_SET = path.resolve("dist-standalone/proto/descriptor_set.pb")
 
 // Load service definitions.
@@ -29,16 +29,18 @@ function generateHandlersAndExports() {
 		const dir = domain.charAt(0).toLowerCase() + domain.slice(1)
 		imports.push(`// ${domain} Service`)
 		handlerSetup.push(`    // ${domain} Service`)
-		handlerSetup.push(`    server.addService(proto.cline.${name}.service, {`)
+		handlerSetup.push(`    server.addService(cline.${name}Service, {`)
 		for (const [rpcName, rpc] of Object.entries(def.service)) {
-			imports.push(`import { ${rpcName} } from "../core/controller/${dir}/${rpcName}"`)
+			imports.push(`import { ${rpcName} } from "@core/controller/${dir}/${rpcName}"`)
+			const requestType = "cline." + rpc.requestType.type.name
 			if (rpc.requestStream) {
 				throw new Error("Request streaming is not supported")
 			}
 			if (rpc.responseStream) {
-				handlerSetup.push(`        ${rpcName}: wrapStreamingResponse(${rpcName}, controller),`)
+				handlerSetup.push(`        ${rpcName}: wrapStreamingResponse<${requestType},void>(${rpcName}, controller),`)
 			} else {
-				handlerSetup.push(`         ${rpcName}: wrapper(${rpcName}, controller),`)
+				const responseType = "cline." + rpc.responseType.type.name
+				handlerSetup.push(`         ${rpcName}: wrapper<${requestType},${responseType}>(${rpcName}, controller),`)
 			}
 		}
 		handlerSetup.push(`    });`)
@@ -58,14 +60,13 @@ const scriptName = path.basename(fileURLToPath(import.meta.url))
 let output = `// GENERATED CODE -- DO NOT EDIT!
 // Generated by ${scriptName}
 import * as grpc from "@grpc/grpc-js"
-import { Controller } from "../core/controller"
-import { GrpcHandlerWrapper, GrpcStreamingResponseHandlerWrapper } from "./grpc-types"
+import { cline } from "@generated/grpc-js"
+import { Controller } from "@core/controller"
+import { GrpcHandlerWrapper, GrpcStreamingResponseHandlerWrapper } from "@/standalone/grpc-types"
 
 ${imports}
-
-export function addServices(
+export function addProtobusServices(
 	server: grpc.Server,
-	proto: any,
 	controller: Controller,
 	wrapper: GrpcHandlerWrapper,
 	wrapStreamingResponse: GrpcStreamingResponseHandlerWrapper,
@@ -74,6 +75,7 @@ ${handlerSetup}
 }
 `
 // Write output file
+fs.mkdirSync(dirname(OUT_FILE), { recursive: true })
 fs.writeFileSync(OUT_FILE, output)
 
 console.log(`Generated service handlers in ${OUT_FILE}.`)

scripts/get-vscode-usages.sh

@@ -12,6 +12,7 @@ git grep -h 'vscode\.' $DIR |
 grep -Ev '//.*vscode' | # remove commented out code
 sed 's|.*vscode\.|vscode.|'| # remove everything before vscode.
 sed 's/[^a-zA-Z0-9_.].*$//' | # remove everything after last identifier
+grep -E '\.[a-z][^.]+$' | # remove types (last part of identifier should be lowercase)
 sort | uniq > $SDK_DEST
 }
 echo Wrote uses of the vscode SDK to $(realpath $SDK_DEST)

scripts/package-standalone.mjs

@@ -36,12 +36,14 @@ if (nativeModules.length > 0) {
 // Zip the build directory (excluding any pre-existing output zip).
 const zipPath = path.join(BUILD_DIR, "standalone.zip")
 const output = fs.createWriteStream(zipPath)
-const archive = archiver("zip", { zlib: { level: 9 } })
+const archive = archiver("zip", { zlib: { level: 3 } })
 
 output.on("close", () => {
 	console.log(`Created ${zipPath} (${(archive.pointer() / 1024 / 1024).toFixed(1)} MB)`)
 })
-
+archive.on("warning", (err) => {
+	console.warn(`Warning: ${err}`)
+})
 archive.on("error", (err) => {
 	throw err
 })
@@ -51,4 +53,20 @@ archive.glob("**/*", {
 	cwd: BUILD_DIR,
 	ignore: ["standalone.zip"],
 })
+
+// Add the whole cline directory under "extension"
+archive.directory(process.cwd(), "extension", (entry) => {
+	// Skip certain directories
+	if (
+		entry.name.startsWith(BUILD_DIR + "/") ||
+		entry.name.startsWith("node_modules/") || // node_modules nearly 1GB.
+		entry.name.startsWith("webview-ui/node_modules/") || // node_modules nearly 1GB.
+		entry.name.match(/(^|\/)\./) // exclude dot directories
+	) {
+		return false
+	}
+	return entry
+})
+
+console.log("Zipping package...")
 await archive.finalize()

src/api/index.ts

@@ -25,6 +25,8 @@ import { AskSageHandler } from "./providers/asksage"
 import { XAIHandler } from "./providers/xai"
 import { SambanovaHandler } from "./providers/sambanova"
 import { CerebrasHandler } from "./providers/cerebras"
+import { SapAiCoreHandler } from "./providers/sapaicore"
+import { ClaudeCodeHandler } from "./providers/claude-code"
 
 export interface ApiHandler {
 	createMessage(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]): ApiStream
@@ -87,6 +89,10 @@ export function buildApiHandler(configuration: ApiConfiguration): ApiHandler {
 			return new SambanovaHandler(options)
 		case "cerebras":
 			return new CerebrasHandler(options)
+		case "sapaicore":
+			return new SapAiCoreHandler(options)
+		case "claude-code":
+			return new ClaudeCodeHandler(options)
 		default:
 			return new AnthropicHandler(options)
 	}

src/api/providers/__tests__/bedrock.test.ts

@@ -1,8 +1,49 @@
 import "should"
 import { AwsBedrockHandler } from "../bedrock"
 import { ApiHandlerOptions } from "@shared/api"
+import { ConverseStreamCommand } from "@aws-sdk/client-bedrock-runtime"
+import { Readable } from "stream"
 
 describe("AwsBedrockHandler", () => {
+	// Helper function to create a mock stream
+	function createMockStream(chunks: any[]): Readable {
+		const stream = new Readable({
+			objectMode: true,
+			read() {
+				if (chunks.length > 0) {
+					this.push(chunks.shift())
+				} else {
+					this.push(null)
+				}
+			},
+		})
+		return stream
+	}
+
+	// Helper function to collect generator results
+	async function collectGeneratorResults(generator: AsyncGenerator<any>): Promise<any[]> {
+		const results: any[] = []
+		for await (const item of generator) {
+			results.push(item)
+		}
+		return results
+	}
+
+	// Mock AWS Bedrock client
+	class MockBedrockClient {
+		private streamChunks: any[]
+
+		constructor(streamChunks: any[]) {
+			this.streamChunks = streamChunks
+		}
+
+		async send(_command: any): Promise<any> {
+			return {
+				stream: createMockStream(this.streamChunks),
+			}
+		}
+	}
+
 	describe("withTempEnv", () => {
 		// Store original env vars for cleanup
 		const originalEnv: Record<string, string | undefined> = {}
@@ -142,4 +183,511 @@ describe("AwsBedrockHandler", () => {
 			process.env["AWS_PROFILE"]!.should.equal(preAWSProfile)
 		})
 	})
+
+	const mockOptions: ApiHandlerOptions = {
+		apiModelId: "anthropic.claude-3-7-sonnet-20250219-v1:0",
+		awsRegion: "us-east-1",
+		awsAccessKey: "test-key",
+		awsSecretKey: "test-secret",
+		awsSessionToken: "",
+		awsUseProfile: false,
+		awsProfile: "",
+		awsBedrockUsePromptCache: false,
+		awsUseCrossRegionInference: false,
+		awsBedrockEndpoint: "",
+		awsBedrockCustomSelected: false,
+		awsBedrockCustomModelBaseId: undefined,
+		thinkingBudgetTokens: 1600,
+	}
+
+	const mockModelInfo = {
+		maxTokens: 8192,
+		contextWindow: 200000,
+		supportsPromptCache: true,
+		supportsImages: true,
+		inputPrice: 3.0,
+		outputPrice: 15.0,
+		cacheWritesPrice: 3.75,
+		cacheReadsPrice: 0.3,
+	}
+
+	describe("executeConverseStream", () => {
+		let handler: AwsBedrockHandler
+
+		beforeEach(() => {
+			handler = new AwsBedrockHandler(mockOptions)
+		})
+
+		describe("reasoning content handling (deprecated)", () => {
+			// These tests are for the old reasoningContent API that may be deprecated
+			// Keep them for backward compatibility but they may fail with new API
+		})
+
+		describe("thinking response handling (new API structure)", () => {
+			it("should handle thinking response in additionalModelResponseFields", async () => {
+				const mockChunks = [
+					{ messageStart: { role: "assistant" } },
+					{
+						metadata: {
+							additionalModelResponseFields: {
+								thinkingResponse: {
+									reasoning: [
+										{
+											type: "text",
+											text: "まず与えられた数値50.653の立方根を求める必要があります。",
+											signature: "sig1",
+										},
+										{
+											type: "text",
+											text: "立方根を近似するために数値を3乗したときの誤差を調整していきます。",
+											signature: "sig2",
+										},
+									],
+								},
+							},
+						},
+					},
+					{ contentBlockDelta: { delta: { text: "50.653の立方根は約3.707です。" }, contentBlockIndex: 0 } },
+					{ contentBlockStop: { contentBlockIndex: 0 } },
+					{ messageStop: { stopReason: "end_turn" } },
+					{ metadata: { usage: { inputTokens: 100, outputTokens: 50 } } },
+				]
+
+				const mockClient = new MockBedrockClient(mockChunks)
+				const command = new ConverseStreamCommand({ modelId: "test-model", messages: [] })
+
+				// Replace getBedrockClient with our mock
+				const originalGetBedrockClient = handler["getBedrockClient"]
+				handler["getBedrockClient"] = async () => mockClient as any
+
+				const generator = handler["executeConverseStream"](command, mockModelInfo)
+				const results = await collectGeneratorResults(generator)
+
+				// Restore original method
+				handler["getBedrockClient"] = originalGetBedrockClient
+
+				// Verify thinking steps are yielded before the final answer
+				results.should.have.length(4)
+				results[0].type.should.equal("reasoning")
+				results[0].reasoning.should.equal("まず与えられた数値50.653の立方根を求める必要があります。")
+				results[1].type.should.equal("reasoning")
+				results[1].reasoning.should.equal("立方根を近似するために数値を3乗したときの誤差を調整していきます。")
+				results[2].type.should.equal("text")
+				results[2].text.should.equal("50.653の立方根は約3.707です。")
+				results[3].type.should.equal("usage")
+			})
+
+			it("should not parse thinking tags in text content", async () => {
+				const mockChunks = [
+					{ messageStart: { role: "assistant" } },
+					// Regular text that contains thinking tags should NOT be parsed as thinking
+					{
+						contentBlockDelta: {
+							delta: { text: "Let me explain <thinking>this is not real thinking</thinking> in the text." },
+							contentBlockIndex: 0,
+						},
+					},
+					{ contentBlockStop: { contentBlockIndex: 0 } },
+					{ messageStop: { stopReason: "end_turn" } },
+				]
+
+				const mockClient = new MockBedrockClient(mockChunks)
+				const command = new ConverseStreamCommand({ modelId: "test-model", messages: [] })
+
+				// Replace getBedrockClient with our mock
+				const originalGetBedrockClient = handler["getBedrockClient"]
+				handler["getBedrockClient"] = async () => mockClient as any
+
+				const generator = handler["executeConverseStream"](command, mockModelInfo)
+				const results = await collectGeneratorResults(generator)
+
+				// Restore original method
+				handler["getBedrockClient"] = originalGetBedrockClient
+
+				// Verify that thinking tags are treated as regular text
+				results.should.have.length(1)
+				results[0].type.should.equal("text")
+				results[0].text.should.equal("Let me explain <thinking>this is not real thinking</thinking> in the text.")
+			})
+
+			it("should handle thinking response with empty reasoning array", async () => {
+				const mockChunks = [
+					{ messageStart: { role: "assistant" } },
+					{
+						metadata: {
+							additionalModelResponseFields: {
+								thinkingResponse: {
+									reasoning: [],
+								},
+							},
+						},
+					},
+					{ contentBlockDelta: { delta: { text: "Direct response without thinking" }, contentBlockIndex: 0 } },
+					{ contentBlockStop: { contentBlockIndex: 0 } },
+					{ messageStop: { stopReason: "end_turn" } },
+				]
+
+				const mockClient = new MockBedrockClient(mockChunks)
+				const command = new ConverseStreamCommand({ modelId: "test-model", messages: [] })
+
+				// Replace getBedrockClient with our mock
+				const originalGetBedrockClient = handler["getBedrockClient"]
+				handler["getBedrockClient"] = async () => mockClient as any
+
+				const generator = handler["executeConverseStream"](command, mockModelInfo)
+				const results = await collectGeneratorResults(generator)
+
+				// Restore original method
+				handler["getBedrockClient"] = originalGetBedrockClient
+
+				// Verify only text is returned when reasoning array is empty
+				results.should.have.length(1)
+				results[0].type.should.equal("text")
+				results[0].text.should.equal("Direct response without thinking")
+			})
+
+			it("should handle thinking response interleaved with text chunks", async () => {
+				const mockChunks = [
+					{ messageStart: { role: "assistant" } },
+					// First, some thinking
+					{
+						metadata: {
+							additionalModelResponseFields: {
+								thinkingResponse: {
+									reasoning: [{ type: "text", text: "Initial thought process", signature: "sig1" }],
+								},
+							},
+						},
+					},
+					// Then some text
+					{ contentBlockDelta: { delta: { text: "Based on my analysis" }, contentBlockIndex: 0 } },
+					// More thinking
+					{
+						metadata: {
+							additionalModelResponseFields: {
+								thinkingResponse: {
+									reasoning: [{ type: "text", text: "Additional consideration", signature: "sig2" }],
+								},
+							},
+						},
+					},
+					// Final text
+					{ contentBlockDelta: { delta: { text: ", here is the answer." }, contentBlockIndex: 0 } },
+					{ contentBlockStop: { contentBlockIndex: 0 } },
+					{ messageStop: { stopReason: "end_turn" } },
+				]
+
+				const mockClient = new MockBedrockClient(mockChunks)
+				const command = new ConverseStreamCommand({ modelId: "test-model", messages: [] })
+
+				// Replace getBedrockClient with our mock
+				const originalGetBedrockClient = handler["getBedrockClient"]
+				handler["getBedrockClient"] = async () => mockClient as any
+
+				const generator = handler["executeConverseStream"](command, mockModelInfo)
+				const results = await collectGeneratorResults(generator)
+
+				// Restore original method
+				handler["getBedrockClient"] = originalGetBedrockClient
+
+				// Verify interleaved thinking and text
+				results.should.have.length(4)
+				results[0].type.should.equal("reasoning")
+				results[0].reasoning.should.equal("Initial thought process")
+				results[1].type.should.equal("text")
+				results[1].text.should.equal("Based on my analysis")
+				results[2].type.should.equal("reasoning")
+				results[2].reasoning.should.equal("Additional consideration")
+				results[3].type.should.equal("text")
+				results[3].text.should.equal(", here is the answer.")
+			})
+		})
+
+		describe("multiple content blocks", () => {
+			it("should handle multiple content blocks (reasoning + text)", async () => {
+				const mockChunks = [
+					{ messageStart: { role: "assistant" } },
+					// Text block only - reasoning is now in additionalModelResponseFields
+					{ contentBlockDelta: { delta: { text: "Here is " }, contentBlockIndex: 0 } },
+					{ contentBlockDelta: { delta: { text: "my response" }, contentBlockIndex: 0 } },
+					{ contentBlockStop: { contentBlockIndex: 0 } },
+					{ messageStop: { stopReason: "end_turn" } },
+				]
+
+				const mockClient = new MockBedrockClient(mockChunks)
+				const command = new ConverseStreamCommand({ modelId: "test-model", messages: [] })
+
+				// Replace getBedrockClient with our mock
+				const originalGetBedrockClient = handler["getBedrockClient"]
+				handler["getBedrockClient"] = async () => mockClient as any
+
+				const generator = handler["executeConverseStream"](command, mockModelInfo)
+				const results = await collectGeneratorResults(generator)
+
+				// Restore original method
+				handler["getBedrockClient"] = originalGetBedrockClient
+
+				// Verify text chunks are yielded correctly
+				results.should.have.length(2)
+				results[0].type.should.equal("text")
+				results[0].text.should.equal("Here is ")
+				results[1].type.should.equal("text")
+				results[1].text.should.equal("my response")
+			})
+
+			it("should handle real-world Japanese content", async () => {
+				const mockChunks = [
+					{ messageStart: { role: "assistant" } },
+					// Text block with Japanese response
+					{ contentBlockDelta: { delta: { text: "# 生成AIの仕組み - 10歳の君にも分かる説明" }, contentBlockIndex: 0 } },
+					{ contentBlockStop: { contentBlockIndex: 0 } },
+					{ messageStop: { stopReason: "end_turn" } },
+				]
+
+				const mockClient = new MockBedrockClient(mockChunks)
+				const command = new ConverseStreamCommand({ modelId: "test-model", messages: [] })
+
+				// Replace getBedrockClient with our mock
+				const originalGetBedrockClient = handler["getBedrockClient"]
+				handler["getBedrockClient"] = async () => mockClient as any
+
+				const generator = handler["executeConverseStream"](command, mockModelInfo)
+				const results = await collectGeneratorResults(generator)
+
+				// Restore original method
+				handler["getBedrockClient"] = originalGetBedrockClient
+
+				// Verify Japanese content is handled correctly
+				results.should.have.length(1)
+				results[0].type.should.equal("text")
+				results[0].text.should.equal("# 生成AIの仕組み - 10歳の君にも分かる説明")
+			})
+
+			it("should handle interleaved content blocks", async () => {
+				const mockChunks = [
+					{ messageStart: { role: "assistant" } },
+					// Interleaved text blocks
+					{ contentBlockDelta: { delta: { text: "Text 1" }, contentBlockIndex: 0 } },
+					{ contentBlockDelta: { delta: { text: " Text 2" }, contentBlockIndex: 0 } },
+					// Stop blocks
+					{ contentBlockStop: { contentBlockIndex: 0 } },
+					{ messageStop: { stopReason: "end_turn" } },
+				]
+
+				const mockClient = new MockBedrockClient(mockChunks)
+				const command = new ConverseStreamCommand({ modelId: "test-model", messages: [] })
+
+				// Replace getBedrockClient with our mock
+				const originalGetBedrockClient = handler["getBedrockClient"]
+				handler["getBedrockClient"] = async () => mockClient as any
+
+				const generator = handler["executeConverseStream"](command, mockModelInfo)
+				const results = await collectGeneratorResults(generator)
+
+				// Restore original method
+				handler["getBedrockClient"] = originalGetBedrockClient
+
+				// Verify text chunks are yielded correctly
+				results.should.have.length(2)
+				results[0].type.should.equal("text")
+				results[0].text.should.equal("Text 1")
+				results[1].type.should.equal("text")
+				results[1].text.should.equal(" Text 2")
+			})
+		})
+
+		describe("error handling", () => {
+			it("should handle internalServerException", async () => {
+				const mockChunks = [
+					{ messageStart: { role: "assistant" } },
+					{ internalServerException: { message: "Internal server error occurred" } },
+				]
+
+				const mockClient = new MockBedrockClient(mockChunks)
+				const command = new ConverseStreamCommand({ modelId: "test-model", messages: [] })
+
+				// Replace getBedrockClient with our mock
+				const originalGetBedrockClient = handler["getBedrockClient"]
+				handler["getBedrockClient"] = async () => mockClient as any
+
+				const generator = handler["executeConverseStream"](command, mockModelInfo)
+				const results = await collectGeneratorResults(generator)
+
+				// Restore original method
+				handler["getBedrockClient"] = originalGetBedrockClient
+
+				// Verify error was handled
+				results.should.have.length(1)
+				results[0].type.should.equal("text")
+				results[0].text.should.equal("[ERROR] Internal server error: Internal server error occurred")
+			})
+
+			it("should handle throttlingException", async () => {
+				const mockChunks = [
+					{ messageStart: { role: "assistant" } },
+					{ throttlingException: { message: "Rate limit exceeded" } },
+				]
+
+				const mockClient = new MockBedrockClient(mockChunks)
+				const command = new ConverseStreamCommand({ modelId: "test-model", messages: [] })
+
+				// Replace getBedrockClient with our mock
+				const originalGetBedrockClient = handler["getBedrockClient"]
+				handler["getBedrockClient"] = async () => mockClient as any
+
+				const generator = handler["executeConverseStream"](command, mockModelInfo)
+				const results = await collectGeneratorResults(generator)
+
+				// Restore original method
+				handler["getBedrockClient"] = originalGetBedrockClient
+
+				// Verify error was handled
+				results.should.have.length(1)
+				results[0].type.should.equal("text")
+				results[0].text.should.equal("[ERROR] Throttling error: Rate limit exceeded")
+			})
+		})
+
+		describe("usage tracking", () => {
+			it("should track usage with cache tokens", async () => {
+				const mockChunks = [
+					{ messageStart: { role: "assistant" } },
+					{ contentBlockDelta: { delta: { text: "Response" }, contentBlockIndex: 0 } },
+					{ contentBlockStop: { contentBlockIndex: 0 } },
+					{ messageStop: { stopReason: "end_turn" } },
+					{
+						metadata: {
+							usage: {
+								inputTokens: 100,
+								outputTokens: 50,
+								cacheReadInputTokens: 20,
+								cacheWriteInputTokens: 30,
+							},
+						},
+					},
+				]
+
+				const mockClient = new MockBedrockClient(mockChunks)
+				const command = new ConverseStreamCommand({ modelId: "test-model", messages: [] })
+
+				// Replace getBedrockClient with our mock
+				const originalGetBedrockClient = handler["getBedrockClient"]
+				handler["getBedrockClient"] = async () => mockClient as any
+
+				const generator = handler["executeConverseStream"](command, mockModelInfo)
+				const results = await collectGeneratorResults(generator)
+
+				// Restore original method
+				handler["getBedrockClient"] = originalGetBedrockClient
+
+				// Verify usage tracking
+				results.should.have.length(2)
+				results[0].type.should.equal("text")
+				results[0].text.should.equal("Response")
+				results[1].type.should.equal("usage")
+				results[1].inputTokens.should.equal(100)
+				results[1].outputTokens.should.equal(50)
+				results[1].cacheReadTokens.should.equal(20)
+				results[1].cacheWriteTokens.should.equal(30)
+			})
+		})
+	})
+
+	describe("getModelId", () => {
+		it("should return raw model ID for custom models", async () => {
+			const customOptions: ApiHandlerOptions = {
+				...mockOptions,
+				awsBedrockCustomSelected: true,
+				apiModelId:
+					"arn:aws:bedrock:us-west-2:123456789012:custom-model/anthropic.claude-3-5-sonnet-20241022-v2:0/Qk8MMyLmRd",
+			}
+			const customHandler = new AwsBedrockHandler(customOptions)
+
+			const modelId = await customHandler.getModelId()
+			modelId.should.equal(
+				"arn:aws:bedrock:us-west-2:123456789012:custom-model/anthropic.claude-3-5-sonnet-20241022-v2:0/Qk8MMyLmRd",
+			)
+		})
+
+		it("should not encode custom model IDs with slashes", async () => {
+			const customOptions: ApiHandlerOptions = {
+				...mockOptions,
+				awsBedrockCustomSelected: true,
+				apiModelId: "my-namespace/my-custom-model",
+			}
+			const customHandler = new AwsBedrockHandler(customOptions)
+
+			const modelId = await customHandler.getModelId()
+			modelId.should.equal("my-namespace/my-custom-model")
+			modelId.should.not.match(/%2F/)
+		})
+
+		it("should apply cross-region prefix for non-custom models when enabled", async () => {
+			const crossRegionOptions: ApiHandlerOptions = {
+				...mockOptions,
+				awsUseCrossRegionInference: true,
+				awsRegion: "us-west-2",
+			}
+			const crossRegionHandler = new AwsBedrockHandler(crossRegionOptions)
+
+			const modelId = await crossRegionHandler.getModelId()
+			modelId.should.equal("us.anthropic.claude-3-7-sonnet-20250219-v1:0")
+		})
+
+		it("should apply EU cross-region prefix", async () => {
+			const euOptions: ApiHandlerOptions = {
+				...mockOptions,
+				awsUseCrossRegionInference: true,
+				awsRegion: "eu-central-1",
+			}
+			const euHandler = new AwsBedrockHandler(euOptions)
+
+			const modelId = await euHandler.getModelId()
+			modelId.should.equal("eu.anthropic.claude-3-7-sonnet-20250219-v1:0")
+		})
+
+		it("should apply APAC cross-region prefix", async () => {
+			const apacOptions: ApiHandlerOptions = {
+				...mockOptions,
+				awsUseCrossRegionInference: true,
+				awsRegion: "ap-northeast-1",
+			}
+			const apacHandler = new AwsBedrockHandler(apacOptions)
+
+			const modelId = await apacHandler.getModelId()
+			modelId.should.equal("apac.anthropic.claude-3-7-sonnet-20250219-v1:0")
+		})
+
+		it("should not apply cross-region prefix for custom models even when enabled", async () => {
+			const customCrossRegionOptions: ApiHandlerOptions = {
+				...mockOptions,
+				awsBedrockCustomSelected: true,
+				apiModelId: "arn:aws:bedrock:us-west-2:123456789012:custom-model/my-model",
+				awsUseCrossRegionInference: true,
+			}
+			const customCrossRegionHandler = new AwsBedrockHandler(customCrossRegionOptions)
+
+			const modelId = await customCrossRegionHandler.getModelId()
+			modelId.should.equal("arn:aws:bedrock:us-west-2:123456789012:custom-model/my-model")
+		})
+
+		it("should handle UltraThink model ARN correctly", async () => {
+			const ultraThinkOptions: ApiHandlerOptions = {
+				...mockOptions,
+				awsBedrockCustomSelected: true,
+				apiModelId:
+					"arn:aws:bedrock:us-west-2:123456789012:custom-model/anthropic.claude-3-5-sonnet-20241022-v2:0/Qk8MMyLmRd",
+				awsBedrockCustomModelBaseId: "anthropic.claude-3-5-sonnet-20241022-v2:0",
+			}
+			const ultraThinkHandler = new AwsBedrockHandler(ultraThinkOptions)
+
+			const modelId = await ultraThinkHandler.getModelId()
+			// Should return the raw ARN without any encoding
+			modelId.should.equal(
+				"arn:aws:bedrock:us-west-2:123456789012:custom-model/anthropic.claude-3-5-sonnet-20241022-v2:0/Qk8MMyLmRd",
+			)
+			modelId.should.not.match(/%2F/)
+			modelId.should.not.match(/%3A/)
+		})
+	})
 })

src/api/providers/bedrock.ts

@@ -1,4 +1,3 @@
-import AnthropicBedrock from "@anthropic-ai/bedrock-sdk"
 import { Anthropic } from "@anthropic-ai/sdk"
 import { withRetry } from "../retry"
 import { ApiHandler } from "../"
@@ -14,6 +13,81 @@ import {
 	InvokeModelWithResponseStreamCommand,
 } from "@aws-sdk/client-bedrock-runtime"
 
+// Import proper AWS SDK types
+import type { Message, ContentBlock } from "@aws-sdk/client-bedrock-runtime"
+
+// Extend AWS SDK types to include additionalModelResponseFields
+interface ExtendedMetadata {
+	usage?: {
+		inputTokens?: number
+		outputTokens?: number
+		cacheReadInputTokens?: number
+		cacheWriteInputTokens?: number
+	}
+	additionalModelResponseFields?: {
+		thinkingResponse?: {
+			reasoning?: Array<{
+				type: string
+				text?: string
+				signature?: string
+			}>
+		}
+	}
+}
+
+// Define types for stream response content blocks
+interface ContentBlockStart {
+	contentBlockIndex?: number
+	start?: {
+		type?: string
+		thinking?: string
+	}
+	contentBlock?: {
+		type?: string
+		thinking?: string
+	}
+	type?: string
+	thinking?: string
+}
+
+// Define types for stream response deltas
+interface ContentBlockDelta {
+	contentBlockIndex?: number
+	delta?: {
+		type?: string
+		thinking?: string
+		text?: string
+		reasoningContent?: {
+			text?: string
+		}
+	}
+}
+
+// Define types for supported content types
+type SupportedContentType = "text" | "image" | "thinking"
+
+interface ContentItem {
+	type: SupportedContentType
+	text?: string
+	source?: {
+		data: string | Buffer | Uint8Array
+		media_type?: string
+	}
+}
+
+// Define cache point type for AWS Bedrock
+interface CachePointContentBlock {
+	cachePoint: {
+		type: "default"
+	}
+}
+
+// Define provider options type based on AWS SDK patterns
+interface ProviderChainOptions {
+	ignoreCache?: boolean
+	profile?: string
+}
+
 // https://docs.anthropic.com/en/api/claude-on-amazon-bedrock
 export class AwsBedrockHandler implements ApiHandler {
 	private options: ApiHandlerOptions
@@ -22,7 +96,7 @@ export class AwsBedrockHandler implements ApiHandler {
 		this.options = options
 	}
 
-	@withRetry()
+	@withRetry({ maxRetries: 4 })
 	async *createMessage(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]): ApiStream {
 		// cross region inference requires prefixing the model id with the region
 		const modelId = await this.getModelId()
@@ -46,146 +120,8 @@ export class AwsBedrockHandler implements ApiHandler {
 			return
 		}
 
-		const budget_tokens = this.options.thinkingBudgetTokens || 0
-		const reasoningOn =
-			(baseModelId.includes("3-7") || baseModelId.includes("sonnet-4") || baseModelId.includes("opus-4")) &&
-			budget_tokens !== 0
-				? true
-				: false
-
-		// Get model info and message indices for caching
-		const userMsgIndices = messages.reduce((acc, msg, index) => (msg.role === "user" ? [...acc, index] : acc), [] as number[])
-		const lastUserMsgIndex = userMsgIndices[userMsgIndices.length - 1] ?? -1
-		const secondLastMsgUserIndex = userMsgIndices[userMsgIndices.length - 2] ?? -1
-
-		// Create anthropic client, using sessions created or renewed after this handler's
-		// initialization, and allowing for session renewal if necessary as well
-		const client = await this.getAnthropicClient()
-
-		// Use withTempEnv to ensure environment variables are properly restored
-		const stream = await AwsBedrockHandler.withTempEnv(
-			() => {
-				// AWS SDK prioritizes AWS_PROFILE over AWS_ACCESS_KEY_ID/AWS_SECRET_ACCESS_KEY pair
-				// If this is set as an env variable already (ie. from ~/.zshrc) it will override credentials configured by Cline
-				// Temporarily remove AWS_PROFILE to ensure our credentials are used
-				delete process.env["AWS_PROFILE"]
-			},
-			async () => {
-				return await client.messages.create({
-					model: modelId,
-					max_tokens: model.info.maxTokens || 8192,
-					thinking: reasoningOn ? { type: "enabled", budget_tokens: budget_tokens } : undefined,
-					temperature: reasoningOn ? undefined : 0,
-					system: [
-						{
-							text: systemPrompt,
-							type: "text",
-							...(this.options.awsBedrockUsePromptCache === true && {
-								cache_control: { type: "ephemeral" },
-							}),
-						},
-					],
-					messages: messages.map((message, index) => {
-						if (index === lastUserMsgIndex || index === secondLastMsgUserIndex) {
-							return {
-								...message,
-								content:
-									typeof message.content === "string"
-										? [
-												{
-													type: "text",
-													text: message.content,
-													...(this.options.awsBedrockUsePromptCache === true && {
-														cache_control: { type: "ephemeral" },
-													}),
-												},
-											]
-										: message.content.map((content, contentIndex) =>
-												contentIndex === message.content.length - 1
-													? {
-															...content,
-															...(this.options.awsBedrockUsePromptCache === true && {
-																cache_control: { type: "ephemeral" },
-															}),
-														}
-													: content,
-											),
-							}
-						}
-						return message
-					}),
-					stream: true,
-				})
-			},
-		)
-
-		for await (const chunk of stream) {
-			switch (chunk?.type) {
-				case "message_start":
-					const usage = chunk.message.usage
-					yield {
-						type: "usage",
-						inputTokens: usage.input_tokens || 0,
-						outputTokens: usage.output_tokens || 0,
-						cacheWriteTokens: usage.cache_creation_input_tokens || undefined,
-						cacheReadTokens: usage.cache_read_input_tokens || undefined,
-					}
-					break
-				case "message_delta":
-					yield {
-						type: "usage",
-						inputTokens: 0,
-						outputTokens: chunk.usage.output_tokens || 0,
-					}
-					break
-				case "content_block_start":
-					switch (chunk.content_block.type) {
-						case "thinking":
-							yield {
-								type: "reasoning",
-								reasoning: chunk.content_block.thinking || "",
-							}
-							break
-						case "redacted_thinking":
-							// Handle redacted thinking blocks - we still mark it as reasoning
-							// but note that the content is encrypted
-							yield {
-								type: "reasoning",
-								reasoning: "[Redacted thinking block]",
-							}
-							break
-						case "text":
-							if (chunk.index > 0) {
-								yield {
-									type: "text",
-									text: "\n",
-								}
-							}
-							yield {
-								type: "text",
-								text: chunk.content_block.text,
-							}
-							break
-					}
-					break
-				case "content_block_delta":
-					switch (chunk.delta.type) {
-						case "thinking_delta":
-							yield {
-								type: "reasoning",
-								reasoning: chunk.delta.thinking,
-							}
-							break
-						case "text_delta":
-							yield {
-								type: "text",
-								text: chunk.delta.text,
-							}
-							break
-					}
-					break
-			}
-		}
+		// Default: Use Anthropic Converse API for all Anthropic models
+		yield* this.createAnthropicMessage(systemPrompt, messages, modelId, model)
 	}
 
 	getModel(): { id: string; info: ModelInfo } {
@@ -197,11 +133,20 @@ export class AwsBedrockHandler implements ApiHandler {
 
 		const customSelected = this.options.awsBedrockCustomSelected
 		const baseModel = this.options.awsBedrockCustomModelBaseId
-		if (customSelected && modelId && baseModel && baseModel in bedrockModels) {
-			// Use the user-input model ID but inherit capabilities from the base model
+
+		// Handle custom models
+		if (customSelected && modelId) {
+			// If base model is provided and valid, use its capabilities
+			if (baseModel && baseModel in bedrockModels) {
+				return {
+					id: modelId,
+					info: bedrockModels[baseModel],
+				}
+			}
+			// For custom models without valid base model in bedrock model list, use default model's capabilities
 			return {
 				id: modelId,
-				info: bedrockModels[baseModel],
+				info: bedrockModels[bedrockDefaultModelId],
 			}
 		}
 
@@ -224,7 +169,7 @@ export class AwsBedrockHandler implements ApiHandler {
 		sessionToken?: string
 	}> {
 		// Configure provider options
-		const providerOptions: any = {}
+		const providerOptions: ProviderChainOptions = {}
 		if (this.options.awsUseProfile) {
 			// For profile-based auth, always use ignoreCache to detect credential file changes
 			// This solves the AWS Identity Manager issue where credential files change externally
@@ -276,30 +221,11 @@ export class AwsBedrockHandler implements ApiHandler {
 		})
 	}
 
-	/**
-	 * Creates an AnthropicBedrock client with the appropriate credentials
-	 */
-	private async getAnthropicClient(): Promise<AnthropicBedrock> {
-		const credentials = await this.getAwsCredentials()
-
-		// Return an AnthropicBedrock client with the resolved/assumed credentials.
-		return new AnthropicBedrock({
-			awsAccessKey: credentials.accessKeyId,
-			awsSecretKey: credentials.secretAccessKey,
-			awsSessionToken: credentials.sessionToken,
-			awsRegion: this.getRegion(),
-			...(this.options.awsBedrockEndpoint && { baseURL: this.options.awsBedrockEndpoint }),
-		})
-	}
-
 	/**
 	 * Gets the appropriate model ID, accounting for cross-region inference if enabled.
-	 * If the model ID is an ARN that contains a slash, you will get the URL encoded ARN.
+	 * For custom models, returns the raw model ID without any encoding.
 	 */
 	async getModelId(): Promise<string> {
-		if (this.options.awsBedrockCustomSelected && this.getModel().id.includes("/")) {
-			return encodeURIComponent(this.getModel().id)
-		}
 		if (!this.options.awsBedrockCustomSelected && this.options.awsUseCrossRegionInference) {
 			const regionPrefix = this.getRegion().slice(0, 3)
 			switch (regionPrefix) {
@@ -536,126 +462,318 @@ export class AwsBedrockHandler implements ApiHandler {
 	}
 
 	/**
-	 * Creates a message using Amazon Nova models through AWS Bedrock
-	 * Implements support for Amazon Nova models
+	 * Executes a Converse API stream command and handles the response
+	 * Common implementation for both Anthropic and Nova models
 	 */
-	private async *createNovaMessage(
-		systemPrompt: string,
-		messages: Anthropic.Messages.MessageParam[],
-		modelId: string,
-		model: { id: string; info: ModelInfo },
-	): ApiStream {
-		// Get Bedrock client with proper credentials
-		const client = await this.getBedrockClient()
-
-		// Format messages for Nova model
-		const formattedMessages = this.formatNovaMessages(messages)
-
-		// Prepare request for Nova model
-		const command = new ConverseStreamCommand({
-			modelId: modelId,
-			messages: formattedMessages,
-			system: systemPrompt ? [{ text: systemPrompt }] : undefined,
-			inferenceConfig: {
-				maxTokens: model.info.maxTokens || 5000,
-				temperature: 0,
-				// topP: 0.9, // Alternative: use topP instead of temperature
-			},
-		})
-
-		// Execute the streaming request and handle response
+	private async *executeConverseStream(command: ConverseStreamCommand, modelInfo: ModelInfo): ApiStream {
 		try {
+			const client = await this.getBedrockClient()
 			const response = await client.send(command)
 
 			if (response.stream) {
-				let hasReportedInputTokens = false
+				// Buffer content by contentBlockIndex to handle multi-block responses correctly
+				const contentBuffers: Record<number, string> = {}
+				const blockTypes = new Map<number, "reasoning" | "text">()
 
 				for await (const chunk of response.stream) {
+					// Debug logging to see actual response structure
+					// console.log("Bedrock chunk:", JSON.stringify(chunk, null, 2))
+
+					// Handle thinking response in additionalModelResponseFields (LangChain format)
+					const metadata = chunk.metadata as ExtendedMetadata | undefined
+					if (metadata?.additionalModelResponseFields?.thinkingResponse) {
+						const thinkingResponse = metadata.additionalModelResponseFields.thinkingResponse
+						if (thinkingResponse.reasoning && Array.isArray(thinkingResponse.reasoning)) {
+							for (const reasoningBlock of thinkingResponse.reasoning) {
+								if (reasoningBlock.type === "text" && reasoningBlock.text) {
+									yield {
+										type: "reasoning",
+										reasoning: reasoningBlock.text,
+									}
+								}
+							}
+						}
+					}
+
 					// Handle metadata events with token usage information
 					if (chunk.metadata?.usage) {
-						// Report complete token usage from the model itself
 						const inputTokens = chunk.metadata.usage.inputTokens || 0
 						const outputTokens = chunk.metadata.usage.outputTokens || 0
+						const cacheReadInputTokens = chunk.metadata.usage.cacheReadInputTokens || 0
+						const cacheWriteInputTokens = chunk.metadata.usage.cacheWriteInputTokens || 0
+
 						yield {
 							type: "usage",
 							inputTokens,
 							outputTokens,
-							totalCost: calculateApiCostOpenAI(model.info, inputTokens, outputTokens, 0, 0),
-						}
-						hasReportedInputTokens = true
-					}
-
-					// Handle content delta (text generation)
-					if (chunk.contentBlockDelta?.delta?.text) {
-						yield {
-							type: "text",
-							text: chunk.contentBlockDelta.delta.text,
+							cacheReadTokens: cacheReadInputTokens,
+							cacheWriteTokens: cacheWriteInputTokens,
+							totalCost: calculateApiCostOpenAI(
+								modelInfo,
+								inputTokens,
+								outputTokens,
+								cacheWriteInputTokens,
+								cacheReadInputTokens,
+							),
 						}
 					}
 
-					// Handle reasoning content if present
-					if (chunk.contentBlockDelta?.delta?.reasoningContent?.text) {
-						yield {
-							type: "reasoning",
-							reasoning: chunk.contentBlockDelta.delta.reasoningContent.text,
+					// Handle content block start - check if Bedrock uses Anthropic SDK format
+					if (chunk.contentBlockStart) {
+						const blockStart = chunk.contentBlockStart as ContentBlockStart
+						const blockIndex = chunk.contentBlockStart.contentBlockIndex
+
+						// Check for thinking block in various possible formats
+						if (
+							blockStart.start?.type === "thinking" ||
+							blockStart.contentBlock?.type === "thinking" ||
+							blockStart.type === "thinking"
+						) {
+							if (blockIndex !== undefined) {
+								blockTypes.set(blockIndex, "reasoning")
+								// Initialize content if provided
+								const initialContent =
+									blockStart.start?.thinking || blockStart.contentBlock?.thinking || blockStart.thinking || ""
+								if (initialContent) {
+									yield {
+										type: "reasoning",
+										reasoning: initialContent,
+									}
+								}
+							}
 						}
 					}
 
-					// Handle errors
-					if (chunk.internalServerException) {
-						yield {
-							type: "text",
-							text: `[ERROR] Internal server error: ${chunk.internalServerException.message}`,
-						}
-					} else if (chunk.modelStreamErrorException) {
-						yield {
-							type: "text",
-							text: `[ERROR] Model stream error: ${chunk.modelStreamErrorException.message}`,
-						}
-					} else if (chunk.validationException) {
-						yield {
-							type: "text",
-							text: `[ERROR] Validation error: ${chunk.validationException.message}`,
-						}
-					} else if (chunk.throttlingException) {
-						yield {
-							type: "text",
-							text: `[ERROR] Throttling error: ${chunk.throttlingException.message}`,
-						}
-					} else if (chunk.serviceUnavailableException) {
-						yield {
-							type: "text",
-							text: `[ERROR] Service unavailable: ${chunk.serviceUnavailableException.message}`,
+					// Handle content block delta - accumulate content by block index
+					if (chunk.contentBlockDelta) {
+						const blockIndex = chunk.contentBlockDelta.contentBlockIndex
+
+						if (blockIndex !== undefined) {
+							// Initialize buffer for this block if it doesn't exist
+							if (!(blockIndex in contentBuffers)) {
+								contentBuffers[blockIndex] = ""
+							}
+
+							// Check if this is a thinking block
+							const blockType = blockTypes.get(blockIndex)
+							const delta = chunk.contentBlockDelta.delta as ContentBlockDelta["delta"]
+
+							// Handle thinking delta (Anthropic SDK format)
+							if (delta?.type === "thinking_delta" || delta?.thinking) {
+								const thinkingContent = delta.thinking || delta.text || ""
+								if (thinkingContent) {
+									yield {
+										type: "reasoning",
+										reasoning: thinkingContent,
+									}
+								}
+							} else if (delta?.reasoningContent?.text) {
+								// Handle reasoning content (Bedrock format)
+								const reasoningText = delta.reasoningContent.text
+								if (reasoningText) {
+									yield {
+										type: "reasoning",
+										reasoning: reasoningText,
+									}
+								}
+							} else if (chunk.contentBlockDelta.delta?.text) {
+								// Handle regular text content
+								const textContent = chunk.contentBlockDelta.delta.text
+								contentBuffers[blockIndex] += textContent
+
+								// Stream based on block type
+								if (blockType === "reasoning") {
+									yield {
+										type: "reasoning",
+										reasoning: textContent,
+									}
+								} else {
+									yield {
+										type: "text",
+										text: textContent,
+									}
+								}
+							}
 						}
 					}
+
+					// Handle content block stop - clean up buffers
+					if (chunk.contentBlockStop) {
+						const blockIndex = chunk.contentBlockStop.contentBlockIndex
+
+						if (blockIndex !== undefined) {
+							// Clean up buffers and tracking for this block
+							delete contentBuffers[blockIndex]
+							blockTypes.delete(blockIndex)
+						}
+					}
+
+					// Handle errors with unified error handling
+					yield* this.handleBedrockStreamError(chunk)
 				}
 			}
 		} catch (error) {
-			console.error("Error processing Nova model response:", error)
+			console.error("Error processing Converse API response:", error)
 			yield {
 				type: "text",
-				text: `[ERROR] Failed to process Nova response: ${error instanceof Error ? error.message : String(error)}`,
+				text: `[ERROR] Failed to process response: ${error instanceof Error ? error.message : String(error)}`,
 			}
 		}
 	}
 
 	/**
-	 * Formats messages for Amazon Nova models according to the SDK specification
+	 * Handles Bedrock stream errors in a unified way
 	 */
-	private formatNovaMessages(messages: Anthropic.Messages.MessageParam[]): { role: ConversationRole; content: any[] }[] {
+	private *handleBedrockStreamError(chunk: any): Generator<{ type: "text"; text: string }> {
+		if (chunk.internalServerException) {
+			yield {
+				type: "text",
+				text: `[ERROR] Internal server error: ${chunk.internalServerException.message}`,
+			}
+		} else if (chunk.modelStreamErrorException) {
+			yield {
+				type: "text",
+				text: `[ERROR] Model stream error: ${chunk.modelStreamErrorException.message}`,
+			}
+		} else if (chunk.validationException) {
+			yield {
+				type: "text",
+				text: `[ERROR] Validation error: ${chunk.validationException.message}`,
+			}
+		} else if (chunk.throttlingException) {
+			yield {
+				type: "text",
+				text: `[ERROR] Throttling error: ${chunk.throttlingException.message}`,
+			}
+		} else if (chunk.serviceUnavailableException) {
+			yield {
+				type: "text",
+				text: `[ERROR] Service unavailable: ${chunk.serviceUnavailableException.message}`,
+			}
+		}
+	}
+
+	/**
+	 * Prepares system messages with optional caching support
+	 */
+	private prepareSystemMessages(systemPrompt: string, enableCaching: boolean): any[] | undefined {
+		if (!systemPrompt) {
+			return undefined
+		}
+
+		if (enableCaching) {
+			return [{ text: systemPrompt }, { cachePoint: { type: "default" } }]
+		}
+
+		return [{ text: systemPrompt }]
+	}
+
+	/**
+	 * Gets inference configuration for different model types
+	 */
+	private getInferenceConfig(modelInfo: ModelInfo, modelType: "anthropic" | "nova"): any {
+		// For Anthropic models with thinking enabled, temperature must be 1
+		if (modelType === "anthropic") {
+			const budget_tokens = this.options.thinkingBudgetTokens || 0
+			const baseModelId =
+				(this.options.awsBedrockCustomSelected ? this.options.awsBedrockCustomModelBaseId : this.getModel().id) ||
+				this.getModel().id
+			const reasoningOn = this.shouldEnableReasoning(baseModelId, budget_tokens)
+
+			return {
+				maxTokens: modelInfo.maxTokens || 8192,
+				temperature: reasoningOn ? 1 : 0,
+			}
+		}
+
+		return {
+			maxTokens: modelInfo.maxTokens || (modelType === "nova" ? 5000 : 8192),
+			temperature: 0,
+		}
+	}
+
+	/**
+	 * Determines if reasoning should be enabled for Claude models
+	 */
+	private shouldEnableReasoning(baseModelId: string, budgetTokens: number): boolean {
+		return (
+			(baseModelId.includes("3-7") || baseModelId.includes("sonnet-4") || baseModelId.includes("opus-4")) &&
+			budgetTokens !== 0
+		)
+	}
+
+	/**
+	 * Creates a message using Anthropic Claude models through AWS Bedrock Converse API
+	 * Implements support for Anthropic Claude models using the unified Converse API
+	 */
+	private async *createAnthropicMessage(
+		systemPrompt: string,
+		messages: Anthropic.Messages.MessageParam[],
+		modelId: string,
+		model: { id: string; info: ModelInfo },
+	): ApiStream {
+		// Format messages for Anthropic model using unified formatter
+		const formattedMessages = this.formatMessagesForConverseAPI(messages)
+
+		// Get model info and message indices for caching
+		const userMsgIndices = messages.reduce((acc, msg, index) => (msg.role === "user" ? [...acc, index] : acc), [] as number[])
+		const lastUserMsgIndex = userMsgIndices[userMsgIndices.length - 1] ?? -1
+		const secondLastMsgUserIndex = userMsgIndices[userMsgIndices.length - 2] ?? -1
+
+		// Apply caching controls to messages if enabled
+		const messagesWithCache = this.options.awsBedrockUsePromptCache
+			? this.applyCacheControlToMessages(formattedMessages, lastUserMsgIndex, secondLastMsgUserIndex)
+			: formattedMessages
+
+		// Prepare system message with caching support
+		const systemMessages = this.prepareSystemMessages(systemPrompt, this.options.awsBedrockUsePromptCache || false)
+
+		// Get thinking configuration
+		const budget_tokens = this.options.thinkingBudgetTokens || 0
+		const baseModelId =
+			(this.options.awsBedrockCustomSelected ? this.options.awsBedrockCustomModelBaseId : this.getModel().id) ||
+			this.getModel().id
+		const reasoningOn = this.shouldEnableReasoning(baseModelId, budget_tokens)
+
+		// Prepare request for Anthropic model using Converse API
+		const command = new ConverseStreamCommand({
+			modelId: modelId,
+			messages: messagesWithCache,
+			system: systemMessages,
+			inferenceConfig: this.getInferenceConfig(model.info, "anthropic"),
+			// Add thinking configuration as per LangChain documentation
+			additionalModelRequestFields: reasoningOn
+				? {
+						thinking: {
+							type: "enabled",
+							budget_tokens: budget_tokens,
+						},
+					}
+				: undefined,
+		})
+
+		// Execute the streaming request using unified handler
+		yield* this.executeConverseStream(command, model.info)
+	}
+
+	/**
+	 * Formats messages for models using the Converse API specification
+	 * Used by both Anthropic and Nova models to avoid code duplication
+	 */
+	private formatMessagesForConverseAPI(messages: Anthropic.Messages.MessageParam[]): Message[] {
 		return messages.map((message) => {
 			// Determine role (user or assistant)
 			const role = message.role === "user" ? ConversationRole.USER : ConversationRole.ASSISTANT
 
 			// Process content based on type
-			let content: any[] = []
+			let content: ContentBlock[] = []
 
 			if (typeof message.content === "string") {
 				// Simple text content
 				content = [{ text: message.content }]
 			} else if (Array.isArray(message.content)) {
-				// Convert Anthropic content format to Nova content format
-				content = message.content
+				// Convert Anthropic content format to Converse API content format
+				const processedContent = message.content
 					.map((item) => {
 						// Text content
 						if (item.type === "text") {
@@ -664,55 +782,16 @@ export class AwsBedrockHandler implements ApiHandler {
 
 						// Image content
 						if (item.type === "image") {
-							// Handle different image source formats
-							let imageData: Uint8Array
-							let format = "jpeg" // default format
-
-							// Extract format from media_type if available
-							if (item.source.media_type) {
-								// Extract format from media_type (e.g., "image/jpeg" -> "jpeg")
-								const formatMatch = item.source.media_type.match(/image\/(\w+)/)
-								if (formatMatch && formatMatch[1]) {
-									format = formatMatch[1]
-									// Ensure format is one of the allowed values
-									if (!["png", "jpeg", "gif", "webp"].includes(format)) {
-										format = "jpeg" // Default to jpeg if not supported
-									}
-								}
-							}
-
-							// Get image data
-							try {
-								if (typeof item.source.data === "string") {
-									// Handle base64 encoded data
-									const base64Data = item.source.data.replace(/^data:image\/\w+;base64,/, "")
-									imageData = new Uint8Array(Buffer.from(base64Data, "base64"))
-								} else if (item.source.data && typeof item.source.data === "object") {
-									// Try to convert to Uint8Array
-									imageData = new Uint8Array(Buffer.from(item.source.data as any))
-								} else {
-									console.error("Unsupported image data format")
-									return null // Skip this item if format is not supported
-								}
-							} catch (error) {
-								console.error("Could not convert image data to Uint8Array:", error)
-								return null // Skip this item if conversion fails
-							}
-
-							return {
-								image: {
-									format,
-									source: {
-										bytes: imageData,
-									},
-								},
-							}
+							return this.processImageContent(item)
 						}
 
-						// Return null for unsupported content types
+						// Log unsupported content types for debugging
+						console.warn(`Unsupported content type: ${(item as ContentItem).type}`)
 						return null
 					})
-					.filter(Boolean) // Remove any null items
+					.filter((item): item is ContentBlock => item !== null)
+
+				content = processedContent
 			}
 
 			// Return formatted message
@@ -722,4 +801,129 @@ export class AwsBedrockHandler implements ApiHandler {
 			}
 		})
 	}
+
+	/**
+	 * Processes image content with proper error handling and user notification
+	 */
+	private processImageContent(item: any): ContentBlock | null {
+		let imageData: Uint8Array
+		let format: "png" | "jpeg" | "gif" | "webp" = "jpeg" // default format
+
+		// Extract format from media_type if available
+		if (item.source.media_type) {
+			// Extract format from media_type (e.g., "image/jpeg" -> "jpeg")
+			const formatMatch = item.source.media_type.match(/image\/(\w+)/)
+			if (formatMatch && formatMatch[1]) {
+				const extractedFormat = formatMatch[1]
+				// Ensure format is one of the allowed values
+				if (["png", "jpeg", "gif", "webp"].includes(extractedFormat)) {
+					format = extractedFormat as "png" | "jpeg" | "gif" | "webp"
+				}
+			}
+		}
+
+		// Get image data with improved error handling
+		try {
+			if (typeof item.source.data === "string") {
+				// Handle base64 encoded data
+				const base64Data = item.source.data.replace(/^data:image\/\w+;base64,/, "")
+				imageData = new Uint8Array(Buffer.from(base64Data, "base64"))
+			} else if (item.source.data && typeof item.source.data === "object") {
+				// Try to convert to Uint8Array
+				imageData = new Uint8Array(Buffer.from(item.source.data as Buffer | Uint8Array))
+			} else {
+				throw new Error("Unsupported image data format")
+			}
+
+			return {
+				image: {
+					format,
+					source: {
+						bytes: imageData,
+					},
+				},
+			}
+		} catch (error) {
+			console.error("Failed to process image content:", error)
+			// Return a text content indicating the error instead of null
+			// This ensures users are aware of the issue
+			return {
+				text: `[ERROR: Failed to process image - ${error instanceof Error ? error.message : "Unknown error"}]`,
+			}
+		}
+	}
+
+	/**
+	 * Applies cache control to messages for prompt caching using AWS Bedrock's cachePoint system
+	 * AWS Bedrock uses cachePoint objects instead of Anthropic's cache_control approach
+	 */
+	private applyCacheControlToMessages(
+		messages: Message[],
+		lastUserMsgIndex: number,
+		secondLastMsgUserIndex: number,
+	): Message[] {
+		return messages.map((message, index) => {
+			// Add cachePoint to the last user message and second-to-last user message
+			if (index === lastUserMsgIndex || index === secondLastMsgUserIndex) {
+				// Clone the message to avoid modifying the original
+				const messageWithCache = { ...message }
+
+				if (messageWithCache.content && Array.isArray(messageWithCache.content)) {
+					// Add cachePoint to the end of the content array
+					messageWithCache.content = [
+						...messageWithCache.content,
+						{
+							cachePoint: {
+								type: "default",
+							},
+						} as CachePointContentBlock, // Properly typed cache point for AWS SDK
+					]
+				}
+
+				return messageWithCache
+			}
+
+			return message
+		})
+	}
+
+	/**
+	 * Creates a message using Amazon Nova models through AWS Bedrock
+	 * Implements support for Amazon Nova models with caching support
+	 */
+	private async *createNovaMessage(
+		systemPrompt: string,
+		messages: Anthropic.Messages.MessageParam[],
+		modelId: string,
+		model: { id: string; info: ModelInfo },
+	): ApiStream {
+		// Format messages for Nova model using unified formatter
+		const formattedMessages = this.formatMessagesForConverseAPI(messages)
+
+		// Get model info and message indices for caching (for Nova models that support it)
+		const userMsgIndices = messages.reduce((acc, msg, index) => (msg.role === "user" ? [...acc, index] : acc), [] as number[])
+		const lastUserMsgIndex = userMsgIndices[userMsgIndices.length - 1] ?? -1
+		const secondLastMsgUserIndex = userMsgIndices[userMsgIndices.length - 2] ?? -1
+
+		// Apply caching controls to messages if model supports caching and option is enabled
+		const messagesWithCache =
+			this.options.awsBedrockUsePromptCache && model.info.supportsPromptCache
+				? this.applyCacheControlToMessages(formattedMessages, lastUserMsgIndex, secondLastMsgUserIndex)
+				: formattedMessages
+
+		// Prepare system message with caching support for Nova models that support it
+		const enableCaching = this.options.awsBedrockUsePromptCache && model.info.supportsPromptCache
+		const systemMessages = this.prepareSystemMessages(systemPrompt, enableCaching || false)
+
+		// Prepare request for Nova model
+		const command = new ConverseStreamCommand({
+			modelId: modelId,
+			messages: messagesWithCache,
+			system: systemMessages,
+			inferenceConfig: this.getInferenceConfig(model.info, "nova"),
+		})
+
+		// Execute the streaming request using unified handler
+		yield* this.executeConverseStream(command, model.info)
+	}
 }

src/api/providers/claude-code.ts

@@ -0,0 +1,165 @@
+import type { Anthropic } from "@anthropic-ai/sdk"
+import { claudeCodeDefaultModelId, ClaudeCodeModelId, claudeCodeModels, type ApiHandlerOptions } from "@/shared/api"
+import { type ApiHandler } from ".."
+import { ApiStreamUsageChunk, type ApiStream } from "../transform/stream"
+import { withRetry } from "../retry"
+import { runClaudeCode } from "@/integrations/claude-code/run"
+import { ClaudeCodeMessage } from "@/integrations/claude-code/types"
+
+export class ClaudeCodeHandler implements ApiHandler {
+	private options: ApiHandlerOptions
+
+	constructor(options: ApiHandlerOptions) {
+		this.options = options
+	}
+
+	@withRetry({
+		maxRetries: 4,
+		baseDelay: 2000,
+		maxDelay: 15000,
+	})
+	async *createMessage(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]): ApiStream {
+		const claudeProcess = runClaudeCode({
+			systemPrompt,
+			messages,
+			path: this.options.claudeCodePath,
+			modelId: this.getModel().id,
+		})
+
+		const dataQueue: string[] = []
+		let processError = null
+		let errorOutput = ""
+		let exitCode: number | null = null
+
+		claudeProcess.stdout.on("data", (data) => {
+			const output = data.toString()
+			const lines = output.split("\n").filter((line: string) => line.trim() !== "")
+
+			for (const line of lines) {
+				dataQueue.push(line)
+			}
+		})
+
+		claudeProcess.stderr.on("data", (data) => {
+			errorOutput += data.toString()
+		})
+
+		claudeProcess.on("close", (code) => {
+			exitCode = code
+		})
+
+		claudeProcess.on("error", (error) => {
+			processError = error
+		})
+
+		// Usage is included with assistant messages,
+		// but cost is included in the result chunk
+		let usage: ApiStreamUsageChunk = {
+			type: "usage",
+			inputTokens: 0,
+			outputTokens: 0,
+			cacheReadTokens: 0,
+			cacheWriteTokens: 0,
+		}
+
+		while (exitCode !== 0 || dataQueue.length > 0) {
+			if (dataQueue.length === 0) {
+				await new Promise((resolve) => setImmediate(resolve))
+			}
+
+			if (exitCode !== null && exitCode !== 0) {
+				throw new Error(
+					`Claude Code process exited with code ${exitCode}.${errorOutput ? ` Error output: ${errorOutput.trim()}` : ""}`,
+				)
+			}
+
+			const data = dataQueue.shift()
+			if (!data) {
+				continue
+			}
+
+			const chunk = this.attemptParseChunk(data)
+
+			if (!chunk) {
+				yield {
+					type: "text",
+					text: data || "",
+				}
+
+				continue
+			}
+
+			if (chunk.type === "system" && chunk.subtype === "init") {
+				continue
+			}
+
+			if (chunk.type === "assistant" && "message" in chunk) {
+				const message = chunk.message
+
+				if (message.stop_reason !== null && message.stop_reason !== "tool_use") {
+					const errorMessage = message.content[0]?.text || `Claude Code stopped with reason: ${message.stop_reason}`
+
+					if (errorMessage.includes("Invalid model name")) {
+						throw new Error(
+							errorMessage +
+								`\n\nAPI keys and subscription plans allow different models. Make sure the selected model is included in your plan.`,
+						)
+					}
+
+					throw new Error(errorMessage)
+				}
+
+				for (const content of message.content) {
+					if (content.type === "text") {
+						yield {
+							type: "text",
+							text: content.text,
+						}
+					} else {
+						console.warn("Unsupported content type:", content.type)
+					}
+				}
+
+				usage.inputTokens += message.usage.input_tokens
+				usage.outputTokens += message.usage.output_tokens
+				usage.cacheReadTokens = (usage.cacheReadTokens || 0) + (message.usage.cache_read_input_tokens || 0)
+				usage.cacheWriteTokens = (usage.cacheWriteTokens || 0) + (message.usage.cache_creation_input_tokens || 0)
+
+				continue
+			}
+
+			if (chunk.type === "result" && "result" in chunk) {
+				usage.totalCost = chunk.cost_usd || 0
+
+				yield usage
+			}
+
+			if (processError) {
+				throw processError
+			}
+		}
+	}
+
+	getModel() {
+		const modelId = this.options.apiModelId
+		if (modelId && modelId in claudeCodeModels) {
+			const id = modelId as ClaudeCodeModelId
+			return { id, info: claudeCodeModels[id] }
+		}
+
+		return {
+			id: claudeCodeDefaultModelId,
+			info: claudeCodeModels[claudeCodeDefaultModelId],
+		}
+	}
+
+	// TOOD: Validate instead of parsing
+	private attemptParseChunk(data: string): ClaudeCodeMessage | null {
+		try {
+			return JSON.parse(data)
+		} catch (error) {
+			console.error("Error parsing chunk:", error)
+			return null
+		}
+	}
+}

src/api/providers/cline.ts

@@ -74,6 +74,16 @@ export class ClineHandler implements ApiHandler {
 			}
 
 			if (!didOutputUsage && chunk.usage) {
+				// @ts-ignore-next-line
+				let totalCost = chunk.usage.cost || 0
+				const modelId = this.getModel().id
+				const provider = modelId.split("/")[0]
+
+				// If provider is x-ai, set totalCost to 0 (we're doing a promo)
+				if (provider === "x-ai") {
+					totalCost = 0
+				}
+
 				yield {
 					type: "usage",
 					cacheWriteTokens: 0,
@@ -81,7 +91,7 @@ export class ClineHandler implements ApiHandler {
 					inputTokens: chunk.usage.prompt_tokens || 0,
 					outputTokens: chunk.usage.completion_tokens || 0,
 					// @ts-ignore-next-line
-					totalCost: chunk.usage.cost || 0,
+					totalCost,
 				}
 				didOutputUsage = true
 			}
@@ -124,7 +134,10 @@ export class ClineHandler implements ApiHandler {
 	}
 
 	getModel(): { id: string; info: ModelInfo } {
-		const modelId = this.options.openRouterModelId
+		let modelId = this.options.openRouterModelId
+		if (modelId === "x-ai/grok-3") {
+			modelId = "x-ai/grok-3-beta"
+		}
 		const modelInfo = this.options.openRouterModelInfo
 		if (modelId && modelInfo) {
 			return { id: modelId, info: modelInfo }

src/api/providers/gemini.ts

@@ -2,6 +2,7 @@ import type { Anthropic } from "@anthropic-ai/sdk"
 // Restore GenerateContentConfig import and add GenerateContentResponseUsageMetadata
 import { GoogleGenAI, type GenerateContentConfig, type GenerateContentResponseUsageMetadata } from "@google/genai"
 import { withRetry } from "../retry"
+import { Part } from "@google/genai"
 import { ApiHandler } from "../"
 import { ApiHandlerOptions, geminiDefaultModelId, GeminiModelId, geminiModels, ModelInfo } from "@shared/api"
 import { convertAnthropicMessageToGemini } from "../transform/gemini-format"
@@ -96,9 +97,10 @@ export class GeminiHandler implements ApiHandler {
 		}
 
 		// Add thinking config if the model supports it
-		if (info.thinkingConfig?.outputPrice !== undefined && maxBudget > 0) {
+		if (thinkingBudget > 0) {
 			requestConfig.thinkingConfig = {
 				thinkingBudget: thinkingBudget,
+				includeThoughts: true,
 			}
 		}
 
@@ -111,6 +113,7 @@ export class GeminiHandler implements ApiHandler {
 		let promptTokens = 0
 		let outputTokens = 0
 		let cacheReadTokens = 0
+		let thoughtsTokenCount = 0 // Initialize thought token counts
 		let lastUsageMetadata: GenerateContentResponseUsageMetadata | undefined
 
 		try {
@@ -130,6 +133,31 @@ export class GeminiHandler implements ApiHandler {
 					isFirstSdkChunk = false
 				}
 
+				// Handle thinking content from Gemini's response
+				const candidateForThoughts = chunk?.candidates?.[0]
+				const partsForThoughts = candidateForThoughts?.content?.parts
+				let thoughts = "" // Initialize as empty string
+
+				if (partsForThoughts) {
+					// This ensures partsForThoughts is a Part[] array
+					for (const part of partsForThoughts) {
+						const { thought, text } = part as Part
+						if (thought && text) {
+							// Ensure part.text exists
+							// Handle the thought part
+							thoughts += text + "\n" // Append thought and a newline
+						}
+					}
+				}
+
+				if (thoughts.trim() !== "") {
+					yield {
+						type: "reasoning",
+						reasoning: thoughts.trim(),
+					}
+					thoughts = "" // Reset thoughts after yielding
+				}
+
 				if (chunk.text) {
 					yield {
 						type: "text",
@@ -141,6 +169,7 @@ export class GeminiHandler implements ApiHandler {
 					lastUsageMetadata = chunk.usageMetadata
 					promptTokens = lastUsageMetadata.promptTokenCount ?? promptTokens
 					outputTokens = lastUsageMetadata.candidatesTokenCount ?? outputTokens
+					thoughtsTokenCount = lastUsageMetadata.thoughtsTokenCount ?? thoughtsTokenCount
 					cacheReadTokens = lastUsageMetadata.cachedContentTokenCount ?? cacheReadTokens
 				}
 			}
@@ -151,12 +180,14 @@ export class GeminiHandler implements ApiHandler {
 					info,
 					inputTokens: promptTokens,
 					outputTokens,
+					thoughtsTokenCount,
 					cacheReadTokens,
 				})
 				yield {
 					type: "usage",
-					inputTokens: promptTokens,
+					inputTokens: promptTokens - cacheReadTokens,
 					outputTokens,
+					thoughtsTokenCount,
 					cacheReadTokens,
 					cacheWriteTokens: 0,
 					totalCost,
@@ -239,11 +270,13 @@ export class GeminiHandler implements ApiHandler {
 		info,
 		inputTokens,
 		outputTokens,
+		thoughtsTokenCount = 0,
 		cacheReadTokens = 0,
 	}: {
 		info: ModelInfo
 		inputTokens: number
 		outputTokens: number
+		thoughtsTokenCount: number
 		cacheReadTokens?: number
 	}) {
 		// Exit early if any required pricing information is missing
@@ -275,18 +308,18 @@ export class GeminiHandler implements ApiHandler {
 		const inputTokensCost = inputPrice * (uncachedInputTokens / 1_000_000)
 
 		// 2. Output token costs
-		const outputTokensCost = outputPrice * (outputTokens / 1_000_000)
+		const responseTokensCost = outputPrice * ((outputTokens + thoughtsTokenCount) / 1_000_000)
 
 		// 3. Cache read costs (immediate)
 		const cacheReadCost = (cacheReadTokens ?? 0) > 0 ? cacheReadsPrice * ((cacheReadTokens ?? 0) / 1_000_000) : 0
 
 		// Calculate total immediate cost (excluding cache write/storage costs)
-		const totalCost = inputTokensCost + outputTokensCost + cacheReadCost
+		const totalCost = inputTokensCost + responseTokensCost + cacheReadCost
 
 		// Create the trace object for debugging
 		const trace: Record<string, { price: number; tokens: number; cost: number }> = {
 			input: { price: inputPrice, tokens: uncachedInputTokens, cost: inputTokensCost },
-			output: { price: outputPrice, tokens: outputTokens, cost: outputTokensCost },
+			output: { price: outputPrice, tokens: outputTokens, cost: responseTokensCost },
 		}
 
 		// Only include cache read costs in the trace (cache write costs are tracked separately)

src/api/providers/openrouter.ts

@@ -139,7 +139,10 @@ export class OpenRouterHandler implements ApiHandler {
 	}
 
 	getModel(): { id: string; info: ModelInfo } {
-		const modelId = this.options.openRouterModelId
+		let modelId = this.options.openRouterModelId
+		if (modelId === "x-ai/grok-3") {
+			modelId = "x-ai/grok-3-beta"
+		}
 		const modelInfo = this.options.openRouterModelInfo
 		if (modelId && modelInfo) {
 			return { id: modelId, info: modelInfo }

src/api/providers/qwen.ts

@@ -53,20 +53,37 @@ export class QwenHandler implements ApiHandler {
 	async *createMessage(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]): ApiStream {
 		const model = this.getModel()
 		const isDeepseekReasoner = model.id.includes("deepseek-r1")
+		const isReasoningModelFamily = model.id.includes("qwen3") || ["qwen-plus-latest", "qwen-turbo-latest"].includes(model.id)
+
 		let openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
 			{ role: "system", content: systemPrompt },
 			...convertToOpenAiMessages(messages),
 		]
-		if (isDeepseekReasoner) {
+
+		let temperature: number | undefined = 0
+		// Configuration for extended thinking
+		const budgetTokens = this.options.thinkingBudgetTokens || 0
+		const reasoningOn = budgetTokens !== 0 ? true : false
+		const thinkingArgs = isReasoningModelFamily
+			? {
+					enable_thinking: reasoningOn,
+					thinking_budget: reasoningOn ? budgetTokens : undefined,
+				}
+			: undefined
+
+		if (isDeepseekReasoner || (reasoningOn && isReasoningModelFamily)) {
 			openAiMessages = convertToR1Format([{ role: "user", content: systemPrompt }, ...messages])
+			temperature = undefined
 		}
+
 		const stream = await this.client.chat.completions.create({
 			model: model.id,
 			max_completion_tokens: model.info.maxTokens,
 			messages: openAiMessages,
 			stream: true,
 			stream_options: { include_usage: true },
-			...(model.id === "deepseek-r1" ? {} : { temperature: 0 }),
+			temperature,
+			...thinkingArgs,
 		})
 
 		for await (const chunk of stream) {

src/api/providers/sapaicore.ts

@@ -0,0 +1,540 @@
+import { Anthropic } from "@anthropic-ai/sdk"
+import axios from "axios"
+import OpenAI from "openai"
+import { ApiHandler } from "../"
+import { ApiHandlerOptions, ModelInfo, sapAiCoreDefaultModelId, SapAiCoreModelId, sapAiCoreModels } from "../../shared/api"
+import { convertToOpenAiMessages } from "../transform/openai-format"
+import { ApiStream } from "../transform/stream"
+
+interface Deployment {
+	id: string
+	name: string
+}
+interface Token {
+	access_token: string
+	expires_in: number
+	scope: string
+	jti: string
+	token_type: string
+	expires_at: number
+}
+export class SapAiCoreHandler implements ApiHandler {
+	private options: ApiHandlerOptions
+	private token?: Token
+	private deployments?: Deployment[]
+
+	constructor(options: ApiHandlerOptions) {
+		this.options = options
+	}
+
+	private async authenticate(): Promise<Token> {
+		const payload = {
+			grant_type: "client_credentials",
+			client_id: this.options.sapAiCoreClientId || "",
+			client_secret: this.options.sapAiCoreClientSecret || "",
+		}
+
+		const tokenUrl = (this.options.sapAiCoreTokenUrl || "").replace(/\/+$/, "") + "/oauth/token"
+		const response = await axios.post(tokenUrl, payload, {
+			headers: { "Content-Type": "application/x-www-form-urlencoded" },
+		})
+		const token = response.data as Token
+		token.expires_at = Date.now() + token.expires_in * 1000
+		return token
+	}
+
+	private async getToken(): Promise<string> {
+		if (!this.token || this.token.expires_at < Date.now()) {
+			this.token = await this.authenticate()
+		}
+		return this.token.access_token
+	}
+
+	private async getAiCoreDeployments(): Promise<Deployment[]> {
+		if (this.options.sapAiCoreClientSecret === "") {
+			return [{ id: "notconfigured", name: "ai-core-not-configured" }]
+		}
+
+		const token = await this.getToken()
+		const headers = {
+			Authorization: `Bearer ${token}`,
+			"AI-Resource-Group": this.options.sapAiResourceGroup || "default",
+			"Content-Type": "application/json",
+		}
+
+		const url = `${this.options.sapAiCoreBaseUrl}/v2/lm/deployments?$top=10000&$skip=0`
+
+		try {
+			const response = await axios.get(url, { headers })
+			const deployments = response.data.resources
+
+			return deployments
+				.filter((deployment: any) => deployment.targetStatus === "RUNNING")
+				.map((deployment: any) => {
+					const model = deployment.details?.resources?.backend_details?.model
+					if (!model?.name || !model?.version) {
+						return null // Skip this row
+					}
+					return {
+						id: deployment.id,
+						name: `${model.name}:${model.version}`,
+					}
+				})
+				.filter((deployment: any) => deployment !== null)
+		} catch (error) {
+			console.error("Error fetching deployments:", error)
+			throw new Error("Failed to fetch deployments")
+		}
+	}
+
+	private async getDeploymentForModel(modelId: string): Promise<string> {
+		// If deployments are not fetched yet or the model is not found in the fetched deployments, fetch deployments
+		if (!this.deployments || !this.hasDeploymentForModel(modelId)) {
+			this.deployments = await this.getAiCoreDeployments()
+		}
+
+		const deployment = this.deployments.find((d) => {
+			const deploymentBaseName = d.name.split(":")[0].toLowerCase()
+			const modelBaseName = modelId.split(":")[0].toLowerCase()
+			return deploymentBaseName === modelBaseName
+		})
+
+		if (!deployment) {
+			throw new Error(`No running deployment found for model ${modelId}`)
+		}
+
+		return deployment.id
+	}
+
+	private hasDeploymentForModel(modelId: string): boolean {
+		return this.deployments?.some((d) => d.name.split(":")[0].toLowerCase() === modelId.split(":")[0].toLowerCase()) ?? false
+	}
+
+	async *createMessage(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]): ApiStream {
+		const token = await this.getToken()
+		const headers = {
+			Authorization: `Bearer ${token}`,
+			"AI-Resource-Group": this.options.sapAiResourceGroup || "default",
+			"Content-Type": "application/json",
+		}
+
+		const model = this.getModel()
+		const deploymentId = await this.getDeploymentForModel(model.id)
+
+		const anthropicModels = [
+			"anthropic--claude-3.7-sonnet",
+			"anthropic--claude-3.5-sonnet",
+			"anthropic--claude-3-sonnet",
+			"anthropic--claude-3-haiku",
+			"anthropic--claude-3-opus",
+		]
+
+		const openAIModels = ["gpt-4o", "gpt-4", "gpt-4o-mini", "o1", "gpt-4.1", "gpt-4.1-nano", "o3-mini", "o3", "o4-mini"]
+
+		let url: string
+		let payload: any
+		if (anthropicModels.includes(model.id)) {
+			url = `${this.options.sapAiCoreBaseUrl}/v2/inference/deployments/${deploymentId}/invoke-with-response-stream`
+
+			if (model.id === "anthropic--claude-3.7-sonnet") {
+				url = `${this.options.sapAiCoreBaseUrl}/v2/inference/deployments/${deploymentId}/converse-stream`
+				payload = {
+					inferenceConfig: {
+						maxTokens: model.info.maxTokens,
+						temperature: 0.0,
+					},
+					system: systemPrompt ? [{ text: systemPrompt }] : undefined,
+					messages: this.formatAnthropicMessages(messages),
+				}
+			} else {
+				payload = {
+					max_tokens: model.info.maxTokens,
+					system: systemPrompt,
+					messages,
+					anthropic_version: "bedrock-2023-05-31",
+				}
+			}
+		} else if (openAIModels.includes(model.id)) {
+			let openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
+				{ role: "system", content: systemPrompt },
+				...convertToOpenAiMessages(messages),
+			]
+
+			url = `${this.options.sapAiCoreBaseUrl}/v2/inference/deployments/${deploymentId}/chat/completions?api-version=2024-12-01-preview`
+			payload = {
+				stream: true,
+				messages: openAiMessages,
+				max_tokens: model.info.maxTokens,
+				temperature: 0.0,
+				frequency_penalty: 0,
+				presence_penalty: 0,
+				stop: null,
+				stream_options: { include_usage: true },
+			}
+
+			if (["o1", "o3-mini", "o3", "o4-mini"].includes(model.id)) {
+				delete payload.max_tokens
+				delete payload.temperature
+			}
+
+			if (model.id === "o3-mini") {
+				delete payload.stream
+				delete payload.stream_options
+			}
+		} else {
+			throw new Error(`Unsupported model: ${model.id}`)
+		}
+
+		try {
+			const response = await axios.post(url, JSON.stringify(payload, null, 2), {
+				headers,
+				responseType: "stream",
+			})
+
+			if (model.id === "o3-mini") {
+				const response = await axios.post(url, JSON.stringify(payload, null, 2), { headers })
+
+				// Yield the usage information
+				if (response.data.usage) {
+					yield {
+						type: "usage",
+						inputTokens: response.data.usage.prompt_tokens,
+						outputTokens: response.data.usage.completion_tokens,
+					}
+				}
+
+				// Yield the content
+				if (response.data.choices && response.data.choices.length > 0) {
+					yield {
+						type: "text",
+						text: response.data.choices[0].message.content,
+					}
+				}
+
+				// Final usage yield
+				if (response.data.usage) {
+					yield {
+						type: "usage",
+						inputTokens: response.data.usage.prompt_tokens,
+						outputTokens: response.data.usage.completion_tokens,
+					}
+				}
+			} else if (openAIModels.includes(model.id)) {
+				yield* this.streamCompletionGPT(response.data, model)
+			} else if (model.id === "anthropic--claude-3.7-sonnet") {
+				yield* this.streamCompletionSonnet37(response.data, model)
+			} else {
+				yield* this.streamCompletion(response.data, model)
+			}
+		} catch (error) {
+			if (error.response) {
+				// The request was made and the server responded with a status code
+				// that falls out of the range of 2xx
+				console.error("Error status:", error.response.status)
+				console.error("Error data:", error.response.data)
+				console.error("Error headers:", error.response.headers)
+
+				if (error.response.status === 404) {
+					console.error("404 Error reason:", error.response.data)
+					throw new Error(`404 Not Found: ${error.response.data}`)
+				}
+			} else if (error.request) {
+				// The request was made but no response was received
+				console.error("Error request:", error.request)
+				throw new Error("No response received from server")
+			} else {
+				// Something happened in setting up the request that triggered an Error
+				console.error("Error message:", error.message)
+				throw new Error(`Error setting up request: ${error.message}`)
+			}
+
+			throw new Error("Failed to create message")
+		}
+	}
+
+	private async *streamCompletion(
+		stream: any,
+		model: { id: SapAiCoreModelId; info: ModelInfo },
+	): AsyncGenerator<any, void, unknown> {
+		let usage = { input_tokens: 0, output_tokens: 0 }
+
+		try {
+			for await (const chunk of stream) {
+				const lines = chunk.toString().split("\n").filter(Boolean)
+				for (const line of lines) {
+					if (line.startsWith("data: ")) {
+						const jsonData = line.slice(6)
+						try {
+							const data = JSON.parse(jsonData)
+							console.log("Received data:", data)
+							if (data.type === "message_start") {
+								usage.input_tokens = data.message.usage.input_tokens
+								yield {
+									type: "usage",
+									inputTokens: usage.input_tokens,
+									outputTokens: usage.output_tokens,
+								}
+							} else if (data.type === "content_block_start" || data.type === "content_block_delta") {
+								const contentBlock = data.type === "content_block_start" ? data.content_block : data.delta
+
+								if (contentBlock.type === "text" || contentBlock.type === "text_delta") {
+									yield {
+										type: "text",
+										text: contentBlock.text || "",
+									}
+								}
+							} else if (data.type === "message_delta") {
+								if (data.usage) {
+									usage.output_tokens = data.usage.output_tokens
+									yield {
+										type: "usage",
+										inputTokens: 0,
+										outputTokens: data.usage.output_tokens,
+									}
+								}
+							}
+						} catch (error) {
+							console.error("Failed to parse JSON data:", error)
+						}
+					}
+				}
+			}
+		} catch (error) {
+			console.error("Error streaming completion:", error)
+			throw error
+		}
+	}
+
+	private async *streamCompletionSonnet37(
+		stream: any,
+		model: { id: SapAiCoreModelId; info: ModelInfo },
+	): AsyncGenerator<any, void, unknown> {
+		function toStrictJson(str: string): string {
+			// Wrap it in parentheses so JS will treat it as an expression
+			const obj = new Function("return " + str)()
+			return JSON.stringify(obj)
+		}
+
+		let usage = { input_tokens: 0, output_tokens: 0 }
+
+		try {
+			// Iterate over the stream and process each chunk
+			for await (const chunk of stream) {
+				const lines = chunk.toString().split("\n").filter(Boolean)
+
+				for (const line of lines) {
+					if (line.startsWith("data: ")) {
+						const jsonData = line.slice(6)
+
+						try {
+							// Parse the incoming JSON data from the stream
+							const data = JSON.parse(toStrictJson(jsonData))
+							console.log("Received data:", data)
+
+							// Handle metadata (token usage)
+							if (data.metadata?.usage) {
+								const inputTokens = data.metadata.usage.inputTokens || 0
+								const outputTokens = data.metadata.usage.outputTokens || 0
+
+								yield {
+									type: "usage",
+									inputTokens,
+									outputTokens,
+								}
+							}
+
+							// Handle content block delta (text generation)
+							if (data.contentBlockDelta) {
+								if (data.contentBlockDelta?.delta?.text) {
+									yield {
+										type: "text",
+										text: data.contentBlockDelta.delta.text,
+									}
+								}
+
+								// Handle reasoning content if present
+								if (data.contentBlockDelta?.delta?.reasoningContent?.text) {
+									yield {
+										type: "reasoning",
+										reasoning: data.contentBlockDelta.delta.reasoningContent.text,
+									}
+								}
+							}
+						} catch (error) {
+							console.error("Failed to parse JSON data:", error)
+							yield {
+								type: "text",
+								text: `[ERROR] Failed to parse response data: ${error instanceof Error ? error.message : String(error)}`,
+							}
+						}
+					}
+				}
+			}
+		} catch (error) {
+			console.error("Error streaming completion:", error)
+			yield {
+				type: "text",
+				text: `[ERROR] Failed to process stream: ${error instanceof Error ? error.message : String(error)}`,
+			}
+		}
+	}
+
+	private async *streamCompletionGPT(
+		stream: any,
+		model: { id: SapAiCoreModelId; info: ModelInfo },
+	): AsyncGenerator<any, void, unknown> {
+		let currentContent = ""
+		let inputTokens = 0
+		let outputTokens = 0
+
+		try {
+			for await (const chunk of stream) {
+				const lines = chunk.toString().split("\n").filter(Boolean)
+				for (const line of lines) {
+					if (line.trim() === "data: [DONE]") {
+						// End of stream, yield final usage
+						yield {
+							type: "usage",
+							inputTokens,
+							outputTokens,
+						}
+						return
+					}
+
+					if (line.startsWith("data: ")) {
+						const jsonData = line.slice(6)
+						try {
+							const data = JSON.parse(jsonData)
+							console.log("Received GPT data:", data)
+
+							if (data.choices && data.choices.length > 0) {
+								const choice = data.choices[0]
+								if (choice.delta && choice.delta.content) {
+									yield {
+										type: "text",
+										text: choice.delta.content,
+									}
+									currentContent += choice.delta.content
+								}
+							}
+
+							// Handle usage information
+							if (data.usage) {
+								inputTokens = data.usage.prompt_tokens || inputTokens
+								outputTokens = data.usage.completion_tokens || outputTokens
+								yield {
+									type: "usage",
+									inputTokens,
+									outputTokens,
+								}
+							}
+
+							if (data.choices && data.choices[0].finish_reason === "stop") {
+								// Final usage yield, if not already provided
+								if (!data.usage) {
+									yield {
+										type: "usage",
+										inputTokens,
+										outputTokens,
+									}
+								}
+							}
+						} catch (error) {
+							console.error("Failed to parse GPT JSON data:", error)
+						}
+					}
+				}
+			}
+		} catch (error) {
+			console.error("Error streaming GPT completion:", error)
+			throw error
+		}
+	}
+
+	createUserReadableRequest(
+		userContent: Array<
+			Anthropic.TextBlockParam | Anthropic.ImageBlockParam | Anthropic.ToolUseBlockParam | Anthropic.ToolResultBlockParam
+		>,
+	): any {
+		return {
+			model: this.getModel().id,
+			max_tokens: this.getModel().info.maxTokens,
+			system: "(see SYSTEM_PROMPT in src/ClaudeDev.ts)",
+			messages: [{ conversation_history: "..." }, { role: "user", content: userContent }],
+			tools: "(see tools in src/ClaudeDev.ts)",
+			tool_choice: { type: "auto" },
+		}
+	}
+
+	getModel(): { id: SapAiCoreModelId; info: ModelInfo } {
+		const modelId = this.options.apiModelId
+		if (modelId && modelId in sapAiCoreModels) {
+			const id = modelId as SapAiCoreModelId
+			return { id, info: sapAiCoreModels[id] }
+		}
+		return { id: sapAiCoreDefaultModelId, info: sapAiCoreModels[sapAiCoreDefaultModelId] }
+	}
+
+	private getValidImageFormat(mediaType: string): string {
+		const format = mediaType.split("/")[1]?.toLowerCase()
+		const validFormats = ["png", "jpeg", "gif", "webp"]
+
+		if (validFormats.includes(format)) {
+			return format
+		}
+		throw new Error(`Unsupported image format: ${format}`)
+	}
+
+	private formatAnthropicMessages(messages: Anthropic.Messages.MessageParam[]): any[] {
+		return messages.map((m) => {
+			const contentBlocks: any[] = []
+
+			if (typeof m.content === "string") {
+				contentBlocks.push({ text: m.content })
+			} else if (Array.isArray(m.content)) {
+				for (const block of m.content) {
+					if (block.type === "text") {
+						if (!block.text) {
+							throw new Error('Text block is missing the "text" field.')
+						}
+						contentBlocks.push({ text: block.text })
+					} else if (block.type === "image") {
+						if (!block.source) {
+							throw new Error('Image block is missing the "source" field.')
+						}
+
+						const { type, media_type, data } = block.source
+
+						if (!type || !media_type || !data) {
+							throw new Error('Image source must have "type", "media_type", and "data" fields.')
+						}
+
+						if (type !== "base64") {
+							throw new Error(`Unsupported image source type: ${type}. Only "base64" is supported.`)
+						}
+
+						const format = this.getValidImageFormat(media_type)
+
+						contentBlocks.push({
+							image: {
+								format,
+								source: {
+									bytes: data,
+								},
+							},
+						})
+					} else {
+						throw new Error(`Unsupported content block type: ${block.type}`)
+					}
+				}
+			} else {
+				throw new Error("Unsupported content format.")
+			}
+
+			return {
+				role: m.role,
+				content: contentBlocks,
+			}
+		})
+	}
+}

src/api/retry.ts

@@ -57,7 +57,7 @@ export function withRetry(options: RetryOptions = {}) {
 					const handlerInstance = this as any
 					if (handlerInstance.options?.onRetryAttempt) {
 						try {
-							handlerInstance.options.onRetryAttempt(attempt + 1, maxRetries, delay, error)
+							await handlerInstance.options.onRetryAttempt(attempt + 1, maxRetries, delay, error)
 						} catch (e) {
 							console.error("Error in onRetryAttempt callback:", e)
 						}

src/api/transform/stream.ts

@@ -17,5 +17,6 @@ export interface ApiStreamUsageChunk {
 	outputTokens: number
 	cacheWriteTokens?: number
 	cacheReadTokens?: number
+	thoughtsTokenCount?: number // openrouter
 	totalCost?: number // openrouter
 }

src/core/assistant-message/diff.test.ts

@@ -1,9 +1,9 @@
-import { constructNewFileContent as cnfc2 } from "./diff"
+import { constructNewFileContent as cnfc } from "./diff"
 import { describe, it } from "mocha"
 import { expect } from "chai"
 
-async function cnfc(diffContent: string, originalContent: string, isFinal: boolean): Promise<string> {
-	return cnfc2(diffContent, originalContent, isFinal, "v1")
+async function cnfc2(diffContent: string, originalContent: string, isFinal: boolean): Promise<string> {
+	return cnfc(diffContent, originalContent, isFinal, "v2")
 }
 
 describe("constructNewFileContent", () => {
@@ -19,14 +19,34 @@ new content
 			isFinal: true,
 		},
 		{
-			name: "full file replacement",
-			original: "old content",
-			diff: `------- SEARCH
+			name: "malformed search - mixed symbols",
+			original: "line1\nline2\nline3",
+			diff: `<<-- SEARCH
+line2
 =======
-new content
+replaced
 +++++++ REPLACE`,
-			expected: "new content\n",
-			isFinal: true,
+			shouldThrow: true,
+		},
+		{
+			name: "malformed search - insufficient dashes",
+			original: "line1\nline2\nline3",
+			diff: `-- SEARCH
+line2
+=======
+replaced
++++++++ REPLACE`,
+			shouldThrow: true,
+		},
+		{
+			name: "malformed search - missing space",
+			original: "line1\nline2\nline3",
+			diff: `-------SEARCH
+line2
+=======
+replaced
++++++++ REPLACE`,
+			shouldThrow: true,
 		},
 		{
 			name: "exact match replacement",
@@ -139,17 +159,33 @@ replaced
 	]
 	//.filter(({name}) => name === "multiple ordered replacements")
 	//.filter(({name}) => name === "delete then replace")
-	testCases.forEach(({ name, original, diff, expected, isFinal }) => {
+	testCases.forEach(({ name, original, diff, expected, isFinal, shouldThrow }) => {
 		it(`should handle ${name} case correctly`, async () => {
-			const result1 = await cnfc(diff, original, isFinal)
-			const result2 = await cnfc2(diff, original, isFinal)
-			const equal = result1 === result2
-			const equal2 = result1 === expected
-			// Verify both implementations produce same result
-			expect(result1).to.equal(result2)
+			if (shouldThrow) {
+				try {
+					await cnfc(diff, original, isFinal ?? true)
+					expect.fail("Expected an error to be thrown")
+				} catch (err) {
+					expect(err).to.be.an("error")
+				}
 
-			// Verify result matches expected
-			expect(result1).to.equal(expected)
+				try {
+					await cnfc2(diff, original, isFinal ?? true)
+					expect.fail("Expected an error to be thrown")
+				} catch (err) {
+					expect(err).to.be.an("error")
+				}
+			} else {
+				const result1 = await cnfc(diff, original, isFinal ?? true)
+				const result2 = await cnfc2(diff, original, isFinal ?? true)
+				const equal = result1 === result2
+				const equal2 = result1 === expected
+				// Verify both implementations produce same result
+				expect(result1).to.equal(result2)
+
+				// Verify result matches expected
+				expect(result1).to.equal(expected)
+			}
 		})
 	})
 
@@ -175,4 +211,136 @@ replaced
 			expect(err).to.be.an("error")
 		}
 	})
+
+	it("should handle missing final REPLACE marker when isFinal is true", async () => {
+		const original = "line1\nline2\nline3"
+		const diff = `------- SEARCH
+line2
+=======
+replaced`
+		// Note: missing +++++++ REPLACE marker
+
+		const result1 = await cnfc(diff, original, true) // isFinal = true
+
+		// Should still work and replace line2 with "replaced"
+		const expected = "line1\nreplaced\nline3"
+
+		expect(result1).to.equal(expected)
+	})
+
+	it("should handle missing final REPLACE marker with multiple lines of replacement", async () => {
+		const original = "function test() {\n\tconst a = 1;\n\treturn a;\n}"
+		const diff = `------- SEARCH
+	const a = 1;
+	return a;
+=======
+	const a = 42;
+	console.log('updated');
+	return a;`
+		// Note: missing +++++++ REPLACE marker
+
+		const result1 = await cnfc(diff, original, true) // isFinal = true
+		const expected = "function test() {\n\tconst a = 42;\n\tconsole.log('updated');\n\treturn a;\n}"
+
+		expect(result1).to.equal(expected)
+	})
+
+	// 	it("should NOT process incomplete replacement when isFinal is false", async () => {
+	// 		const original = "line1\nline2\nline3"
+	// 		const diff = `------- SEARCH
+	// line2
+	// =======
+	// replaced`
+	// 		// Note: missing +++++++ REPLACE marker AND isFinal = false
+
+	// 		const result1 = await cnfc(diff, original, false) // isFinal = false
+
+	// 		// Should not make any changes since the block is incomplete
+	// 		const expected = "line1\nline2\nline3"
+
+	// 		expect(result1).to.equal(expected)
+	// 	})
+})
+
+// Test cases for out-of-order search/replace blocks
+
+describe("Diff Format Out of Order Cases", () => {
+	it("should handle out-of-order replacements with different positions", async () => {
+		const isFinal = true
+		const original = "first\nsecond\nthird\nfourth\n"
+		const diff = `------- SEARCH
+fourth
+=======
+new fourth
++++++++ REPLACE
+------- SEARCH
+second
+=======
+new second
++++++++ REPLACE`
+		const result1 = await cnfc(diff, original, isFinal)
+		const expectedResult = "first\nnew second\nthird\nnew fourth\n"
+		expect(result1).to.equal(expectedResult)
+	})
+
+	it("should handle multiple out-of-order replacements", async () => {
+		const isFinal = true
+		const original = "one\ntwo\nthree\nfour\nfive\n"
+		const diff = `------- SEARCH
+four
+=======
+fourth
++++++++ REPLACE
+------- SEARCH
+two
+=======
+second
++++++++ REPLACE
+------- SEARCH
+five
+=======
+fifth
++++++++ REPLACE`
+		const result1 = await cnfc(diff, original, isFinal)
+		const expectedResult = "one\nsecond\nthree\nfourth\nfifth\n"
+		expect(result1).to.equal(expectedResult)
+	})
+
+	it("should handle out-of-order replacements with indentation", async () => {
+		const isFinal = true
+		const original = "function test() {\n\tconst a = 1;\n\tconst b = 2;\n\tconst c = 3;\n\n}"
+		const diff = `------- SEARCH
+	const c = 3;
+=======
+	const c = 30;
++++++++ REPLACE
+------- SEARCH
+	const a = 1;
+=======
+	const a = 10;
++++++++ REPLACE`
+		const result1 = await cnfc(diff, original, isFinal)
+		const expectedResult = "function test() {\n\tconst a = 10;\n\tconst b = 2;\n\tconst c = 30;\n\n}"
+		expect(result1).to.equal(expectedResult)
+	})
+
+	it("should handle out-of-order replacements with empty lines", async () => {
+		const isFinal = true
+		const original = "header\n\nbody\n\nfooter\n"
+		const diff = `------- SEARCH
+footer
+=======
+new footer
++++++++ REPLACE
+------- SEARCH
+
+body
+
+=======
+new body content
++++++++ REPLACE`
+		const result1 = await cnfc(diff, original, isFinal)
+		const expectedResult = "header\nnew body content\nnew footer\n"
+		expect(result1).to.equal(expectedResult)
+	})
 })

src/core/assistant-message/diff.ts

@@ -4,6 +4,28 @@ const REPLACE_BLOCK_END = "+++++++ REPLACE"
 
 const SEARCH_BLOCK_CHAR = "-"
 const REPLACE_BLOCK_CHAR = "+"
+const LEGACY_SEARCH_BLOCK_CHAR = "<"
+const LEGACY_REPLACE_BLOCK_CHAR = ">"
+
+// Replace the exact string constants with flexible regex patterns
+const SEARCH_BLOCK_START_REGEX = /^[-]{3,} SEARCH$/
+const SEARCH_BLOCK_END_REGEX = /^[=]{3,}$/
+const REPLACE_BLOCK_END_REGEX = /^[+]{3,} REPLACE$/
+const LEGACY_SEARCH_BLOCK_START_REGEX = /^[<]{3,} SEARCH$/
+const LEGACY_REPLACE_BLOCK_END_REGEX = /^[>]{3,} REPLACE$/
+
+// Helper functions to check if a line matches the flexible patterns
+function isSearchBlockStart(line: string): boolean {
+	return SEARCH_BLOCK_START_REGEX.test(line) || LEGACY_SEARCH_BLOCK_START_REGEX.test(line)
+}
+
+function isSearchBlockEnd(line: string): boolean {
+	return SEARCH_BLOCK_END_REGEX.test(line)
+}
+
+function isReplaceBlockEnd(line: string): boolean {
+	return REPLACE_BLOCK_END_REGEX.test(line) || LEGACY_REPLACE_BLOCK_END_REGEX.test(line)
+}
 
 /**
  * Attempts a line-trimmed fallback match for the given search content in the original content.
@@ -211,7 +233,7 @@ export async function constructNewFileContent(
 	diffContent: string,
 	originalContent: string,
 	isFinal: boolean,
-	version: "v1" | "v2" = "v2",
+	version: "v1" | "v2" = "v1",
 ): Promise<string> {
 	const constructor = constructNewFileContentVersionMapping[version]
 	if (!constructor) {
@@ -228,9 +250,6 @@ const constructNewFileContentVersionMapping: Record<
 	v2: constructNewFileContentV2,
 } as const
 
-/**
- * @deprecated
- */
 async function constructNewFileContentV1(diffContent: string, originalContent: string, isFinal: boolean): Promise<string> {
 	let result = ""
 	let lastProcessedIndex = 0
@@ -243,6 +262,10 @@ async function constructNewFileContentV1(diffContent: string, originalContent: s
 	let searchMatchIndex = -1
 	let searchEndIndex = -1
 
+	// Track all replacements to handle out-of-order edits
+	let replacements: Array<{ start: number; end: number; content: string }> = []
+	let pendingOutOfOrderReplacement = false
+
 	let lines = diffContent.split("\n")
 
 	// If the last line looks like a partial marker but isn't recognized,
@@ -250,23 +273,27 @@ async function constructNewFileContentV1(diffContent: string, originalContent: s
 	const lastLine = lines[lines.length - 1]
 	if (
 		lines.length > 0 &&
-		(lastLine.startsWith(SEARCH_BLOCK_CHAR) || lastLine.startsWith("=") || lastLine.startsWith(REPLACE_BLOCK_CHAR)) &&
-		lastLine !== SEARCH_BLOCK_START &&
-		lastLine !== SEARCH_BLOCK_END &&
-		lastLine !== REPLACE_BLOCK_END
+		(lastLine.startsWith(SEARCH_BLOCK_CHAR) ||
+			lastLine.startsWith(LEGACY_SEARCH_BLOCK_CHAR) ||
+			lastLine.startsWith("=") ||
+			lastLine.startsWith(REPLACE_BLOCK_CHAR) ||
+			lastLine.startsWith(LEGACY_REPLACE_BLOCK_CHAR)) &&
+		!isSearchBlockStart(lastLine) &&
+		!isSearchBlockEnd(lastLine) &&
+		!isReplaceBlockEnd(lastLine)
 	) {
 		lines.pop()
 	}
 
 	for (const line of lines) {
-		if (line === SEARCH_BLOCK_START) {
+		if (isSearchBlockStart(line)) {
 			inSearch = true
 			currentSearchContent = ""
 			currentReplaceContent = ""
 			continue
 		}
 
-		if (line === SEARCH_BLOCK_END) {
+		if (isSearchBlockEnd(line)) {
 			inSearch = false
 			inReplace = true
 
@@ -284,9 +311,12 @@ async function constructNewFileContentV1(diffContent: string, originalContent: s
 					searchMatchIndex = 0
 					searchEndIndex = 0
 				} else {
-					// Complete file replacement scenario: treat the entire file as matched
-					searchMatchIndex = 0
-					searchEndIndex = originalContent.length
+					// ERROR: Empty search block with non-empty file indicates malformed SEARCH marker
+					throw new Error(
+						"Empty SEARCH block detected with non-empty file. This usually indicates a malformed SEARCH marker.\n" +
+							"Please ensure your SEARCH marker follows the correct format:\n" +
+							"- Use '------- SEARCH' (7+ dashes + space + SEARCH)\n",
+					)
 				}
 			} else {
 				// Add check for inefficient full-file search
@@ -314,31 +344,51 @@ async function constructNewFileContentV1(diffContent: string, originalContent: s
 						if (blockMatch) {
 							;[searchMatchIndex, searchEndIndex] = blockMatch
 						} else {
-							throw new Error(
-								`The SEARCH block:\n${currentSearchContent.trimEnd()}\n...does not match anything in the file or was searched out of order in the provided blocks.`,
-							)
+							// Last resort: search the entire file from the beginning
+							const fullFileIndex = originalContent.indexOf(currentSearchContent, 0)
+							if (fullFileIndex !== -1) {
+								// Found in the file - could be out of order
+								searchMatchIndex = fullFileIndex
+								searchEndIndex = fullFileIndex + currentSearchContent.length
+								if (searchMatchIndex < lastProcessedIndex) {
+									pendingOutOfOrderReplacement = true
+								}
+							} else {
+								throw new Error(
+									`The SEARCH block:\n${currentSearchContent.trimEnd()}\n...does not match anything in the file.`,
+								)
+							}
 						}
 					}
 				}
 			}
 
-			// Output everything up to the match location
-			result += originalContent.slice(lastProcessedIndex, searchMatchIndex)
+			// Check if this is an out-of-order replacement
+			if (searchMatchIndex < lastProcessedIndex) {
+				pendingOutOfOrderReplacement = true
+			}
+
+			// For in-order replacements, output everything up to the match location
+			if (!pendingOutOfOrderReplacement) {
+				result += originalContent.slice(lastProcessedIndex, searchMatchIndex)
+			}
 			continue
 		}
 
-		if (line === REPLACE_BLOCK_END) {
+		if (isReplaceBlockEnd(line)) {
 			// Finished one replace block
 
-			// // Remove the artificially added linebreak in the last line of the REPLACE block
-			// if (result.endsWith("\r\n")) {
-			// 	result = result.slice(0, -2)
-			// } else if (result.endsWith("\n")) {
-			// 	result = result.slice(0, -1)
-			// }
+			// Store this replacement
+			replacements.push({
+				start: searchMatchIndex,
+				end: searchEndIndex,
+				content: currentReplaceContent,
+			})
 
-			// Advance lastProcessedIndex to after the matched section
-			lastProcessedIndex = searchEndIndex
+			// If this was an in-order replacement, advance lastProcessedIndex
+			if (!pendingOutOfOrderReplacement) {
+				lastProcessedIndex = searchEndIndex
+			}
 
 			// Reset for next block
 			inSearch = false
@@ -347,6 +397,7 @@ async function constructNewFileContentV1(diffContent: string, originalContent: s
 			currentReplaceContent = ""
 			searchMatchIndex = -1
 			searchEndIndex = -1
+			pendingOutOfOrderReplacement = false
 			continue
 		}
 
@@ -358,16 +409,59 @@ async function constructNewFileContentV1(diffContent: string, originalContent: s
 			currentSearchContent += line + "\n"
 		} else if (inReplace) {
 			currentReplaceContent += line + "\n"
-			// Output replacement lines immediately if we know the insertion point
-			if (searchMatchIndex !== -1) {
+			// Only output replacement lines immediately for in-order replacements
+			if (searchMatchIndex !== -1 && !pendingOutOfOrderReplacement) {
 				result += line + "\n"
 			}
 		}
 	}
 
-	// If this is the final chunk, append any remaining original content
-	if (isFinal && lastProcessedIndex < originalContent.length) {
-		result += originalContent.slice(lastProcessedIndex)
+	// If this is the final chunk, we need to apply all replacements and build the final result
+	if (isFinal) {
+		// Handle the case where we're still in replace mode when processing ends
+		// and this is the final chunk - treat it as if we encountered the REPLACE marker
+		if (inReplace && searchMatchIndex !== -1) {
+			// Store this replacement
+			replacements.push({
+				start: searchMatchIndex,
+				end: searchEndIndex,
+				content: currentReplaceContent,
+			})
+
+			// If this was an in-order replacement, advance lastProcessedIndex
+			if (!pendingOutOfOrderReplacement) {
+				lastProcessedIndex = searchEndIndex
+			}
+
+			// Reset state
+			inSearch = false
+			inReplace = false
+			currentSearchContent = ""
+			currentReplaceContent = ""
+			searchMatchIndex = -1
+			searchEndIndex = -1
+			pendingOutOfOrderReplacement = false
+		}
+		// end of handling missing replace marker
+
+		// Sort replacements by start position
+		replacements.sort((a, b) => a.start - b.start)
+
+		// Rebuild the entire result by applying all replacements
+		result = ""
+		let currentPos = 0
+
+		for (const replacement of replacements) {
+			// Add original content up to this replacement
+			result += originalContent.slice(currentPos, replacement.start)
+			// Add the replacement content
+			result += replacement.content
+			// Move position to after the replaced section
+			currentPos = replacement.end
+		}
+
+		// Add any remaining original content
+		result += originalContent.slice(currentPos)
 	}
 
 	return result
@@ -487,7 +581,7 @@ class NewFileContentConstructor {
 		pendingNonStandardLineLimit: number,
 	): number {
 		let removeLineCount = 0
-		if (line === SEARCH_BLOCK_START) {
+		if (isSearchBlockStart(line)) {
 			removeLineCount = this.trimPendingNonStandardTrailingEmptyLines(pendingNonStandardLineLimit)
 			if (removeLineCount > 0) {
 				pendingNonStandardLineLimit = pendingNonStandardLineLimit - removeLineCount
@@ -497,7 +591,7 @@ class NewFileContentConstructor {
 				canWritependingNonStandardLines && (this.pendingNonStandardLines.length = 0)
 			}
 			this.activateSearchState()
-		} else if (line === SEARCH_BLOCK_END) {
+		} else if (isSearchBlockEnd(line)) {
 			// 校验非标内容
 			if (!this.isSearchingActive()) {
 				this.tryFixSearchBlock(pendingNonStandardLineLimit)
@@ -505,7 +599,7 @@ class NewFileContentConstructor {
 			}
 			this.activateReplaceState()
 			this.beforeReplace()
-		} else if (line === REPLACE_BLOCK_END) {
+		} else if (isReplaceBlockEnd(line)) {
 			if (!this.isReplacingActive()) {
 				this.tryFixReplaceBlock(pendingNonStandardLineLimit)
 				canWritependingNonStandardLines && (this.pendingNonStandardLines.length = 0)
@@ -528,7 +622,6 @@ class NewFileContentConstructor {
 			} else {
 				let appendToPendingNonStandardLines = canWritependingNonStandardLines
 				if (appendToPendingNonStandardLines) {
-					console.log("unstandard line:" + line)
 					// 处理非标内容
 					this.pendingNonStandardLines.push(line)
 				}
@@ -613,7 +706,7 @@ class NewFileContentConstructor {
 		if (!lineLimit) {
 			throw new Error("Invalid SEARCH/REPLACE block structure - no lines available to process")
 		}
-		let searchTagRegexp = /^[-]{3,} SEARCH$/
+		let searchTagRegexp = /^([-]{3,}|[<]{3,}) SEARCH$/
 		const searchTagIndex = this.findLastMatchingLineIndex(searchTagRegexp, lineLimit)
 		if (searchTagIndex !== -1) {
 			let fixLines = this.pendingNonStandardLines.slice(searchTagIndex, lineLimit)
@@ -664,7 +757,7 @@ class NewFileContentConstructor {
 			throw new Error()
 		}
 
-		let replaceEndTagRegexp = /^[+]{3,} REPLACE$/
+		let replaceEndTagRegexp = /^([+]{3,}|[>]{3,}) REPLACE$/
 		const replaceEndTagIndex = this.findLastMatchingLineIndex(replaceEndTagRegexp, lineLimit)
 		const likeReplaceEndTag = replaceEndTagIndex === lineLimit - 1
 		if (likeReplaceEndTag) {
@@ -713,7 +806,11 @@ export async function constructNewFileContentV2(diffContent: string, originalCon
 	const lastLine = lines[lines.length - 1]
 	if (
 		lines.length > 0 &&
-		(lastLine.startsWith(SEARCH_BLOCK_CHAR) || lastLine.startsWith("=") || lastLine.startsWith(REPLACE_BLOCK_CHAR)) &&
+		(lastLine.startsWith(SEARCH_BLOCK_CHAR) ||
+			lastLine.startsWith(LEGACY_SEARCH_BLOCK_CHAR) ||
+			lastLine.startsWith("=") ||
+			lastLine.startsWith(REPLACE_BLOCK_CHAR) ||
+			lastLine.startsWith(LEGACY_REPLACE_BLOCK_CHAR)) &&
 		lastLine !== SEARCH_BLOCK_START &&
 		lastLine !== SEARCH_BLOCK_END &&
 		lastLine !== REPLACE_BLOCK_END

src/core/assistant-message/diff_edge_cases.test.ts

@@ -1,9 +1,9 @@
-import { constructNewFileContent as cnfc2 } from "./diff"
+import { constructNewFileContent as cnfc } from "./diff"
 import { describe, it } from "mocha"
 import { expect } from "chai"
 
-async function cnfc(diffContent: string, originalContent: string, isFinal: boolean): Promise<string> {
-	return cnfc2(diffContent, originalContent, isFinal, "v1")
+async function cnfc2(diffContent: string, originalContent: string, isFinal: boolean): Promise<string> {
+	return cnfc(diffContent, originalContent, isFinal, "v2")
 }
 
 describe("Diff Format Edge Cases", () => {
@@ -17,8 +17,9 @@ new content
 +++++++ REPLACE`
 		const result1 = await cnfc(diff, original, isFinal)
 		const result2 = await cnfc2(diff, original, isFinal)
-		expect(result1).to.equal("new content\n")
-		expect(result2).to.equal("before\nnew content\nafter")
+		const expectedResult = "before\nnew content\nafter"
+		expect(result1).to.equal(expectedResult)
+		expect(result2).to.equal(expectedResult)
 	})
 
 	it("should handle SEARCH prefix symbols - more than 7", async () => {
@@ -31,8 +32,9 @@ new content
 +++++++ REPLACE`
 		const result1 = await cnfc(diff, original, isFinal)
 		const result2 = await cnfc2(diff, original, isFinal)
-		expect(result1).to.equal("new content\n")
-		expect(result2).to.equal("before\nnew content\nafter")
+		const expectedResult = "before\nnew content\nafter"
+		expect(result1).to.equal(expectedResult)
+		expect(result2).to.equal(expectedResult)
 	})
 
 	it("should handle SEARCH - less than 7 and REPLACE = less than 7", async () => {
@@ -45,8 +47,9 @@ new content
 +++++++ REPLACE`
 		const result1 = await cnfc(diff, original, isFinal)
 		const result2 = await cnfc2(diff, original, isFinal)
-		expect(result1).to.equal("r")
-		expect(result2).to.equal("before\nnew content\nafter")
+		const expectedResult = "before\nnew content\nafter"
+		expect(result1).to.equal(expectedResult)
+		expect(result2).to.equal(expectedResult)
 	})
 
 	it("should handle SEARCH - less than 7 and REPLACE = more than 7", async () => {
@@ -59,7 +62,7 @@ new content
 +++++++ REPLACE`
 		const result1 = await cnfc(diff, original, isFinal)
 		const result2 = await cnfc2(diff, original, isFinal)
-		expect(result1).to.equal("r")
+		expect(result1).to.equal("before\nnew content\nafter")
 		expect(result2).to.equal("before\nnew content\nafter")
 	})
 
@@ -73,8 +76,9 @@ new content
 +++++++ REPLACE`
 		const result1 = await cnfc(diff, original, isFinal)
 		const result2 = await cnfc2(diff, original, isFinal)
-		expect(result1).to.equal("r")
-		expect(result2).to.equal("before\nnew content\nafter")
+		const expectedResult = "before\nnew content\nafter"
+		expect(result1).to.equal(expectedResult)
+		expect(result2).to.equal(expectedResult)
 	})
 
 	it("should handle SEARCH - more than 7 and REPLACE = less than 7", async () => {
@@ -87,8 +91,9 @@ new content
 +++++++ REPLACE`
 		const result1 = await cnfc(diff, original, isFinal)
 		const result2 = await cnfc2(diff, original, isFinal)
-		expect(result1).to.equal("r")
-		expect(result2).to.equal("before\nnew content\nafter")
+		const expectedResult = "before\nnew content\nafter"
+		expect(result1).to.equal(expectedResult)
+		expect(result2).to.equal(expectedResult)
 	})
 
 	it("should handle consecutive SEARCH-REPLACE with second block SEARCH - less than 7", async () => {
@@ -106,8 +111,9 @@ second new content
 +++++++ REPLACE`
 		const result1 = await cnfc(diff, original, isFinal)
 		const result2 = await cnfc2(diff, original, isFinal)
-		expect(result1).to.equal("before\nfirst new content\nsecond new content\n")
-		expect(result2).to.equal("before\nfirst new content\nafter\nsecond new content\nend")
+		const expectedResult = "before\nfirst new content\nafter\nsecond new content\nend"
+		expect(result1).to.equal(expectedResult)
+		expect(result2).to.equal(expectedResult)
 	})
 
 	it("should handle consecutive SEARCH-REPLACE with second block SEARCH - less than 7 and REPLACE = less than 7", async () => {
@@ -125,7 +131,8 @@ second new content
 +++++++ REPLACE`
 		const result1 = await cnfc(diff, original, isFinal)
 		const result2 = await cnfc2(diff, original, isFinal)
-		expect(result1).to.equal("before\nfirst new content\nd")
-		expect(result2).to.equal("before\nfirst new content\nafter\nsecond new content\nend")
+		const expectedResult = "before\nfirst new content\nafter\nsecond new content\nend"
+		expect(result1).to.equal(expectedResult)
+		expect(result2).to.equal(expectedResult)
 	})
 })

src/core/assistant-message/diff_edge_cases2.test.ts

@@ -1,361 +1,361 @@
-import { constructNewFileContent as cnfc2 } from "./diff"
-import { describe, it } from "mocha"
-import { expect } from "chai"
+// import { constructNewFileContent as cnfc } from "./diff"
+// import { describe, it } from "mocha"
+// import { expect } from "chai"
 
-async function cnfc(diffContent: string, originalContent: string, isFinal: boolean): Promise<string> {
-	return cnfc2(diffContent, originalContent, isFinal, "v1")
-}
+// async function cnfc2(diffContent: string, originalContent: string, isFinal: boolean): Promise<string> {
+// 	return cnfc(diffContent, originalContent, isFinal, "v2")
+// }
 
-describe("Diff Format Edge Cases", () => {
-	it("should handle missing search block", async () => {
-		const original = "line1\nline2"
-		const diff = `=======
-new content
-+++++++ REPLACE`
-		const result1 = await cnfc(diff, original, true)
-		expect(result1).to.equal("new content\n")
-		try {
-			await cnfc2(diff, original, true)
-			expect.fail("Expected an error to be thrown")
-		} catch (err) {
-			expect(err).to.be.an("error")
-		}
-	})
+// describe("Diff Format Edge Cases", () => {
+// 	it("should handle missing search block", async () => {
+// 		const original = "line1\nline2"
+// 		const diff = `=======
+// new content
+// +++++++ REPLACE`
+// 		const result1 = await cnfc(diff, original, true)
+// 		expect(result1).to.equal("new content\n")
+// 		try {
+// 			await cnfc2(diff, original, true)
+// 			expect.fail("Expected an error to be thrown")
+// 		} catch (err) {
+// 			expect(err).to.be.an("error")
+// 		}
+// 	})
 
-	it("should handle consecutive search blocks", async () => {
-		const original = "text"
-		const diff = `------- SEARCH
-=======
-replaced
-+++++++ REPLACE
-------- SEARCH
-=======
-another
-+++++++ REPLACE`
-		const result1 = await cnfc(diff, original, true)
-		expect(result1).to.equal("replaced\nanother\n")
-		try {
-			await cnfc2(diff, original, true)
-			expect.fail("Expected an error to be thrown")
-		} catch (err) {
-			expect(err).to.be.an("error")
-		}
-	})
+// 	it("should handle consecutive search blocks", async () => {
+// 		const original = "text"
+// 		const diff = `------- SEARCH
+// =======
+// replaced
+// +++++++ REPLACE
+// ------- SEARCH
+// =======
+// another
+// +++++++ REPLACE`
+// 		const result1 = await cnfc(diff, original, true)
+// 		expect(result1).to.equal("replaced\nanother\n")
+// 		try {
+// 			await cnfc2(diff, original, true)
+// 			expect.fail("Expected an error to be thrown")
+// 		} catch (err) {
+// 			expect(err).to.be.an("error")
+// 		}
+// 	})
 
-	it("should handle reverse markers order", async () => {
-		const original = "content"
-		const diff = `+++++++ SEARCH
-=======
-invalid
-------- REPLACE`
-		const result1 = await cnfc(diff, original, true)
-		expect(result1).to.equal("invalid\ncontent")
-		try {
-			await cnfc2(diff, original, true)
-			expect.fail("Expected an error to be thrown")
-		} catch (err) {
-			expect(err).to.be.an("error")
-		}
-	})
+// 	it("should handle reverse markers order", async () => {
+// 		const original = "content"
+// 		const diff = `+++++++ SEARCH
+// =======
+// invalid
+// ------- REPLACE`
+// 		const result1 = await cnfc(diff, original, true)
+// 		expect(result1).to.equal("invalid\ncontent")
+// 		try {
+// 			await cnfc2(diff, original, true)
+// 			expect.fail("Expected an error to be thrown")
+// 		} catch (err) {
+// 			expect(err).to.be.an("error")
+// 		}
+// 	})
 
-	it("should handle incomplete block structure", async () => {
-		const original = "valid text"
-		const diff = `------- SEARCH
-text
-+++++++ REPLACE`
-		const result1 = await cnfc(diff, original, true)
-		expect(result1).to.equal("t")
-		try {
-			await cnfc2(diff, original, true)
-			expect.fail("Expected an error to be thrown")
-		} catch (err) {
-			expect(err).to.be.an("error")
-		}
-	})
+// 	it("should handle incomplete block structure", async () => {
+// 		const original = "valid text"
+// 		const diff = `------- SEARCH
+// text
+// +++++++ REPLACE`
+// 		const result1 = await cnfc(diff, original, true)
+// 		expect(result1).to.equal("t")
+// 		try {
+// 			await cnfc2(diff, original, true)
+// 			expect.fail("Expected an error to be thrown")
+// 		} catch (err) {
+// 			expect(err).to.be.an("error")
+// 		}
+// 	})
 
-	it("should handle empty search block", async () => {
-		const original = "any content"
-		const diff = `------- SEARCH
-=======
-inserted
-+++++++ REPLACE`
-		const result1 = await cnfc(diff, original, true)
-		const result2 = await cnfc2(diff, original, true)
-		expect(result1).to.equal("inserted\n")
-		expect(result1).to.equal(result2)
-	})
+// 	it("should handle empty search block", async () => {
+// 		const original = "any content"
+// 		const diff = `------- SEARCH
+// =======
+// inserted
+// +++++++ REPLACE`
+// 		const result1 = await cnfc(diff, original, true)
+// 		const result2 = await cnfc2(diff, original, true)
+// 		expect(result1).to.equal("inserted\n")
+// 		expect(result1).to.equal(result2)
+// 	})
 
-	it("should handle mixed line endings", async () => {
-		const original = "line1\r\nline2"
-		const diff = `------- SEARCH
-line1\r
-=======
-line1
-+++++++ REPLACE`
-		const result1 = await cnfc(diff, original, true)
-		const result2 = await cnfc2(diff, original, true)
-		expect(result1).to.equal("line1\nline2")
-		expect(result1).to.equal(result2)
-	})
+// 	it("should handle mixed line endings", async () => {
+// 		const original = "line1\r\nline2"
+// 		const diff = `------- SEARCH
+// line1\r
+// =======
+// line1
+// +++++++ REPLACE`
+// 		const result1 = await cnfc(diff, original, true)
+// 		const result2 = await cnfc2(diff, original, true)
+// 		expect(result1).to.equal("line1\nline2")
+// 		expect(result1).to.equal(result2)
+// 	})
 
-	it("should handle special characters in search", async () => {
-		const original = "text with $^.*\nend"
-		const diff = `------- SEARCH
-$^.*
-=======
-replaced
-+++++++ REPLACE`
-		const result1 = await cnfc(diff, original, true)
-		const result2 = await cnfc2(diff, original, true)
-		expect(result1).to.equal("text with replaced\nend")
-		expect(result1).to.equal(result2)
-	})
+// 	it("should handle special characters in search", async () => {
+// 		const original = "text with $^.*\nend"
+// 		const diff = `------- SEARCH
+// $^.*
+// =======
+// replaced
+// +++++++ REPLACE`
+// 		const result1 = await cnfc(diff, original, true)
+// 		const result2 = await cnfc2(diff, original, true)
+// 		expect(result1).to.equal("text with replaced\nend")
+// 		expect(result1).to.equal(result2)
+// 	})
 
-	it("should handle special regex chars and nested search markers", async () => {
-		const original = `text with $^.*\n--- SEARCH\nend`
-		const diff = `------- SEARCH
-$^.*
-=======
-replaced
-+++++++ REPLACE
+// 	it("should handle special regex chars and nested search markers", async () => {
+// 		const original = `text with $^.*\n--- SEARCH\nend`
+// 		const diff = `------- SEARCH
+// $^.*
+// =======
+// replaced
+// +++++++ REPLACE
 
-------- SEARCH
---- SEARCH
-=======
-before
-+++++++ REPLACE`
-		const result1 = await cnfc(diff, original, true)
-		const result2 = await cnfc2(diff, original, true)
-		expect(result1).to.equal("text with replaced\nbefore\nend")
-		expect(result1).to.equal(result2)
-	})
+// ------- SEARCH
+// --- SEARCH
+// =======
+// before
+// +++++++ REPLACE`
+// 		const result1 = await cnfc(diff, original, true)
+// 		const result2 = await cnfc2(diff, original, true)
+// 		expect(result1).to.equal("text with replaced\nbefore\nend")
+// 		expect(result1).to.equal(result2)
+// 	})
 
-	it("cnfc2 should handle invalid search marker format", async () => {
-		const original = `text with $^.*\n--- SEARCH\nend`
-		const diff = `--- SEARCH
-$^.*
-=======
-replaced
-+++++++ REPLACE
+// 	it("cnfc2 should handle invalid search marker format", async () => {
+// 		const original = `text with $^.*\n--- SEARCH\nend`
+// 		const diff = `--- SEARCH
+// $^.*
+// =======
+// replaced
+// +++++++ REPLACE
 
-------- SEARCH
---- SEARCH
-=======
-before
-+++++++ REPLACE`
-		try {
-			await cnfc(diff, original, true)
-			expect.fail("Expected an error to be thrown")
-		} catch (err) {
-			expect(err).to.be.an("error")
-		}
-		const result2 = await cnfc2(diff, original, true)
-		expect(result2).to.equal("text with replaced\nbefore\nend")
-	})
+// ------- SEARCH
+// --- SEARCH
+// =======
+// before
+// +++++++ REPLACE`
+// 		try {
+// 			await cnfc(diff, original, true)
+// 			expect.fail("Expected an error to be thrown")
+// 		} catch (err) {
+// 			expect(err).to.be.an("error")
+// 		}
+// 		const result2 = await cnfc2(diff, original, true)
+// 		expect(result2).to.equal("text with replaced\nbefore\nend")
+// 	})
 
-	it("cnfc2 should throw error for incomplete search marker", async () => {
-		const original = `text with $^.*\n--- SEARCH\nend`
-		const diff = `--- SEARCH
-$^.*
-=======
-replaced
-+++++++ REPLACE
+// 	it("cnfc2 should throw error for incomplete search marker", async () => {
+// 		const original = `text with $^.*\n--- SEARCH\nend`
+// 		const diff = `--- SEARCH
+// $^.*
+// =======
+// replaced
+// +++++++ REPLACE
 
------- SEARCH
---- SEARCH
-=======
-before
-+++++++ REPLACE`
-		const result1 = await cnfc(diff, original, true)
-		expect(result1).to.equal("replaced\nbefore\n")
-		try {
-			await cnfc2(diff, original, true)
-			expect.fail("Expected an error to be thrown")
-		} catch (err) {
-			expect(err).to.be.an("error")
-		}
-	})
+// ------ SEARCH
+// --- SEARCH
+// =======
+// before
+// +++++++ REPLACE`
+// 		const result1 = await cnfc(diff, original, true)
+// 		expect(result1).to.equal("replaced\nbefore\n")
+// 		try {
+// 			await cnfc2(diff, original, true)
+// 			expect.fail("Expected an error to be thrown")
+// 		} catch (err) {
+// 			expect(err).to.be.an("error")
+// 		}
+// 	})
 
-	it("cnfc2 should handle custom nested search markers", async () => {
-		const original = `text with $^.*\n--- SEARCH2\nend`
-		const diff = `--- SEARCH
-$^.*
-=======
-replaced
-+++++++ REPLACE
+// 	it("cnfc2 should handle custom nested search markers", async () => {
+// 		const original = `text with $^.*\n--- SEARCH2\nend`
+// 		const diff = `--- SEARCH
+// $^.*
+// =======
+// replaced
+// +++++++ REPLACE
 
------- SEARCH
---- SEARCH2
-=======
-before
-+++++++ REPLACE`
-		const result1 = await cnfc(diff, original, true)
-		const result2 = await cnfc2(diff, original, true)
-		expect(result1).to.equal("replaced\nbefore\n")
-		expect(result2).to.equal("text with replaced\nbefore\nend")
-	})
+// ------ SEARCH
+// --- SEARCH2
+// =======
+// before
+// +++++++ REPLACE`
+// 		const result1 = await cnfc(diff, original, true)
+// 		const result2 = await cnfc2(diff, original, true)
+// 		expect(result1).to.equal("replaced\nbefore\n")
+// 		expect(result2).to.equal("text with replaced\nbefore\nend")
+// 	})
 
-	it("cnfc2 should handle text containing nested search markers", async () => {
-		const original = `text with $^.*\ntext with --- SEARCH2\nend`
-		const diff = `--- SEARCH
-$^.*
-=======
-replaced
-+++++++ REPLACE
+// 	it("cnfc2 should handle text containing nested search markers", async () => {
+// 		const original = `text with $^.*\ntext with --- SEARCH2\nend`
+// 		const diff = `--- SEARCH
+// $^.*
+// =======
+// replaced
+// +++++++ REPLACE
 
------- SEARCH
-text with --- SEARCH2
-=======
-before
-+++++++ REPLACE`
-		const result1 = await cnfc(diff, original, true)
-		const result2 = await cnfc2(diff, original, true)
-		expect(result1).to.equal("replaced\nbefore\n")
-		expect(result2).to.equal("text with replaced\nbefore\nend")
-	})
+// ------ SEARCH
+// text with --- SEARCH2
+// =======
+// before
+// +++++++ REPLACE`
+// 		const result1 = await cnfc(diff, original, true)
+// 		const result2 = await cnfc2(diff, original, true)
+// 		expect(result1).to.equal("replaced\nbefore\n")
+// 		expect(result2).to.equal("text with replaced\nbefore\nend")
+// 	})
 
-	it("cnfc2 should handle missing replacement marker in lenient mode", async () => {
-		const original = `text with $^.*\ntext with --- SEARCH2\nend`
-		const diff = `--- SEARCH
-$^.*
-=======
-replaced
-+++++++ REPLACE
+// 	it("cnfc2 should handle missing replacement marker in lenient mode", async () => {
+// 		const original = `text with $^.*\ntext with --- SEARCH2\nend`
+// 		const diff = `--- SEARCH
+// $^.*
+// =======
+// replaced
+// +++++++ REPLACE
 
------- SEARCH
-text with --- SEARCH2
-=======
-before`
-		const result1 = await cnfc(diff, original, false)
-		const result2 = await cnfc2(diff, original, false)
-		expect(result1).to.equal("replaced\nbefore\n")
-		expect(result2).to.equal("text with replaced\nbefore\n")
-	})
+// ------ SEARCH
+// text with --- SEARCH2
+// =======
+// before`
+// 		const result1 = await cnfc(diff, original, false)
+// 		const result2 = await cnfc2(diff, original, false)
+// 		expect(result1).to.equal("replaced\nbefore\n")
+// 		expect(result2).to.equal("text with replaced\nbefore\n")
+// 	})
 
-	it("cnfc2 should throw error for missing replacement marker in strict mode", async () => {
-		const original = `text with $^.*\ntext with --- SEARCH2\nend`
-		const diff = `--- SEARCH
-$^.*
-=======
-replaced
-+++++++ REPLACE
+// 	it("cnfc2 should throw error for missing replacement marker in strict mode", async () => {
+// 		const original = `text with $^.*\ntext with --- SEARCH2\nend`
+// 		const diff = `--- SEARCH
+// $^.*
+// =======
+// replaced
+// +++++++ REPLACE
 
------- SEARCH
-text with --- SEARCH2
-=======
-before`
-		const result1 = await cnfc(diff, original, true)
-		expect(result1).to.equal("replaced\nbefore\n")
-		try {
-			await cnfc2(diff, original, true)
-			expect.fail("Expected an error to be thrown")
-		} catch (err) {
-			expect(err).to.be.an("error")
-		}
-	})
+// ------ SEARCH
+// text with --- SEARCH2
+// =======
+// before`
+// 		const result1 = await cnfc(diff, original, true)
+// 		expect(result1).to.equal("replaced\nbefore\n")
+// 		try {
+// 			await cnfc2(diff, original, true)
+// 			expect.fail("Expected an error to be thrown")
+// 		} catch (err) {
+// 			expect(err).to.be.an("error")
+// 		}
+// 	})
 
-	it("cnfc2 should handle long text with multiple search-replace blocks", async () => {
-		const original = `This is a long text with multiple sections.
-Section 1: Lorem ipsum dolor sit amet
-Section 2: consectetur adipiscing elit
-Section 3: sed do eiusmod tempor
-Section 4: incididunt ut labore
-Section 5: et dolore magna aliqua`
+// 	it("cnfc2 should handle long text with multiple search-replace blocks", async () => {
+// 		const original = `This is a long text with multiple sections.
+// Section 1: Lorem ipsum dolor sit amet
+// Section 2: consectetur adipiscing elit
+// Section 3: sed do eiusmod tempor
+// Section 4: incididunt ut labore
+// Section 5: et dolore magna aliqua`
 
-		const diff = `--- SEARCH
-Section 1: Lorem ipsum dolor sit amet
-=======
-Section 1: Replaced text
-+++++++ REPLACE
+// 		const diff = `--- SEARCH
+// Section 1: Lorem ipsum dolor sit amet
+// =======
+// Section 1: Replaced text
+// +++++++ REPLACE
 
-------- SEARCH
-Section 3: sed do eiusmod tempor
-=======
-Section 3: Modified content
-+++++++ REPLACE
+// ------- SEARCH
+// Section 3: sed do eiusmod tempor
+// =======
+// Section 3: Modified content
+// +++++++ REPLACE
 
-------- SEARCH
-Section 5: et dolore magna aliqua
-=======
-Section 5: Final replacement
-+++++++ REPLACE`
+// ------- SEARCH
+// Section 5: et dolore magna aliqua
+// =======
+// Section 5: Final replacement
+// +++++++ REPLACE`
 
-		const expected = `This is a long text with multiple sections.
-Section 1: Replaced text
-Section 2: consectetur adipiscing elit
-Section 3: Modified content
-Section 4: incididunt ut labore
-Section 5: Final replacement
-`
+// 		const expected = `This is a long text with multiple sections.
+// Section 1: Replaced text
+// Section 2: consectetur adipiscing elit
+// Section 3: Modified content
+// Section 4: incididunt ut labore
+// Section 5: Final replacement
+// `
 
-		const result = await cnfc2(diff, original, true)
-		expect(result).to.equal(expected)
-	})
+// 		const result = await cnfc2(diff, original, true)
+// 		expect(result).to.equal(expected)
+// 	})
 
-	// Test diff containing special regex characters and nested search markers
-	const diff = `--- SEARCH
-$^.*
-=======
-replaced
-+++++++ REPLACE
+// 	// Test diff containing special regex characters and nested search markers
+// 	const diff = `--- SEARCH
+// $^.*
+// =======
+// replaced
+// +++++++ REPLACE
 
------- SEARCH
---- SEARCH
-=======
-before
-+++++++ REPLACE`
-	// expected1 shows the incremental results when processing the diff line by line
-	// Each element represents the result after processing that line number
-	const expected1 = [
-		"",
-		"",
-		"",
-		"replaced\n",
-		"replaced\n",
-		"replaced\n",
-		"replaced\n",
-		"replaced\n",
-		"replaced\n",
-		"replaced\nbefore\n",
-	]
-	// expected2 shows the results when processing with original content
-	// Each element represents the result after processing that line number
-	const expected2 = [
-		"",
-		"",
-		"text with ",
-		"text with replaced\n",
-		"text with replaced\n",
-		"text with replaced\n",
-		"text with replaced\n",
-		"text with replaced\n",
-		new Error(),
-		new Error(),
-	]
-	const diffLines = diff.split("\n")
-	for (let i = 1; i < diffLines.length; i++) {
-		it(`cnfc2 should handle partial diff configuration (line ${i})`, async () => {
-			const original = `text with $^.*\n--- SEARCH\nend`
-			const result1 = await cnfc(diffLines.slice(0, i).join("\n"), original, i === diffLines.length - 1)
-			expect(result1).to.equal(expected1[i - 1])
-		})
-	}
+// ------ SEARCH
+// --- SEARCH
+// =======
+// before
+// +++++++ REPLACE`
+// 	// expected1 shows the incremental results when processing the diff line by line
+// 	// Each element represents the result after processing that line number
+// 	const expected1 = [
+// 		"",
+// 		"",
+// 		"",
+// 		"replaced\n",
+// 		"replaced\n",
+// 		"replaced\n",
+// 		"replaced\n",
+// 		"replaced\n",
+// 		"replaced\n",
+// 		"replaced\nbefore\n",
+// 	]
+// 	// expected2 shows the results when processing with original content
+// 	// Each element represents the result after processing that line number
+// 	const expected2 = [
+// 		"",
+// 		"",
+// 		"text with ",
+// 		"text with replaced\n",
+// 		"text with replaced\n",
+// 		"text with replaced\n",
+// 		"text with replaced\n",
+// 		"text with replaced\n",
+// 		new Error(),
+// 		new Error(),
+// 	]
+// 	const diffLines = diff.split("\n")
+// 	for (let i = 1; i < diffLines.length; i++) {
+// 		it(`cnfc2 should handle partial diff configuration (line ${i})`, async () => {
+// 			const original = `text with $^.*\n--- SEARCH\nend`
+// 			const result1 = await cnfc(diffLines.slice(0, i).join("\n"), original, i === diffLines.length - 1)
+// 			expect(result1).to.equal(expected1[i - 1])
+// 		})
+// 	}
 
-	for (let i = 1; i < diffLines.length; i++) {
-		it(`cnfc2 should handle partial diff configuration (line ${i})`, async () => {
-			const original = `text with $^.*\n--- SEARCH\nend`
-			let expected = expected2[i - 1]
-			if (expected instanceof Error) {
-				try {
-					await cnfc2(diffLines.slice(0, i).join("\n"), original, true)
-					expect.fail("Expected an error to be thrown")
-				} catch (err) {
-					expect(err).to.be.an("error")
-				}
-			} else {
-				const result2 = await cnfc2(diffLines.slice(0, i).join("\n"), original, i === diffLines.length - 1)
-				expect(result2).to.equal(expected)
-			}
-		})
-	}
-})
+// 	for (let i = 1; i < diffLines.length; i++) {
+// 		it(`cnfc2 should handle partial diff configuration (line ${i})`, async () => {
+// 			const original = `text with $^.*\n--- SEARCH\nend`
+// 			let expected = expected2[i - 1]
+// 			if (expected instanceof Error) {
+// 				try {
+// 					await cnfc2(diffLines.slice(0, i).join("\n"), original, true)
+// 					expect.fail("Expected an error to be thrown")
+// 				} catch (err) {
+// 					expect(err).to.be.an("error")
+// 				}
+// 			} else {
+// 				const result2 = await cnfc2(diffLines.slice(0, i).join("\n"), original, i === diffLines.length - 1)
+// 				expect(result2).to.equal(expected)
+// 			}
+// 		})
+// 	}
+// })

src/core/context/context-management/context-error-handling.ts

@@ -8,10 +8,7 @@ export function checkIsOpenRouterContextWindowError(error: any): boolean {
 
 export function checkIsAnthropicContextWindowError(response: any): boolean {
 	try {
-		return (
-			response?.error?.error?.type === "invalid_request_error" &&
-			response?.error?.error?.message?.includes("prompt is too long")
-		)
+		return response?.error?.error?.type === "invalid_request_error"
 	} catch (e: unknown) {
 		return false
 	}

src/core/context/context-tracking/FileContextTracker.ts

@@ -1,7 +1,10 @@
 import * as path from "path"
 import * as vscode from "vscode"
 import { getTaskMetadata, saveTaskMetadata } from "@core/storage/disk"
+import { getWorkspaceState, updateWorkspaceState } from "@core/storage/state"
+import { getGlobalState } from "@core/storage/state"
 import type { FileMetadataEntry } from "./ContextTrackerTypes"
+import type { ClineMessage } from "@shared/ExtensionMessage"
 
 // This class is responsible for tracking file operations that may result in stale context.
 // If a user modifies a file outside of Cline, the context may become stale and need to be updated.
@@ -11,10 +14,12 @@ import type { FileMetadataEntry } from "./ContextTrackerTypes"
 // a diff edit because the file was modified since Cline last read it.
 
 // FileContextTracker
-//
-// This class is responsible for tracking file operations.
-// If the full contents of a file are pass to Cline via a tool, mention, or edit, the file is marked as active.
-// If a file is modified outside of Cline, we detect and track this change to prevent stale context.
+/**
+This class is responsible for tracking file operations.
+If the full contents of a file are passed to Cline via a tool, mention, or edit, the file is marked as active.
+If a file is modified outside of Cline, we detect and track this change to prevent stale context.
+This is used when restoring a task (non-git "checkpoint" restore), and mid-task.
+*/
 export class FileContextTracker {
 	private context: vscode.ExtensionContext
 	readonly taskId: string
@@ -29,7 +34,9 @@ export class FileContextTracker {
 		this.taskId = taskId
 	}
 
-	// Gets the current working directory or returns undefined if it cannot be determined
+	/**
+	 * Gets the current working directory or returns undefined if it cannot be determined
+	 */
 	private getCwd(): string | undefined {
 		const cwd = vscode.workspace.workspaceFolders?.map((folder) => folder.uri.fsPath).at(0)
 		if (!cwd) {
@@ -38,7 +45,9 @@ export class FileContextTracker {
 		return cwd
 	}
 
-	// File watchers are set up for each file that is tracked in the task metadata.
+	/**
+	 * File watchers are set up for each file that is tracked in the task metadata.
+	 */
 	async setupFileWatcher(filePath: string) {
 		// Only setup watcher if it doesn't already exist for this file
 		if (this.fileWatchers.has(filePath)) {
@@ -70,8 +79,10 @@ export class FileContextTracker {
 		this.fileWatchers.set(filePath, watcher)
 	}
 
-	// Tracks a file operation in metadata and sets up a watcher for the file
-	// This is the main entry point for FileContextTracker and is called when a file is passed to Cline via a tool, mention, or edit.
+	/**
+	 * Tracks a file operation in metadata and sets up a watcher for the file
+	 * This is the main entry point for FileContextTracker and is called when a file is passed to Cline via a tool, mention, or edit.
+	 */
 	async trackFileContext(filePath: string, operation: "read_tool" | "user_edited" | "cline_edited" | "file_mentioned") {
 		try {
 			const cwd = this.getCwd()
@@ -89,9 +100,11 @@ export class FileContextTracker {
 		}
 	}
 
-	// Adds a file to the metadata tracker
-	// This handles the business logic of determining if the file is new, stale, or active.
-	// It also updates the metadata with the latest read/edit dates.
+	/**
+	 * Adds a file to the metadata tracker
+	 * This handles the business logic of determining if the file is new, stale, or active.
+	 * It also updates the metadata with the latest read/edit dates.
+	 */
 	async addFileToFileContextTracker(
 		context: vscode.ExtensionContext,
 		taskId: string,
@@ -154,23 +167,149 @@ export class FileContextTracker {
 		}
 	}
 
-	// Returns (and then clears) the set of recently modified files
+	/**
+	 * Returns (and then clears) the set of recently modified files
+	 */
 	getAndClearRecentlyModifiedFiles(): string[] {
 		const files = Array.from(this.recentlyModifiedFiles)
 		this.recentlyModifiedFiles.clear()
 		return files
 	}
 
-	// Marks a file as edited by Cline to prevent false positives in file watchers
+	/**
+	 * Marks a file as edited by Cline to prevent false positives in file watchers
+	 */
 	markFileAsEditedByCline(filePath: string): void {
 		this.recentlyEditedByCline.add(filePath)
 	}
 
-	// Disposes all file watchers
+	/**
+	 * Disposes all file watchers
+	 */
 	dispose(): void {
 		for (const watcher of this.fileWatchers.values()) {
 			watcher.dispose()
 		}
 		this.fileWatchers.clear()
 	}
+
+	/**
+	 * Detects files that were edited by Cline or users after a specific message timestamp
+	 * This is used when restoring checkpoints to warn about potential file content mismatches
+	 */
+	async detectFilesEditedAfterMessage(messageTs: number, deletedMessages: ClineMessage[]): Promise<string[]> {
+		const editedFiles: string[] = []
+
+		try {
+			// Check task metadata for files that were edited by Cline or users after the message timestamp
+			const taskMetadata = await getTaskMetadata(this.context, this.taskId)
+
+			if (taskMetadata?.files_in_context) {
+				for (const fileEntry of taskMetadata.files_in_context) {
+					const clineEditedAfter = fileEntry.cline_edit_date && fileEntry.cline_edit_date > messageTs
+					const userEditedAfter = fileEntry.user_edit_date && fileEntry.user_edit_date > messageTs
+
+					if (clineEditedAfter || userEditedAfter) {
+						editedFiles.push(fileEntry.path)
+					}
+				}
+			}
+		} catch (error) {
+			console.error("Error checking file context metadata:", error)
+		}
+
+		// Also check deleted task messages for file operations
+		for (const message of deletedMessages) {
+			if (message.say === "tool" && message.text) {
+				try {
+					const toolData = JSON.parse(message.text)
+					if ((toolData.tool === "editedExistingFile" || toolData.tool === "newFileCreated") && toolData.path) {
+						if (!editedFiles.includes(toolData.path)) {
+							editedFiles.push(toolData.path)
+						}
+					}
+				} catch (error) {
+					console.error("Error checking task messages:", error)
+				}
+			}
+		}
+		return [...new Set(editedFiles)]
+	}
+
+	/**
+	 * Stores pending file context warning in workspace state so it persists across task reinitialization
+	 */
+	async storePendingFileContextWarning(files: string[]): Promise<void> {
+		try {
+			const key = `pendingFileContextWarning_${this.taskId}`
+			await updateWorkspaceState(this.context, key, files)
+		} catch (error) {
+			console.error("Error storing pending file context warning:", error)
+		}
+	}
+
+	/**
+	 * Retrieves pending file context warning from workspace state (without clearing it)
+	 */
+	async retrievePendingFileContextWarning(): Promise<string[] | undefined> {
+		try {
+			const key = `pendingFileContextWarning_${this.taskId}`
+			const files = (await getWorkspaceState(this.context, key)) as string[]
+			return files
+		} catch (error) {
+			console.error("Error retrieving pending file context warning:", error)
+		}
+		return undefined
+	}
+
+	/**
+	 * Retrieves and clears pending file context warning from workspace state
+	 */
+	async retrieveAndClearPendingFileContextWarning(): Promise<string[] | undefined> {
+		try {
+			const files = await this.retrievePendingFileContextWarning()
+			if (files) {
+				await updateWorkspaceState(this.context, `pendingFileContextWarning_${this.taskId}`, undefined)
+				return files
+			}
+		} catch (error) {
+			console.error("Error retrieving pending file context warning:", error)
+		}
+		return undefined
+	}
+
+	/**
+	 * Static method to clean up orphaned pending file context warnings at startup
+	 * This removes warnings for tasks that may no longer exist
+	 */
+	static async cleanupOrphanedWarnings(context: vscode.ExtensionContext): Promise<void> {
+		const startTime = Date.now()
+		try {
+			const taskHistory = ((await getGlobalState(context, "taskHistory")) as Array<{ id: string }>) || []
+			const existingTaskIds = new Set(taskHistory.map((task) => task.id))
+			const allStateKeys = context.workspaceState.keys()
+			const pendingWarningKeys = allStateKeys.filter((key) => key.startsWith("pendingFileContextWarning_"))
+
+			const orphanedPendingContextTasks: string[] = []
+			for (const key of pendingWarningKeys) {
+				const taskId = key.replace("pendingFileContextWarning_", "")
+				if (!existingTaskIds.has(taskId)) {
+					orphanedPendingContextTasks.push(key)
+				}
+			}
+
+			if (orphanedPendingContextTasks.length > 0) {
+				for (const key of orphanedPendingContextTasks) {
+					await updateWorkspaceState(context, key, undefined)
+				}
+			}
+
+			const duration = Date.now() - startTime
+			console.log(
+				`FileContextTracker: Processed ${existingTaskIds.size} tasks, found ${pendingWarningKeys.length} pending warnings, ${orphanedPendingContextTasks.length} orphaned, deleted ${orphanedPendingContextTasks.length}, took ${duration}ms`,
+			)
+		} catch (error) {
+			console.error("Error cleaning up orphaned file context warnings:", error)
+		}
+	}
 }

src/core/controller/checkpoints/checkpointRestore.ts

@@ -9,7 +9,7 @@ export async function checkpointRestore(controller: Controller, request: Checkpo
 
 	if (request.number) {
 		// wait for messages to be loaded
-		await pWaitFor(() => controller.task?.isInitialized === true, {
+		await pWaitFor(() => controller.task?.taskState.isInitialized === true, {
 			timeout: 3_000,
 		}).catch(() => {
 			console.error("Failed to init new cline instance")

src/core/controller/file/getRelativePaths.ts

@@ -3,6 +3,8 @@ import { RelativePathsRequest, RelativePaths } from "@shared/proto/file"
 import { FileMethodHandler } from "./index"
 import * as vscode from "vscode"
 import * as path from "path"
+import { UriServiceClient } from "@hosts/host-bridge-client"
+import { Metadata, StringRequest } from "@shared/proto/common"
 
 /**
  * Converts a list of URIs to workspace-relative paths
@@ -17,7 +19,15 @@ export const getRelativePaths: FileMethodHandler = async (
 	const resolvedPaths = await Promise.all(
 		request.uris.map(async (uriString) => {
 			try {
-				const fileUri = vscode.Uri.parse(uriString, true)
+				// Use the host URI service client instead of directly using vscode.Uri.parse
+				const parseResponse = await UriServiceClient.parse(
+					StringRequest.create({
+						metadata: Metadata.create({}),
+						value: uriString,
+					}),
+				)
+				const fileUri = vscode.Uri.parse(`${parseResponse.scheme}://${parseResponse.authority}${parseResponse.path}`)
+				console.log("[DEBUG] UriServiceClient.parse:", fileUri)
 				const relativePathToGet = vscode.workspace.asRelativePath(fileUri, false)
 
 				// If the path is still absolute, it's outside the workspace

src/core/controller/file/subscribeToWorkspaceUpdates.ts

@@ -0,0 +1,58 @@
+import { Controller } from "../index"
+import { EmptyRequest, StringArray } from "@shared/proto/common"
+import { StreamingResponseHandler, getRequestRegistry } from "../grpc-handler"
+
+// Keep track of active subscriptions
+const activeWorkspaceUpdateSubscriptions = new Set<StreamingResponseHandler>()
+
+/**
+ * Subscribe to workspace file updates
+ * @param controller The controller instance
+ * @param request The empty request
+ * @param responseStream The streaming response handler
+ * @param requestId The ID of the request (passed by the gRPC handler)
+ */
+export async function subscribeToWorkspaceUpdates(
+	controller: Controller,
+	request: EmptyRequest,
+	responseStream: StreamingResponseHandler,
+	requestId?: string,
+): Promise<void> {
+	// Add this subscription to the active subscriptions
+	activeWorkspaceUpdateSubscriptions.add(responseStream)
+
+	// Register cleanup when the connection is closed
+	const cleanup = () => {
+		activeWorkspaceUpdateSubscriptions.delete(responseStream)
+	}
+
+	// Register the cleanup function with the request registry if we have a requestId
+	if (requestId) {
+		getRequestRegistry().registerRequest(requestId, cleanup, { type: "workspace_update_subscription" }, responseStream)
+	}
+}
+
+/**
+ * Send a workspace update event to all active subscribers
+ * @param filePaths Array of file paths to send
+ */
+export async function sendWorkspaceUpdateEvent(filePaths: string[]): Promise<void> {
+	// Send the event to all active subscribers
+	const promises = Array.from(activeWorkspaceUpdateSubscriptions).map(async (responseStream) => {
+		try {
+			const event = StringArray.create({
+				values: filePaths,
+			})
+			await responseStream(
+				event,
+				false, // Not the last message
+			)
+		} catch (error) {
+			console.error("Error sending workspace update event:", error)
+			// Remove the subscription if there was an error
+			activeWorkspaceUpdateSubscriptions.delete(responseStream)
+		}
+	})
+
+	await Promise.all(promises)
+}

src/core/controller/file/toggleWorkflow.ts

@@ -0,0 +1,44 @@
+import { Controller } from ".."
+import { Metadata } from "../../../shared/proto/common"
+import { ToggleWorkflowRequest, ClineRulesToggles } from "../../../shared/proto/file"
+import { getWorkspaceState, updateWorkspaceState, getGlobalState, updateGlobalState } from "../../../core/storage/state"
+import { ClineRulesToggles as AppClineRulesToggles } from "../../../shared/cline-rules"
+
+/**
+ * Toggles a workflow on or off
+ * @param controller The controller instance
+ * @param request The request containing the workflow path and enabled state
+ * @returns The updated workflow toggles
+ */
+export async function toggleWorkflow(controller: Controller, request: ToggleWorkflowRequest): Promise<ClineRulesToggles> {
+	const { workflowPath, enabled, isGlobal } = request
+
+	if (!workflowPath || typeof enabled !== "boolean") {
+		console.error("toggleWorkflow: Missing or invalid parameters", {
+			workflowPath,
+			enabled: typeof enabled === "boolean" ? enabled : `Invalid: ${typeof enabled}`,
+		})
+		throw new Error("Missing or invalid parameters for toggleWorkflow")
+	}
+
+	// Update the toggles based on isGlobal flag
+	if (isGlobal) {
+		// Global workflows
+		const toggles = ((await getGlobalState(controller.context, "globalWorkflowToggles")) as AppClineRulesToggles) || {}
+		toggles[workflowPath] = enabled
+		await updateGlobalState(controller.context, "globalWorkflowToggles", toggles)
+		await controller.postStateToWebview()
+
+		// Return the global toggles
+		return ClineRulesToggles.create({ toggles: toggles })
+	} else {
+		// Workspace workflows
+		const toggles = ((await getWorkspaceState(controller.context, "workflowToggles")) as AppClineRulesToggles) || {}
+		toggles[workflowPath] = enabled
+		await updateWorkspaceState(controller.context, "workflowToggles", toggles)
+		await controller.postStateToWebview()
+
+		// Return the workspace toggles
+		return ClineRulesToggles.create({ toggles: toggles })
+	}
+}

src/core/controller/grpc-request-registry.ts

@@ -64,17 +64,17 @@ export class GrpcRequestRegistry {
 	 */
 	public cancelRequest(requestId: string): boolean {
 		const requestInfo = this.activeRequests.get(requestId)
-		if (requestInfo) {
-			try {
-				requestInfo.cleanup()
-				console.log(`[DEBUG] Cleaned up request: ${requestId}`)
-			} catch (error) {
-				console.error(`Error cleaning up request ${requestId}:`, error)
-			}
-			this.activeRequests.delete(requestId)
-			return true
+		if (!requestInfo) {
+			return false
 		}
-		return false
+		try {
+			requestInfo.cleanup()
+			console.log(`[DEBUG] Cleaned up request: ${requestId}`)
+		} catch (error) {
+			console.error(`Error cleaning up request ${requestId}:`, error)
+		}
+		this.activeRequests.delete(requestId)
+		return true
 	}
 
 	/**

src/core/controller/index.ts

@@ -1,23 +1,17 @@
 import { Anthropic } from "@anthropic-ai/sdk"
 import axios from "axios"
-
+import { v4 as uuidv4 } from "uuid"
 import fs from "fs/promises"
 import { setTimeout as setTimeoutPromise } from "node:timers/promises"
 import pWaitFor from "p-wait-for"
 import * as path from "path"
 import * as vscode from "vscode"
 import { handleGrpcRequest, handleGrpcRequestCancel } from "./grpc-handler"
-import { handleModelsServiceRequest } from "./models"
-import { EmptyRequest } from "@shared/proto/common"
 import { buildApiHandler } from "@api/index"
 import { cleanupLegacyCheckpoints } from "@integrations/checkpoints/CheckpointMigration"
 import { downloadTask } from "@integrations/misc/export-markdown"
-import { fetchOpenGraphData } from "@integrations/misc/link-preview"
-import { handleFileServiceRequest } from "./file"
-import { getTheme } from "@integrations/theme/getTheme"
 import WorkspaceTracker from "@integrations/workspace/WorkspaceTracker"
 import { ClineAccountService } from "@services/account/ClineAccountService"
-import { BrowserSession } from "@services/browser/BrowserSession"
 import { McpHub } from "@services/mcp/McpHub"
 import { telemetryService } from "@/services/posthog/telemetry/TelemetryService"
 import { ApiProvider, ModelInfo } from "@shared/api"
@@ -25,38 +19,30 @@ import { ChatContent } from "@shared/ChatContent"
 import { ChatSettings } from "@shared/ChatSettings"
 import { ExtensionMessage, ExtensionState, Platform } from "@shared/ExtensionMessage"
 import { HistoryItem } from "@shared/HistoryItem"
-import { McpDownloadResponse, McpMarketplaceCatalog, McpServer } from "@shared/mcp"
+import { McpMarketplaceCatalog } from "@shared/mcp"
 import { TelemetrySetting } from "@shared/TelemetrySetting"
 import { WebviewMessage } from "@shared/WebviewMessage"
 import { fileExistsAtPath } from "@utils/fs"
 import { getWorkingState } from "@utils/git"
 import { extractCommitMessage } from "@integrations/git/commit-message-generator"
-import { getTotalTasksSize } from "@utils/storage"
-import {
-	ensureMcpServersDirectoryExists,
-	ensureSettingsDirectoryExists,
-	GlobalFileNames,
-	ensureWorkflowsDirectoryExists,
-} from "../storage/disk"
+import { ensureMcpServersDirectoryExists, ensureSettingsDirectoryExists, GlobalFileNames } from "../storage/disk"
 import {
 	getAllExtensionState,
 	getGlobalState,
 	getSecret,
 	getWorkspaceState,
-	resetExtensionState,
 	storeSecret,
 	updateApiConfiguration,
 	updateGlobalState,
 	updateWorkspaceState,
 } from "../storage/state"
-import { Task, cwd } from "../task"
+import { Task } from "../task"
 import { ClineRulesToggles } from "@shared/cline-rules"
 import { sendStateUpdate } from "./state/subscribeToState"
 import { sendAddToInputEvent } from "./ui/subscribeToAddToInput"
 import { sendAuthCallbackEvent } from "./account/subscribeToAuthCallback"
-import { refreshClineRulesToggles } from "@core/context/instructions/user-instructions/cline-rules"
-import { refreshExternalRulesToggles } from "@core/context/instructions/user-instructions/external-rules"
-import { refreshWorkflowToggles } from "@core/context/instructions/user-instructions/workflows"
+import { sendMcpMarketplaceCatalogEvent } from "./mcp/subscribeToMcpMarketplaceCatalog"
+import { sendRelinquishControlEvent } from "./ui/subscribeToRelinquishControl"
 
 /*
 https://github.com/microsoft/vscode-webview-ui-toolkit-samples/blob/main/default/weather-webview/src/providers/WeatherViewProvider.ts
@@ -65,6 +51,7 @@ https://github.com/KumarVariable/vscode-extension-sidebar-html/blob/master/src/c
 */
 
 export class Controller {
+	readonly id: string = uuidv4()
 	private postMessage: (message: ExtensionMessage) => Thenable<boolean> | undefined
 
 	private disposables: vscode.Disposable[] = []
@@ -82,7 +69,7 @@ export class Controller {
 		this.outputChannel.appendLine("ClineProvider instantiated")
 		this.postMessage = postMessage
 
-		this.workspaceTracker = new WorkspaceTracker((msg) => this.postMessageToWebview(msg))
+		this.workspaceTracker = new WorkspaceTracker()
 		this.mcpHub = new McpHub(
 			() => ensureMcpServersDirectoryExists(),
 			() => ensureSettingsDirectoryExists(this.context),
@@ -127,7 +114,7 @@ export class Controller {
 		try {
 			await storeSecret(this.context, "clineApiKey", undefined)
 			await updateGlobalState(this.context, "userInfo", undefined)
-			await updateGlobalState(this.context, "apiProvider", "openrouter")
+			await updateWorkspaceState(this.context, "apiProvider", "openrouter")
 			await this.postStateToWebview()
 			vscode.window.showInformationMessage("Successfully logged out of Cline")
 		} catch (error) {
@@ -143,11 +130,13 @@ export class Controller {
 		await this.clearTask() // ensures that an existing task doesn't exist before starting a new one, although this shouldn't be possible since user must clear task before starting a new one
 		const {
 			apiConfiguration,
-			customInstructions,
 			autoApprovalSettings,
 			browserSettings,
 			chatSettings,
 			shellIntegrationTimeout,
+			terminalReuseEnabled,
+			terminalOutputLineLimit,
+			defaultTerminalProfile,
 			enableCheckpointsSetting,
 			isNewUser,
 			taskHistory,
@@ -182,8 +171,10 @@ export class Controller {
 			browserSettings,
 			chatSettings,
 			shellIntegrationTimeout,
+			terminalReuseEnabled ?? true,
+			terminalOutputLineLimit ?? 500,
+			defaultTerminalProfile ?? "default",
 			enableCheckpointsSetting ?? true,
-			customInstructions,
 			task,
 			images,
 			files,
@@ -215,73 +206,7 @@ export class Controller {
 				await this.setUserInfo(message.user || undefined)
 				await this.postStateToWebview()
 				break
-			case "webviewDidLaunch":
-				this.postStateToWebview()
-				this.workspaceTracker?.populateFilePaths() // don't await
-				getTheme().then((theme) =>
-					this.postMessageToWebview({
-						type: "theme",
-						text: JSON.stringify(theme),
-					}),
-				)
-				// post last cached models in case the call to endpoint fails
-				this.readOpenRouterModels().then((openRouterModels) => {
-					if (openRouterModels) {
-						this.postMessageToWebview({
-							type: "openRouterModels",
-							openRouterModels,
-						})
-					}
-				})
-				// gui relies on model info to be up-to-date to provide the most accurate pricing, so we need to fetch the latest details on launch.
-				// we do this for all users since many users switch between api providers and if they were to switch back to openrouter it would be showing outdated model info if we hadn't retrieved the latest at this point
-				// (see normalizeApiConfiguration > openrouter)
-				// Prefetch marketplace and OpenRouter models
 
-				getGlobalState(this.context, "mcpMarketplaceCatalog").then((mcpMarketplaceCatalog) => {
-					if (mcpMarketplaceCatalog) {
-						this.postMessageToWebview({
-							type: "mcpMarketplaceCatalog",
-							mcpMarketplaceCatalog: mcpMarketplaceCatalog as McpMarketplaceCatalog,
-						})
-					}
-				})
-				this.silentlyRefreshMcpMarketplace()
-				handleModelsServiceRequest(this, "refreshOpenRouterModels", EmptyRequest.create()).then(async (response) => {
-					if (response && response.models) {
-						// update model info in state (this needs to be done here since we don't want to update state while settings is open, and we may refresh models there)
-						const { apiConfiguration } = await getAllExtensionState(this.context)
-						if (apiConfiguration.openRouterModelId && response.models[apiConfiguration.openRouterModelId]) {
-							await updateGlobalState(
-								this.context,
-								"openRouterModelInfo",
-								response.models[apiConfiguration.openRouterModelId],
-							)
-							await this.postStateToWebview()
-						}
-					}
-				})
-				break
-			case "newTask":
-				// Code that should run in response to the hello message command
-				//vscode.window.showInformationMessage(message.text!)
-
-				// Send a message to our webview.
-				// You can send any JSON serializable data.
-				// Could also do this in extension .ts
-				//this.postMessageToWebview({ type: "text", text: `Extension: ${Date.now()}` })
-				// initializing new instance of Cline will make sure that any agentically running promises in old instance don't affect our new task. this essentially creates a fresh slate for the new task
-				await this.initTask(message.text, message.images, message.files)
-				break
-			case "apiConfiguration":
-				if (message.apiConfiguration) {
-					await updateApiConfiguration(this.context, message.apiConfiguration)
-					if (this.task) {
-						this.task.api = buildApiHandler(message.apiConfiguration)
-					}
-				}
-				await this.postStateToWebview()
-				break
 			case "fetchUserCreditsData": {
 				await this.fetchUserCreditsData()
 				break
@@ -290,63 +215,7 @@ export class Controller {
 				await this.fetchMcpMarketplace(message.bool)
 				break
 			}
-			// case "openMcpMarketplaceServerDetails": {
-			// 	if (message.text) {
-			// 		const response = await fetch(`https://api.cline.bot/v1/mcp/marketplace/item?mcpId=${message.mcpId}`)
-			// 		const details: McpDownloadResponse = await response.json()
 
-			// 		if (details.readmeContent) {
-			// 			// Disable markdown preview markers
-			// 			const config = vscode.workspace.getConfiguration("markdown")
-			// 			await config.update("preview.markEditorSelection", false, true)
-
-			// 			// Create URI with base64 encoded markdown content
-			// 			const uri = vscode.Uri.parse(
-			// 				`${DIFF_VIEW_URI_SCHEME}:${details.name} README?${Buffer.from(details.readmeContent).toString("base64")}`,
-			// 			)
-
-			// 			// close existing
-			// 			const tabs = vscode.window.tabGroups.all
-			// 				.flatMap((tg) => tg.tabs)
-			// 				.filter((tab) => tab.label && tab.label.includes("README") && tab.label.includes("Preview"))
-			// 			for (const tab of tabs) {
-			// 				await vscode.window.tabGroups.close(tab)
-			// 			}
-
-			// 			// Show only the preview
-			// 			await vscode.commands.executeCommand("markdown.showPreview", uri, {
-			// 				sideBySide: true,
-			// 				preserveFocus: true,
-			// 			})
-			// 		}
-			// 	}
-
-			// 	this.postMessageToWebview({ type: "relinquishControl" })
-
-			// 	break
-			// }
-			case "toggleWorkflow": {
-				const { workflowPath, enabled, isGlobal } = message
-				if (workflowPath && typeof enabled === "boolean" && typeof isGlobal === "boolean") {
-					if (isGlobal) {
-						const globalWorkflowToggles =
-							((await getGlobalState(this.context, "globalWorkflowToggles")) as ClineRulesToggles) || {}
-						globalWorkflowToggles[workflowPath] = enabled
-						await updateGlobalState(this.context, "globalWorkflowToggles", globalWorkflowToggles)
-						await this.postStateToWebview()
-					} else {
-						const toggles = ((await getWorkspaceState(this.context, "workflowToggles")) as ClineRulesToggles) || {}
-						toggles[workflowPath] = enabled
-						await updateWorkspaceState(this.context, "workflowToggles", toggles)
-						await this.postStateToWebview()
-					}
-				}
-				break
-			}
-			case "fetchLatestMcpServersFromHub": {
-				this.mcpHub?.sendLatestMcpServers()
-				break
-			}
 			// telemetry
 			case "telemetrySetting": {
 				if (message.telemetrySetting) {
@@ -355,48 +224,7 @@ export class Controller {
 				await this.postStateToWebview()
 				break
 			}
-			case "updateSettings": {
-				// api config
-				if (message.apiConfiguration) {
-					await updateApiConfiguration(this.context, message.apiConfiguration)
-					if (this.task) {
-						this.task.api = buildApiHandler(message.apiConfiguration)
-					}
-				}
 
-				// custom instructions
-				await this.updateCustomInstructions(message.customInstructionsSetting)
-
-				// telemetry setting
-				if (message.telemetrySetting) {
-					await this.updateTelemetrySetting(message.telemetrySetting)
-				}
-
-				// plan act setting
-				await updateGlobalState(this.context, "planActSeparateModelsSetting", message.planActSeparateModelsSetting)
-
-				if (typeof message.enableCheckpointsSetting === "boolean") {
-					await updateGlobalState(this.context, "enableCheckpointsSetting", message.enableCheckpointsSetting)
-				}
-
-				if (typeof message.mcpMarketplaceEnabled === "boolean") {
-					await updateGlobalState(this.context, "mcpMarketplaceEnabled", message.mcpMarketplaceEnabled)
-				}
-
-				// chat settings (including preferredLanguage and openAIReasoningEffort)
-				if (message.chatSettings) {
-					await updateGlobalState(this.context, "chatSettings", message.chatSettings)
-					if (this.task) {
-						this.task.chatSettings = message.chatSettings
-					}
-				}
-
-				// after settings are updated, post state to webview
-				await this.postStateToWebview()
-
-				await this.postMessageToWebview({ type: "didUpdateSettings" })
-				break
-			}
 			case "clearAllTaskHistory": {
 				const answer = await vscode.window.showWarningMessage(
 					"What would you like to delete?",
@@ -413,7 +241,7 @@ export class Controller {
 					await this.deleteAllTaskHistory()
 					await this.postStateToWebview()
 				}
-				this.postMessageToWebview({ type: "relinquishControl" })
+				sendRelinquishControlEvent()
 				break
 			}
 			case "grpc_request": {
@@ -428,13 +256,6 @@ export class Controller {
 				}
 				break
 			}
-			case "executeQuickWin":
-				if (message.payload) {
-					const { command, title } = message.payload
-					this.outputChannel.appendLine(`Received executeQuickWin: command='${command}', title='${title}'`)
-					await this.initTask(title)
-				}
-				break
 
 			// Add more switch case statements here as more webview message commands
 			// are created within the webview context (i.e. inside media/main.js)
@@ -464,6 +285,12 @@ export class Controller {
 			previousModeReasoningEffort: newReasoningEffort,
 			previousModeAwsBedrockCustomSelected: newAwsBedrockCustomSelected,
 			previousModeAwsBedrockCustomModelBaseId: newAwsBedrockCustomModelBaseId,
+			previousModeSapAiCoreClientId: newSapAiCoreClientId,
+			previousModeSapAiCoreClientSecret: newSapAiCoreClientSecret,
+			previousModeSapAiCoreBaseUrl: newSapAiCoreBaseUrl,
+			previousModeSapAiCoreTokenUrl: newSapAiCoreTokenUrl,
+			previousModeSapAiCoreResourceGroup: newSapAiResourceGroup,
+			previousModeSapAiCoreModelId: newSapAiCoreModelId,
 			planActSeparateModelsSetting,
 		} = await getAllExtensionState(this.context)
 
@@ -471,9 +298,9 @@ export class Controller {
 
 		if (shouldSwitchModel) {
 			// Save the last model used in this mode
-			await updateGlobalState(this.context, "previousModeApiProvider", apiConfiguration.apiProvider)
-			await updateGlobalState(this.context, "previousModeThinkingBudgetTokens", apiConfiguration.thinkingBudgetTokens)
-			await updateGlobalState(this.context, "previousModeReasoningEffort", apiConfiguration.reasoningEffort)
+			await updateWorkspaceState(this.context, "previousModeApiProvider", apiConfiguration.apiProvider)
+			await updateWorkspaceState(this.context, "previousModeThinkingBudgetTokens", apiConfiguration.thinkingBudgetTokens)
+			await updateWorkspaceState(this.context, "previousModeReasoningEffort", apiConfiguration.reasoningEffort)
 			switch (apiConfiguration.apiProvider) {
 				case "anthropic":
 				case "vertex":
@@ -483,16 +310,16 @@ export class Controller {
 				case "qwen":
 				case "deepseek":
 				case "xai":
-					await updateGlobalState(this.context, "previousModeModelId", apiConfiguration.apiModelId)
+					await updateWorkspaceState(this.context, "previousModeModelId", apiConfiguration.apiModelId)
 					break
 				case "bedrock":
-					await updateGlobalState(this.context, "previousModeModelId", apiConfiguration.apiModelId)
-					await updateGlobalState(
+					await updateWorkspaceState(this.context, "previousModeModelId", apiConfiguration.apiModelId)
+					await updateWorkspaceState(
 						this.context,
 						"previousModeAwsBedrockCustomSelected",
 						apiConfiguration.awsBedrockCustomSelected,
 					)
-					await updateGlobalState(
+					await updateWorkspaceState(
 						this.context,
 						"previousModeAwsBedrockCustomModelBaseId",
 						apiConfiguration.awsBedrockCustomModelBaseId,
@@ -500,34 +327,51 @@ export class Controller {
 					break
 				case "openrouter":
 				case "cline":
-					await updateGlobalState(this.context, "previousModeModelId", apiConfiguration.openRouterModelId)
-					await updateGlobalState(this.context, "previousModeModelInfo", apiConfiguration.openRouterModelInfo)
+					await updateWorkspaceState(this.context, "previousModeModelId", apiConfiguration.openRouterModelId)
+					await updateWorkspaceState(this.context, "previousModeModelInfo", apiConfiguration.openRouterModelInfo)
 					break
 				case "vscode-lm":
 					// Important we don't set modelId to this, as it's an object not string (webview expects model id to be a string)
-					await updateGlobalState(
+					await updateWorkspaceState(
 						this.context,
 						"previousModeVsCodeLmModelSelector",
 						apiConfiguration.vsCodeLmModelSelector,
 					)
 					break
 				case "openai":
-					await updateGlobalState(this.context, "previousModeModelId", apiConfiguration.openAiModelId)
-					await updateGlobalState(this.context, "previousModeModelInfo", apiConfiguration.openAiModelInfo)
+					await updateWorkspaceState(this.context, "previousModeModelId", apiConfiguration.openAiModelId)
+					await updateWorkspaceState(this.context, "previousModeModelInfo", apiConfiguration.openAiModelInfo)
 					break
 				case "ollama":
-					await updateGlobalState(this.context, "previousModeModelId", apiConfiguration.ollamaModelId)
+					await updateWorkspaceState(this.context, "previousModeModelId", apiConfiguration.ollamaModelId)
 					break
 				case "lmstudio":
-					await updateGlobalState(this.context, "previousModeModelId", apiConfiguration.lmStudioModelId)
+					await updateWorkspaceState(this.context, "previousModeModelId", apiConfiguration.lmStudioModelId)
 					break
 				case "litellm":
-					await updateGlobalState(this.context, "previousModeModelId", apiConfiguration.liteLlmModelId)
-					await updateGlobalState(this.context, "previousModeModelInfo", apiConfiguration.liteLlmModelInfo)
+					await updateWorkspaceState(this.context, "previousModeModelId", apiConfiguration.liteLlmModelId)
+					await updateWorkspaceState(this.context, "previousModeModelInfo", apiConfiguration.liteLlmModelInfo)
 					break
 				case "requesty":
-					await updateGlobalState(this.context, "previousModeModelId", apiConfiguration.requestyModelId)
-					await updateGlobalState(this.context, "previousModeModelInfo", apiConfiguration.requestyModelInfo)
+					await updateWorkspaceState(this.context, "previousModeModelId", apiConfiguration.requestyModelId)
+					await updateWorkspaceState(this.context, "previousModeModelInfo", apiConfiguration.requestyModelInfo)
+					break
+				case "sapaicore":
+					await updateWorkspaceState(this.context, "previousModeModelId", apiConfiguration.apiModelId)
+					await updateWorkspaceState(this.context, "previousModeSapAiCoreClientId", apiConfiguration.sapAiCoreClientId)
+					await updateWorkspaceState(
+						this.context,
+						"previousModeSapAiCoreClientSecret",
+						apiConfiguration.sapAiCoreClientSecret,
+					)
+					await updateWorkspaceState(this.context, "previousModeSapAiCoreBaseUrl", apiConfiguration.sapAiCoreBaseUrl)
+					await updateWorkspaceState(this.context, "previousModeSapAiCoreTokenUrl", apiConfiguration.sapAiCoreTokenUrl)
+					await updateWorkspaceState(
+						this.context,
+						"previousModeSapAiCoreResourceGroup",
+						apiConfiguration.sapAiResourceGroup,
+					)
+					await updateWorkspaceState(this.context, "previousModeSapAiCoreModelId", apiConfiguration.sapAiCoreModelId)
 					break
 			}
 
@@ -539,9 +383,9 @@ export class Controller {
 				newReasoningEffort ||
 				newVsCodeLmModelSelector
 			) {
-				await updateGlobalState(this.context, "apiProvider", newApiProvider)
-				await updateGlobalState(this.context, "thinkingBudgetTokens", newThinkingBudgetTokens)
-				await updateGlobalState(this.context, "reasoningEffort", newReasoningEffort)
+				await updateWorkspaceState(this.context, "apiProvider", newApiProvider)
+				await updateWorkspaceState(this.context, "thinkingBudgetTokens", newThinkingBudgetTokens)
+				await updateWorkspaceState(this.context, "reasoningEffort", newReasoningEffort)
 				switch (newApiProvider) {
 					case "anthropic":
 					case "vertex":
@@ -551,38 +395,41 @@ export class Controller {
 					case "qwen":
 					case "deepseek":
 					case "xai":
-						await updateGlobalState(this.context, "apiModelId", newModelId)
+						await updateWorkspaceState(this.context, "apiModelId", newModelId)
 						break
 					case "bedrock":
-						await updateGlobalState(this.context, "apiModelId", newModelId)
-						await updateGlobalState(this.context, "awsBedrockCustomSelected", newAwsBedrockCustomSelected)
-						await updateGlobalState(this.context, "awsBedrockCustomModelBaseId", newAwsBedrockCustomModelBaseId)
+						await updateWorkspaceState(this.context, "apiModelId", newModelId)
+						await updateWorkspaceState(this.context, "awsBedrockCustomSelected", newAwsBedrockCustomSelected)
+						await updateWorkspaceState(this.context, "awsBedrockCustomModelBaseId", newAwsBedrockCustomModelBaseId)
 						break
 					case "openrouter":
 					case "cline":
-						await updateGlobalState(this.context, "openRouterModelId", newModelId)
-						await updateGlobalState(this.context, "openRouterModelInfo", newModelInfo)
+						await updateWorkspaceState(this.context, "openRouterModelId", newModelId)
+						await updateWorkspaceState(this.context, "openRouterModelInfo", newModelInfo)
 						break
 					case "vscode-lm":
-						await updateGlobalState(this.context, "vsCodeLmModelSelector", newVsCodeLmModelSelector)
+						await updateWorkspaceState(this.context, "vsCodeLmModelSelector", newVsCodeLmModelSelector)
 						break
 					case "openai":
-						await updateGlobalState(this.context, "openAiModelId", newModelId)
-						await updateGlobalState(this.context, "openAiModelInfo", newModelInfo)
+						await updateWorkspaceState(this.context, "openAiModelId", newModelId)
+						await updateWorkspaceState(this.context, "openAiModelInfo", newModelInfo)
 						break
 					case "ollama":
-						await updateGlobalState(this.context, "ollamaModelId", newModelId)
+						await updateWorkspaceState(this.context, "ollamaModelId", newModelId)
 						break
 					case "lmstudio":
-						await updateGlobalState(this.context, "lmStudioModelId", newModelId)
+						await updateWorkspaceState(this.context, "lmStudioModelId", newModelId)
 						break
 					case "litellm":
-						await updateGlobalState(this.context, "previousModeModelId", apiConfiguration.liteLlmModelId)
-						await updateGlobalState(this.context, "previousModeModelInfo", apiConfiguration.liteLlmModelInfo)
+						await updateWorkspaceState(this.context, "liteLlmModelId", newModelId)
+						await updateWorkspaceState(this.context, "liteLlmModelInfo", newModelInfo)
 						break
 					case "requesty":
-						await updateGlobalState(this.context, "requestyModelId", newModelId)
-						await updateGlobalState(this.context, "requestyModelInfo", newModelInfo)
+						await updateWorkspaceState(this.context, "requestyModelId", newModelId)
+						await updateWorkspaceState(this.context, "requestyModelInfo", newModelInfo)
+						break
+					case "sapaicore":
+						await updateWorkspaceState(this.context, "apiModelId", newModelId)
 						break
 				}
 
@@ -593,13 +440,13 @@ export class Controller {
 			}
 		}
 
-		await updateGlobalState(this.context, "chatSettings", chatSettings)
+		await updateWorkspaceState(this.context, "chatSettings", chatSettings)
 		await this.postStateToWebview()
 
 		if (this.task) {
 			this.task.chatSettings = chatSettings
-			if (this.task.isAwaitingPlanResponse && didSwitchToActMode) {
-				this.task.didRespondToPlanAskBySwitchingMode = true
+			if (this.task.taskState.isAwaitingPlanResponse && didSwitchToActMode) {
+				this.task.taskState.didRespondToPlanAskBySwitchingMode = true
 				// Use chatContent if provided, otherwise use default message
 				await this.task.handleWebviewAskResponse(
 					"messageResponse",
@@ -624,9 +471,9 @@ export class Controller {
 			await pWaitFor(
 				() =>
 					this.task === undefined ||
-					this.task.isStreaming === false ||
-					this.task.didFinishAbortingStream ||
-					this.task.isWaitingForFirstChunk, // if only first chunk is processed, then there's no need to wait for graceful abort (closes edits, browser, etc)
+					this.task.taskState.isStreaming === false ||
+					this.task.taskState.didFinishAbortingStream ||
+					this.task.taskState.isWaitingForFirstChunk, // if only first chunk is processed, then there's no need to wait for graceful abort (closes edits, browser, etc)
 				{
 					timeout: 3_000,
 				},
@@ -635,21 +482,13 @@ export class Controller {
 			})
 			if (this.task) {
 				// 'abandoned' will prevent this cline instance from affecting future cline instance gui. this may happen if its hanging on a streaming request
-				this.task.abandoned = true
+				this.task.taskState.abandoned = true
 			}
 			await this.initTask(undefined, undefined, undefined, historyItem) // clears task again, so we need to abortTask manually above
 			// await this.postStateToWebview() // new Cline instance will post state when it's ready. having this here sent an empty messages array to webview leading to virtuoso having to reload the entire list
 		}
 	}
 
-	async updateCustomInstructions(instructions?: string) {
-		// User may be clearing the field
-		await updateGlobalState(this.context, "customInstructions", instructions || undefined)
-		if (this.task) {
-			this.task.customInstructions = instructions || undefined
-		}
-	}
-
 	// Account
 
 	async fetchUserCreditsData() {
@@ -684,7 +523,7 @@ export class Controller {
 			await sendAuthCallbackEvent(customToken)
 
 			const clineProvider: ApiProvider = "cline"
-			await updateGlobalState(this.context, "apiProvider", clineProvider)
+			await updateWorkspaceState(this.context, "apiProvider", clineProvider)
 
 			// Update API configuration with the new provider and API key
 			const { apiConfiguration } = await getAllExtensionState(this.context)
@@ -738,10 +577,6 @@ export class Controller {
 			console.error("Failed to fetch MCP marketplace:", error)
 			if (!silent) {
 				const errorMessage = error instanceof Error ? error.message : "Failed to fetch MCP marketplace"
-				await this.postMessageToWebview({
-					type: "mcpMarketplaceCatalog",
-					error: errorMessage,
-				})
 				vscode.window.showErrorMessage(errorMessage)
 			}
 			return undefined
@@ -787,10 +622,7 @@ export class Controller {
 		try {
 			const catalog = await this.fetchMcpMarketplaceFromApi(true)
 			if (catalog) {
-				await this.postMessageToWebview({
-					type: "mcpMarketplaceCatalog",
-					mcpMarketplaceCatalog: catalog,
-				})
+				await sendMcpMarketplaceCatalogEvent(catalog)
 			}
 		} catch (error) {
 			console.error("Failed to silently refresh MCP marketplace:", error)
@@ -818,27 +650,17 @@ export class Controller {
 				| McpMarketplaceCatalog
 				| undefined
 			if (!forceRefresh && cachedCatalog?.items) {
-				await this.postMessageToWebview({
-					type: "mcpMarketplaceCatalog",
-					mcpMarketplaceCatalog: cachedCatalog,
-				})
+				await sendMcpMarketplaceCatalogEvent(cachedCatalog)
 				return
 			}
 
 			const catalog = await this.fetchMcpMarketplaceFromApi(false)
 			if (catalog) {
-				await this.postMessageToWebview({
-					type: "mcpMarketplaceCatalog",
-					mcpMarketplaceCatalog: catalog,
-				})
+				await sendMcpMarketplaceCatalogEvent(catalog)
 			}
 		} catch (error) {
 			console.error("Failed to handle cached MCP marketplace:", error)
 			const errorMessage = error instanceof Error ? error.message : "Failed to handle cached MCP marketplace"
-			await this.postMessageToWebview({
-				type: "mcpMarketplaceCatalog",
-				error: errorMessage,
-			})
 			vscode.window.showErrorMessage(errorMessage)
 		}
 	}
@@ -860,7 +682,7 @@ export class Controller {
 		}
 
 		const openrouter: ApiProvider = "openrouter"
-		await updateGlobalState(this.context, "apiProvider", openrouter)
+		await updateWorkspaceState(this.context, "apiProvider", openrouter)
 		await storeSecret(this.context, "openRouterApiKey", apiKey)
 		await this.postStateToWebview()
 		if (this.task) {
@@ -1018,18 +840,6 @@ export class Controller {
 		throw new Error("Task not found")
 	}
 
-	async showTaskWithId(id: string) {
-		if (id !== this.task?.taskId) {
-			// non-current task
-			const { historyItem } = await this.getTaskWithId(id)
-			await this.initTask(undefined, undefined, undefined, historyItem) // clears existing task
-		}
-		await this.postMessageToWebview({
-			type: "action",
-			action: "chatButtonClicked",
-		})
-	}
-
 	async exportTaskWithId(id: string) {
 		const { historyItem, apiConversationHistory } = await this.getTaskWithId(id)
 		await downloadTask(historyItem.ts, apiConversationHistory)
@@ -1153,30 +963,31 @@ export class Controller {
 
 	async postStateToWebview() {
 		const state = await this.getStateToPostToWebview()
-		// For testing: Bypass gRPC stream and send state directly
-		console.log("[Controller Test Revert] Posting full state via direct 'state' message.")
-		await this.postMessageToWebview({ type: "state", state: state })
-		// await sendStateUpdate(state) // Original line for the GrPC stream
+		await sendStateUpdate(state)
 	}
 
 	async getStateToPostToWebview(): Promise<ExtensionState> {
 		const {
 			apiConfiguration,
 			lastShownAnnouncementId,
-			customInstructions,
 			taskHistory,
 			autoApprovalSettings,
 			browserSettings,
 			chatSettings,
 			userInfo,
 			mcpMarketplaceEnabled,
+			mcpRichDisplayEnabled,
 			telemetrySetting,
 			planActSeparateModelsSetting,
 			enableCheckpointsSetting,
 			globalClineRulesToggles,
 			globalWorkflowToggles,
 			shellIntegrationTimeout,
+			terminalReuseEnabled,
+			defaultTerminalProfile,
 			isNewUser,
+			mcpResponsesCollapsed,
+			terminalOutputLineLimit,
 		} = await getAllExtensionState(this.context)
 
 		const localClineRulesToggles =
@@ -1193,11 +1004,10 @@ export class Controller {
 		return {
 			version: this.context.extension?.packageJSON?.version ?? "",
 			apiConfiguration,
-			customInstructions,
 			uriScheme: vscode.env.uriScheme,
 			currentTaskItem: this.task?.taskId ? (taskHistory || []).find((item) => item.id === this.task?.taskId) : undefined,
-			checkpointTrackerErrorMessage: this.task?.checkpointTrackerErrorMessage,
-			clineMessages: this.task?.clineMessages || [],
+			checkpointTrackerErrorMessage: this.task?.taskState.checkpointTrackerErrorMessage,
+			clineMessages: this.task?.messageStateHandler.getClineMessages() || [],
 			taskHistory: (taskHistory || [])
 				.filter((item) => item.ts && item.task)
 				.sort((a, b) => b.ts - a.ts)
@@ -1209,6 +1019,7 @@ export class Controller {
 			chatSettings,
 			userInfo,
 			mcpMarketplaceEnabled,
+			mcpRichDisplayEnabled,
 			telemetrySetting,
 			planActSeparateModelsSetting,
 			enableCheckpointsSetting: enableCheckpointsSetting ?? true,
@@ -1220,7 +1031,11 @@ export class Controller {
 			localWorkflowToggles: localWorkflowToggles || {},
 			globalWorkflowToggles: globalWorkflowToggles || {},
 			shellIntegrationTimeout,
+			terminalReuseEnabled,
+			defaultTerminalProfile,
 			isNewUser,
+			mcpResponsesCollapsed,
+			terminalOutputLineLimit,
 		}
 	}
 

src/core/controller/mcp/downloadMcp.ts

@@ -3,6 +3,7 @@ import { Empty, StringRequest } from "../../../shared/proto/common"
 import { McpServer, McpDownloadResponse } from "@shared/mcp"
 import axios from "axios"
 import * as vscode from "vscode"
+import { sendChatButtonClickedEvent } from "../ui/subscribeToChatButtonClicked"
 
 /**
  * Download an MCP server from the marketplace
@@ -77,10 +78,7 @@ Here is the project's README to help you get started:\n\n${mcpDetails.readmeCont
 
 		// Initialize task and show chat view
 		await controller.initTask(task)
-		await controller.postMessageToWebview({
-			type: "action",
-			action: "chatButtonClicked",
-		})
+		await sendChatButtonClickedEvent(controller.id)
 
 		// Return an empty response - the client only cares if the call succeeded
 		return Empty.create()

src/core/controller/mcp/getLatestMcpServers.ts

@@ -0,0 +1,25 @@
+import type { Empty } from "@shared/proto/common"
+import { McpServers } from "@shared/proto/mcp"
+import type { Controller } from "../index"
+import { convertMcpServersToProtoMcpServers } from "@/shared/proto-conversions/mcp/mcp-server-conversion"
+
+/**
+ * RPC handler for getting the latest MCP servers
+ * @param controller The controller instance
+ * @param _request Empty request
+ * @returns McpServers response with list of all MCP servers
+ */
+export async function getLatestMcpServers(controller: Controller, _request: Empty): Promise<McpServers> {
+	try {
+		// Get sorted servers from mcpHub using the RPC variant
+		const mcpServers = (await controller.mcpHub?.getLatestMcpServersRPC()) || []
+
+		// Convert to proto format
+		const protoServers = convertMcpServersToProtoMcpServers(mcpServers)
+
+		return McpServers.create({ mcpServers: protoServers })
+	} catch (error) {
+		console.error("Error fetching latest MCP servers:", error)
+		throw error
+	}
+}

src/core/controller/mcp/subscribeToMcpMarketplaceCatalog.ts

@@ -0,0 +1,55 @@
+import { Controller } from "../index"
+import { EmptyRequest } from "@shared/proto/common"
+import { McpMarketplaceCatalog } from "@shared/proto/mcp"
+import { StreamingResponseHandler, getRequestRegistry } from "../grpc-handler"
+
+// Keep track of active subscriptions
+const activeMcpMarketplaceSubscriptions = new Set<StreamingResponseHandler>()
+
+/**
+ * Subscribe to MCP marketplace catalog updates
+ * @param controller The controller instance
+ * @param request The empty request
+ * @param responseStream The streaming response handler
+ * @param requestId The ID of the request (passed by the gRPC handler)
+ */
+export async function subscribeToMcpMarketplaceCatalog(
+	controller: Controller,
+	request: EmptyRequest,
+	responseStream: StreamingResponseHandler,
+	requestId?: string,
+): Promise<void> {
+	// Add this subscription to the active subscriptions
+	activeMcpMarketplaceSubscriptions.add(responseStream)
+
+	// Register cleanup when the connection is closed
+	const cleanup = () => {
+		activeMcpMarketplaceSubscriptions.delete(responseStream)
+	}
+
+	// Register the cleanup function with the request registry if we have a requestId
+	if (requestId) {
+		getRequestRegistry().registerRequest(requestId, cleanup, { type: "mcp_marketplace_subscription" }, responseStream)
+	}
+}
+
+/**
+ * Send an MCP marketplace catalog event to all active subscribers
+ */
+export async function sendMcpMarketplaceCatalogEvent(catalog: McpMarketplaceCatalog): Promise<void> {
+	// Send the event to all active subscribers
+	const promises = Array.from(activeMcpMarketplaceSubscriptions).map(async (responseStream) => {
+		try {
+			await responseStream(
+				catalog,
+				false, // Not the last message
+			)
+		} catch (error) {
+			console.error("Error sending MCP marketplace catalog event:", error)
+			// Remove the subscription if there was an error
+			activeMcpMarketplaceSubscriptions.delete(responseStream)
+		}
+	})
+
+	await Promise.all(promises)
+}

src/core/controller/mcp/subscribeToMcpServers.ts

@@ -0,0 +1,76 @@
+import { Controller } from "../index"
+import { EmptyRequest } from "@shared/proto/common"
+import { McpServers } from "@shared/proto/mcp"
+import { StreamingResponseHandler, getRequestRegistry } from "../grpc-handler"
+import { convertMcpServersToProtoMcpServers } from "@shared/proto-conversions/mcp/mcp-server-conversion"
+
+// Keep track of active subscriptions
+const activeMcpServersSubscriptions = new Set<StreamingResponseHandler>()
+
+/**
+ * Subscribe to MCP servers events
+ * @param controller The controller instance
+ * @param request The empty request
+ * @param responseStream The streaming response handler
+ * @param requestId The ID of the request (passed by the gRPC handler)
+ */
+export async function subscribeToMcpServers(
+	controller: Controller,
+	request: EmptyRequest,
+	responseStream: StreamingResponseHandler,
+	requestId?: string,
+): Promise<void> {
+	// Add this subscription to the active subscriptions
+	activeMcpServersSubscriptions.add(responseStream)
+
+	// Register cleanup when the connection is closed
+	const cleanup = () => {
+		activeMcpServersSubscriptions.delete(responseStream)
+	}
+
+	// Register the cleanup function with the request registry if we have a requestId
+	if (requestId) {
+		getRequestRegistry().registerRequest(requestId, cleanup, { type: "mcpServers_subscription" }, responseStream)
+	}
+
+	// Send initial state if available
+	if (controller.mcpHub) {
+		const mcpServers = controller.mcpHub.getServers()
+		if (mcpServers.length > 0) {
+			try {
+				const protoServers = McpServers.create({
+					mcpServers: convertMcpServersToProtoMcpServers(mcpServers),
+				})
+				await responseStream(
+					protoServers,
+					false, // Not the last message
+				)
+			} catch (error) {
+				console.error("Error sending initial MCP servers:", error)
+				activeMcpServersSubscriptions.delete(responseStream)
+			}
+		}
+	}
+}
+
+/**
+ * Send an MCP servers update to all active subscribers
+ * @param mcpServers The MCP servers to send
+ */
+export async function sendMcpServersUpdate(mcpServers: McpServers): Promise<void> {
+	// Send the event to all active subscribers
+	const promises = Array.from(activeMcpServersSubscriptions).map(async (responseStream) => {
+		try {
+			await responseStream(
+				mcpServers,
+				false, // Not the last message
+			)
+		} catch (error) {
+			console.error("Error sending MCP servers update:", error)
+			// Remove the subscription if there was an error
+			activeMcpServersSubscriptions.delete(responseStream)
+		}
+	})
+
+	await Promise.all(promises)
+}

src/core/controller/mcp/updateMcpTimeout.ts

@@ -12,7 +12,6 @@ export async function updateMcpTimeout(controller: Controller, request: UpdateMc
 	try {
 		if (request.serverName && typeof request.serverName === "string" && typeof request.timeout === "number") {
 			const mcpServers = await controller.mcpHub?.updateServerTimeoutRPC(request.serverName, request.timeout)
-			console.log("mcpServers", mcpServers)
 			const convertedMcpServers = convertMcpServersToProtoMcpServers(mcpServers)
 			console.log("convertedMcpServers", convertedMcpServers)
 			return McpServers.create({ mcpServers: convertedMcpServers })

src/core/controller/models/refreshOpenRouterModels.ts

@@ -15,11 +15,11 @@ import { GlobalFileNames } from "@core/storage/disk"
  */
 export async function refreshOpenRouterModels(
 	controller: Controller,
-	request: EmptyRequest,
+	_request: EmptyRequest,
 ): Promise<OpenRouterCompatibleModelInfo> {
 	const openRouterModelsFilePath = path.join(await ensureCacheDirectoryExists(controller), GlobalFileNames.openRouterModels)
 
-	let models: Record<string, Partial<OpenRouterModelInfo>> = {}
+	let models: Record<string, OpenRouterModelInfo> = {}
 	try {
 		const response = await axios.get("https://openrouter.ai/api/v1/models")
 
@@ -32,15 +32,20 @@ export async function refreshOpenRouterModels(
 				return undefined
 			}
 			for (const rawModel of rawModels) {
-				const modelInfo: Partial<OpenRouterModelInfo> = {
-					maxTokens: rawModel.top_provider?.max_completion_tokens,
-					contextWindow: rawModel.context_length,
-					supportsImages: rawModel.architecture?.modality?.includes("image"),
+				const modelInfo = OpenRouterModelInfo.create({
+					maxTokens: rawModel.top_provider?.max_completion_tokens ?? 0,
+					contextWindow: rawModel.context_length ?? 0,
+					supportsImages: rawModel.architecture?.modality?.includes("image") ?? false,
 					supportsPromptCache: false,
-					inputPrice: parsePrice(rawModel.pricing?.prompt),
-					outputPrice: parsePrice(rawModel.pricing?.completion),
-					description: rawModel.description,
-				}
+					inputPrice: parsePrice(rawModel.pricing?.prompt) ?? 0,
+					outputPrice: parsePrice(rawModel.pricing?.completion) ?? 0,
+					cacheWritesPrice: 0,
+					cacheReadsPrice: 0,
+					description: rawModel.description ?? "",
+					thinkingConfig: rawModel.thinking_config ?? undefined,
+					supportsGlobalEndpoint: rawModel.supports_global_endpoint ?? undefined,
+					tiers: rawModel.tiers ?? [],
+				})
 
 				switch (rawModel.id) {
 					case "anthropic/claude-sonnet-4":
@@ -94,6 +99,11 @@ export async function refreshOpenRouterModels(
 						modelInfo.cacheWritesPrice = 0.14
 						modelInfo.cacheReadsPrice = 0.014
 						break
+					case "x-ai/grok-3-beta":
+						modelInfo.supportsPromptCache = true
+						modelInfo.cacheWritesPrice = 0
+						modelInfo.cacheReadsPrice = 0
+						break
 					default:
 						if (rawModel.id.startsWith("openai/")) {
 							modelInfo.cacheReadsPrice = parsePrice(rawModel.pricing?.input_cache_read)
@@ -112,13 +122,18 @@ export async function refreshOpenRouterModels(
 						break
 				}
 
+				// add new model id
+				if (rawModel.id === "x-ai/grok-3-beta") {
+					models["x-ai/grok-3"] = modelInfo
+				}
+
 				models[rawModel.id] = modelInfo
 			}
 		} else {
 			console.error("Invalid response from OpenRouter API")
 		}
 		await fs.writeFile(openRouterModelsFilePath, JSON.stringify(models))
-		console.log("OpenRouter models fetched and saved", models)
+		console.log("OpenRouter models fetched and saved", JSON.stringify(models).slice(0, 300))
 	} catch (error) {
 		console.error("Error fetching OpenRouter models:", error)
 
@@ -129,30 +144,13 @@ export async function refreshOpenRouterModels(
 		}
 	}
 
-	// Convert the Record<string, Partial<OpenRouterModelInfo>> to Record<string, OpenRouterModelInfo>
-	// by filling in any missing required fields with defaults
-	const typedModels: Record<string, OpenRouterModelInfo> = {}
-	for (const [key, model] of Object.entries(models)) {
-		typedModels[key] = {
-			maxTokens: model.maxTokens ?? 0,
-			contextWindow: model.contextWindow ?? 0,
-			supportsImages: model.supportsImages ?? false,
-			supportsPromptCache: model.supportsPromptCache ?? false,
-			inputPrice: model.inputPrice ?? 0,
-			outputPrice: model.outputPrice ?? 0,
-			cacheWritesPrice: model.cacheWritesPrice ?? 0,
-			cacheReadsPrice: model.cacheReadsPrice ?? 0,
-			description: model.description ?? "",
-		}
-	}
-
-	return OpenRouterCompatibleModelInfo.create({ models: typedModels })
+	return OpenRouterCompatibleModelInfo.create({ models })
 }
 
 /**
  * Reads cached OpenRouter models from disk
  */
-async function readOpenRouterModels(controller: Controller): Promise<Record<string, Partial<OpenRouterModelInfo>> | undefined> {
+async function readOpenRouterModels(controller: Controller): Promise<Record<string, OpenRouterModelInfo> | undefined> {
 	const openRouterModelsFilePath = path.join(await ensureCacheDirectoryExists(controller), GlobalFileNames.openRouterModels)
 	const fileExists = await fileExistsAtPath(openRouterModelsFilePath)
 	if (fileExists) {

src/core/controller/models/refreshRequestyModels.ts

@@ -41,11 +41,6 @@ export async function refreshRequestyModels(controller: Controller, _: EmptyRequ
 				models[model.id] = modelInfo
 			}
 			console.log("Requesty models fetched", models)
-
-			controller.postMessageToWebview({
-				type: "requestyModels",
-				requestyModels: models,
-			})
 		} else {
 			console.error("Invalid response from Requesty API")
 		}

src/core/controller/models/subscribeToOpenRouterModels.ts

@@ -0,0 +1,60 @@
+import { Controller } from "../index"
+import { EmptyRequest } from "@shared/proto/common"
+import { OpenRouterCompatibleModelInfo } from "@shared/proto/models"
+import { StreamingResponseHandler, getRequestRegistry } from "../grpc-handler"
+
+// Keep track of active OpenRouter models subscriptions
+const activeOpenRouterModelsSubscriptions = new Set<StreamingResponseHandler>()
+
+/**
+ * Subscribe to OpenRouter models events
+ * @param controller The controller instance
+ * @param request The empty request
+ * @param responseStream The streaming response handler
+ * @param requestId The ID of the request (passed by the gRPC handler)
+ */
+export async function subscribeToOpenRouterModels(
+	controller: Controller,
+	request: EmptyRequest,
+	responseStream: StreamingResponseHandler,
+	requestId?: string,
+): Promise<void> {
+	console.log("[DEBUG] set up OpenRouter models subscription")
+
+	// Add this subscription to the active subscriptions
+	activeOpenRouterModelsSubscriptions.add(responseStream)
+
+	// Register cleanup when the connection is closed
+	const cleanup = () => {
+		activeOpenRouterModelsSubscriptions.delete(responseStream)
+		console.log("[DEBUG] Cleaned up OpenRouter models subscription")
+	}
+
+	// Register the cleanup function with the request registry if we have a requestId
+	if (requestId) {
+		getRequestRegistry().registerRequest(requestId, cleanup, { type: "openRouterModels_subscription" }, responseStream)
+	}
+}
+
+/**
+ * Send an OpenRouter models event to all active subscribers
+ * @param models The OpenRouter models to send
+ */
+export async function sendOpenRouterModelsEvent(models: OpenRouterCompatibleModelInfo): Promise<void> {
+	// Send the event to all active subscribers
+	const promises = Array.from(activeOpenRouterModelsSubscriptions).map(async (responseStream) => {
+		try {
+			await responseStream(
+				models,
+				false, // Not the last message
+			)
+			console.log("[DEBUG] sending OpenRouter models event")
+		} catch (error) {
+			console.error("Error sending OpenRouter models event:", error)
+			// Remove the subscription if there was an error
+			activeOpenRouterModelsSubscriptions.delete(responseStream)
+		}
+	})
+
+	await Promise.all(promises)
+}

src/core/controller/models/updateApiConfigurationProto.ts

@@ -0,0 +1,43 @@
+import type { Controller } from "../index"
+import { Empty } from "@shared/proto/common"
+import { UpdateApiConfigurationRequest } from "@shared/proto/models"
+import { updateApiConfiguration } from "../../storage/state"
+import { buildApiHandler } from "@api/index"
+import { convertProtoToApiConfiguration } from "@shared/proto-conversions/models/api-configuration-conversion"
+
+/**
+ * Updates API configuration
+ * @param controller The controller instance
+ * @param request The update API configuration request
+ * @returns Empty response
+ */
+export async function updateApiConfigurationProto(
+	controller: Controller,
+	request: UpdateApiConfigurationRequest,
+): Promise<Empty> {
+	try {
+		if (!request.apiConfiguration) {
+			console.log("[APICONFIG: updateApiConfigurationProto] API configuration is required")
+			throw new Error("API configuration is required")
+		}
+
+		// Convert proto ApiConfiguration to application ApiConfiguration
+		const appApiConfiguration = convertProtoToApiConfiguration(request.apiConfiguration)
+
+		// Update the API configuration in storage
+		await updateApiConfiguration(controller.context, appApiConfiguration)
+
+		// Update the task's API handler if there's an active task
+		if (controller.task) {
+			controller.task.api = buildApiHandler(appApiConfiguration)
+		}
+
+		// Post updated state to webview
+		await controller.postStateToWebview()
+
+		return Empty.create()
+	} catch (error) {
+		console.error(`Failed to update API configuration: ${error}`)
+		throw error
+	}
+}

src/core/controller/state/getAvailableTerminalProfiles.ts

@@ -0,0 +1,19 @@
+import { Controller } from "../index"
+import * as proto from "@/shared/proto"
+import { getAvailableTerminalProfiles as getTerminalProfilesFromShell } from "../../../utils/shell"
+
+export async function getAvailableTerminalProfiles(
+	controller: Controller,
+	request: proto.cline.EmptyRequest,
+): Promise<proto.cline.TerminalProfiles> {
+	const profiles = getTerminalProfilesFromShell()
+
+	return proto.cline.TerminalProfiles.create({
+		profiles: profiles.map((profile) => ({
+			id: profile.id,
+			name: profile.name,
+			path: profile.path || "",
+			description: profile.description || "",
+		})),
+	})
+}

src/core/controller/state/resetState.ts

@@ -1,18 +1,25 @@
 import { Controller } from ".."
-import { Empty, EmptyRequest } from "../../../shared/proto/common"
-import { resetExtensionState } from "../../../core/storage/state"
+import { Empty } from "../../../shared/proto/common"
+import { ResetStateRequest } from "../../../shared/proto/state"
+import { resetGlobalState, resetWorkspaceState } from "../../../core/storage/state"
 import * as vscode from "vscode"
+import { sendChatButtonClickedEvent } from "../ui/subscribeToChatButtonClicked"
 
 /**
  * Resets the extension state to its defaults
  * @param controller The controller instance
- * @param request An empty request (no parameters needed)
+ * @param request The reset state request containing the global flag
  * @returns An empty response
  */
-export async function resetState(controller: Controller, request: EmptyRequest): Promise<Empty> {
+export async function resetState(controller: Controller, request: ResetStateRequest): Promise<Empty> {
 	try {
-		vscode.window.showInformationMessage("Resetting state...")
-		await resetExtensionState(controller.context)
+		if (request.global) {
+			vscode.window.showInformationMessage("Resetting global state...")
+			await resetGlobalState(controller.context)
+		} else {
+			vscode.window.showInformationMessage("Resetting workspace state...")
+			await resetWorkspaceState(controller.context)
+		}
 
 		if (controller.task) {
 			controller.task.abortTask()
@@ -22,10 +29,7 @@ export async function resetState(controller: Controller, request: EmptyRequest):
 		vscode.window.showInformationMessage("State reset")
 		await controller.postStateToWebview()
 
-		await controller.postMessageToWebview({
-			type: "action",
-			action: "chatButtonClicked",
-		})
+		await sendChatButtonClickedEvent(controller.id)
 
 		return Empty.create()
 	} catch (error) {

src/core/controller/state/updateDefaultTerminalProfile.ts

@@ -0,0 +1,50 @@
+import * as vscode from "vscode"
+import { Controller } from "../index"
+import * as proto from "@/shared/proto"
+import { updateGlobalState } from "../../storage/state"
+import { TerminalInfo } from "@/integrations/terminal/TerminalRegistry"
+
+export async function updateDefaultTerminalProfile(
+	controller: Controller,
+	request: proto.cline.StringRequest,
+): Promise<proto.cline.TerminalProfileUpdateResponse> {
+	const profileId = request.value
+
+	// Update the terminal profile in the state
+	await updateGlobalState(controller.context, "defaultTerminalProfile", profileId)
+
+	let closedCount = 0
+	let busyTerminals: TerminalInfo[] = []
+
+	// Update the terminal manager of the current task if it exists
+	if (controller.task) {
+		// Call the updated setDefaultTerminalProfile method that returns closed terminal info
+		const result = controller.task.terminalManager.setDefaultTerminalProfile(profileId)
+		closedCount = result.closedCount
+		busyTerminals = result.busyTerminals
+
+		// Show information message if terminals were closed
+		if (closedCount > 0) {
+			vscode.window.showInformationMessage(
+				`Closed ${closedCount} ${closedCount === 1 ? "terminal" : "terminals"} with different profile.`,
+			)
+		}
+
+		// Show warning if there are busy terminals that couldn't be closed
+		if (busyTerminals.length > 0) {
+			vscode.window.showWarningMessage(
+				`${busyTerminals.length} busy ${busyTerminals.length === 1 ? "terminal has" : "terminals have"} a different profile. ` +
+					`Close ${busyTerminals.length === 1 ? "it" : "them"} to use the new profile for all commands.`,
+			)
+		}
+	}
+
+	// Broadcast state update to all webviews
+	await controller.postStateToWebview()
+
+	return proto.cline.TerminalProfileUpdateResponse.create({
+		closedCount,
+		busyTerminalsCount: busyTerminals.length,
+		hasBusyTerminals: busyTerminals.length > 0,
+	})
+}

src/core/controller/state/updateSettings.ts

@@ -0,0 +1,90 @@
+import { Controller } from ".."
+import { Empty } from "../../../shared/proto/common"
+import { UpdateSettingsRequest } from "../../../shared/proto/state"
+import { updateApiConfiguration } from "../../storage/state"
+import { buildApiHandler } from "../../../api"
+import { convertProtoApiConfigurationToApiConfiguration } from "../../../shared/proto-conversions/state/settings-conversion"
+import { convertProtoChatSettingsToChatSettings } from "../../../shared/proto-conversions/state/chat-settings-conversion"
+import { TelemetrySetting } from "@/shared/TelemetrySetting"
+
+/**
+ * Updates multiple extension settings in a single request
+ * @param controller The controller instance
+ * @param request The request containing the settings to update
+ * @returns An empty response
+ */
+export async function updateSettings(controller: Controller, request: UpdateSettingsRequest): Promise<Empty> {
+	try {
+		// Update API configuration
+		if (request.apiConfiguration) {
+			const apiConfiguration = convertProtoApiConfigurationToApiConfiguration(request.apiConfiguration)
+			await updateApiConfiguration(controller.context, apiConfiguration)
+
+			if (controller.task) {
+				controller.task.api = buildApiHandler(apiConfiguration)
+			}
+		}
+
+		// Update telemetry setting
+		if (request.telemetrySetting) {
+			await controller.updateTelemetrySetting(request.telemetrySetting as TelemetrySetting)
+		}
+
+		// Update plan/act separate models setting
+		if (request.planActSeparateModelsSetting !== undefined) {
+			await controller.context.globalState.update("planActSeparateModelsSetting", request.planActSeparateModelsSetting)
+		}
+
+		// Update checkpoints setting
+		if (request.enableCheckpointsSetting !== undefined) {
+			await controller.context.globalState.update("enableCheckpointsSetting", request.enableCheckpointsSetting)
+		}
+
+		// Update MCP marketplace setting
+		if (request.mcpMarketplaceEnabled !== undefined) {
+			await controller.context.globalState.update("mcpMarketplaceEnabled", request.mcpMarketplaceEnabled)
+		}
+
+		// Update MCP responses collapsed setting
+		if (request.mcpResponsesCollapsed !== undefined) {
+			await controller.context.globalState.update("mcpResponsesCollapsed", request.mcpResponsesCollapsed)
+		}
+
+		// Update MCP responses collapsed setting
+		if (request.mcpRichDisplayEnabled !== undefined) {
+			await controller.context.globalState.update("mcpRichDisplayEnabled", request.mcpRichDisplayEnabled)
+		}
+
+		// Update chat settings
+		if (request.chatSettings) {
+			const chatSettings = convertProtoChatSettingsToChatSettings(request.chatSettings)
+			await controller.context.globalState.update("chatSettings", chatSettings)
+			if (controller.task) {
+				controller.task.chatSettings = chatSettings
+			}
+		}
+
+		// Update terminal timeout setting
+		if (request.shellIntegrationTimeout !== undefined) {
+			await controller.context.globalState.update("shellIntegrationTimeout", Number(request.shellIntegrationTimeout))
+		}
+
+		// Update terminal reuse setting
+		if (request.terminalReuseEnabled !== undefined) {
+			await controller.context.globalState.update("terminalReuseEnabled", request.terminalReuseEnabled)
+		}
+
+		// Update terminal output line limit
+		if (request.terminalOutputLineLimit !== undefined) {
+			await controller.context.globalState.update("terminalOutputLineLimit", Number(request.terminalOutputLineLimit))
+		}
+
+		// Post updated state to webview
+		await controller.postStateToWebview()
+
+		return Empty.create()
+	} catch (error) {
+		console.error("Failed to update settings:", error)
+		throw error
+	}
+}

src/core/controller/state/updateTerminalConnectionTimeout.ts

@@ -1,27 +1,18 @@
-import { Controller } from ".."
-import { Int64, Int64Request } from "../../../shared/proto/common"
+import { Controller } from "../index"
+import * as proto from "@/shared/proto"
 import { updateGlobalState } from "../../storage/state"
 
-/**
- * Updates the terminal connection timeout setting
- * @param controller The controller instance
- * @param request The request containing the timeout value in milliseconds
- * @returns The updated timeout value
- */
-export async function updateTerminalConnectionTimeout(controller: Controller, request: Int64Request): Promise<Int64> {
-	try {
-		const timeout = request.value
+export async function updateTerminalConnectionTimeout(
+	controller: Controller,
+	request: proto.cline.Int64Request,
+): Promise<proto.cline.Int64> {
+	const timeoutValue = request.value
 
-		if (typeof timeout === "number" && !isNaN(timeout) && timeout > 0) {
-			// Update the global state directly
-			await updateGlobalState(controller.context, "shellIntegrationTimeout", timeout)
-			return Int64.create({ value: timeout })
-		} else {
-			console.warn(`Invalid shell integration timeout value received: ${timeout}. Expected a positive number.`)
-			throw new Error("Invalid timeout value. Expected a positive number.")
-		}
-	} catch (error) {
-		console.error(`Failed to update terminal connection timeout: ${error}`)
-		throw error
-	}
+	// Update the terminal connection timeout setting in the state
+	await updateGlobalState(controller.context, "shellIntegrationTimeout", timeoutValue)
+
+	// Broadcast state update to all webviews
+	await controller.postStateToWebview()
+
+	return proto.cline.Int64.create({ value: timeoutValue })
 }

src/core/controller/state/updateTerminalReuseEnabled.ts

@@ -0,0 +1,18 @@
+import { Controller } from "../index"
+import * as proto from "@/shared/proto"
+import { updateGlobalState } from "../../storage/state"
+
+export async function updateTerminalReuseEnabled(
+	controller: Controller,
+	request: proto.cline.BooleanRequest,
+): Promise<proto.cline.Empty> {
+	const enabled = request.value
+
+	// Update the terminal reuse setting in the state
+	await updateGlobalState(controller.context, "terminalReuseEnabled", enabled)
+
+	// Broadcast state update to all webviews
+	await controller.postStateToWebview()
+
+	return proto.cline.Empty.create({})
+}

src/core/controller/task/executeQuickWin.ts

@@ -0,0 +1,35 @@
+import { ExecuteQuickWinRequest } from "@shared/proto/task"
+import { Empty } from "@shared/proto/common"
+import type { Controller } from "../index"
+
+/**
+ * Executes a quick win task with command and title
+ * @param controller The controller instance
+ * @param request The execute quick win request
+ * @returns Empty response
+ *
+ * @example
+ * // Usage from webview:
+ * import { TaskServiceClient } from "@/services/grpc-client"
+ * import { ExecuteQuickWinRequest } from "@shared/proto/task"
+ *
+ * const request: ExecuteQuickWinRequest = {
+ *   command: "npm install",
+ *   title: "Install dependencies"
+ * }
+ *
+ * TaskServiceClient.executeQuickWin(request)
+ *   .then(() => console.log("Quick win executed successfully"))
+ *   .catch(error => console.error("Failed to execute quick win:", error))
+ */
+export async function executeQuickWin(controller: Controller, request: ExecuteQuickWinRequest): Promise<Empty> {
+	try {
+		const { command, title } = request
+		console.log(`Received executeQuickWin: command='${command}', title='${title}'`)
+		await controller.initTask(title)
+		return Empty.create({})
+	} catch (error) {
+		console.error("Failed to execute quick win:", error)
+		throw error
+	}
+}

src/core/controller/task/newTask.ts

@@ -1,6 +1,7 @@
 import { Controller } from ".."
 import { Empty } from "../../../shared/proto/common"
 import { NewTaskRequest } from "../../../shared/proto/task"
+import { handleFileServiceRequest } from "../file"
 
 /**
  * Creates a new task with the given text and optional images

src/core/controller/task/showTaskWithId.ts

@@ -1,6 +1,7 @@
 import { Controller } from ".."
 import { StringRequest } from "../../../shared/proto/common"
 import { TaskResponse } from "../../../shared/proto/task"
+import { sendChatButtonClickedEvent } from "../ui/subscribeToChatButtonClicked"
 
 /**
  * Shows a task with the specified ID
@@ -22,10 +23,7 @@ export async function showTaskWithId(controller: Controller, request: StringRequ
 			await controller.initTask(undefined, undefined, undefined, historyItem)
 
 			// Send UI update to show the chat view
-			await controller.postMessageToWebview({
-				type: "action",
-				action: "chatButtonClicked",
-			})
+			await sendChatButtonClickedEvent(controller.id)
 
 			// Return task data for gRPC response
 			return TaskResponse.create({
@@ -49,10 +47,7 @@ export async function showTaskWithId(controller: Controller, request: StringRequ
 		await controller.initTask(undefined, undefined, undefined, fetchedItem)
 
 		// Send UI update to show the chat view
-		await controller.postMessageToWebview({
-			type: "action",
-			action: "chatButtonClicked",
-		})
+		await sendChatButtonClickedEvent(controller.id)
 
 		return TaskResponse.create({
 			id: fetchedItem.id,

src/core/controller/ui/initializeWebview.ts

@@ -0,0 +1,73 @@
+import type { Controller } from "../index"
+import { EmptyRequest, Empty } from "@shared/proto/common"
+import { handleModelsServiceRequest } from "../models"
+import { getAllExtensionState, getGlobalState, updateWorkspaceState } from "../../storage/state"
+import { sendOpenRouterModelsEvent } from "../models/subscribeToOpenRouterModels"
+import { sendMcpMarketplaceCatalogEvent } from "../mcp/subscribeToMcpMarketplaceCatalog"
+import { telemetryService } from "@/services/posthog/telemetry/TelemetryService"
+import { OpenRouterCompatibleModelInfo } from "@/shared/proto/models"
+import { McpMarketplaceCatalog } from "@shared/mcp"
+
+/**
+ * Initialize webview when it launches
+ * @param controller The controller instance
+ * @param request The empty request
+ * @returns Empty response
+ */
+export async function initializeWebview(controller: Controller, request: EmptyRequest): Promise<Empty> {
+	try {
+		// Populate file paths for workspace tracker (don't await)
+		controller.workspaceTracker?.populateFilePaths()
+
+		// Post last cached models in case the call to endpoint fails
+		controller.readOpenRouterModels().then((openRouterModels) => {
+			if (openRouterModels) {
+				sendOpenRouterModelsEvent(OpenRouterCompatibleModelInfo.create({ models: openRouterModels }))
+			}
+		})
+
+		// Refresh OpenRouter models from API
+		handleModelsServiceRequest(controller, "refreshOpenRouterModels", EmptyRequest.create()).then(async (response) => {
+			if (response && response.models) {
+				// Update model info in state (this needs to be done here since we don't want to update state while settings is open, and we may refresh models there)
+				const { apiConfiguration } = await getAllExtensionState(controller.context)
+				if (apiConfiguration.openRouterModelId && response.models[apiConfiguration.openRouterModelId]) {
+					await updateWorkspaceState(
+						controller.context,
+						"openRouterModelInfo",
+						response.models[apiConfiguration.openRouterModelId],
+					)
+					await controller.postStateToWebview()
+				}
+			}
+		})
+
+		// GUI relies on model info to be up-to-date to provide the most accurate pricing, so we need to fetch the latest details on launch.
+		// We do this for all users since many users switch between api providers and if they were to switch back to openrouter it would be showing outdated model info if we hadn't retrieved the latest at this point
+		// (see normalizeApiConfiguration > openrouter)
+		// Prefetch marketplace and OpenRouter models
+
+		// Send cached MCP marketplace catalog if available
+		getGlobalState(controller.context, "mcpMarketplaceCatalog").then((mcpMarketplaceCatalog) => {
+			if (mcpMarketplaceCatalog) {
+				sendMcpMarketplaceCatalogEvent(mcpMarketplaceCatalog as McpMarketplaceCatalog)
+			}
+		})
+
+		// Silently refresh MCP marketplace catalog
+		controller.silentlyRefreshMcpMarketplace()
+
+		// Initialize telemetry service with user's current setting
+		controller.getStateToPostToWebview().then((state) => {
+			const { telemetrySetting } = state
+			const isOptedIn = telemetrySetting !== "disabled"
+			telemetryService.updateTelemetryState(isOptedIn)
+		})
+
+		return Empty.create({})
+	} catch (error) {
+		console.error("Failed to initialize webview:", error)
+		// Return empty response even on error to not break the frontend
+		return Empty.create({})
+	}
+}

src/core/controller/ui/scrollToSettings.ts

@@ -1,15 +1,15 @@
 import { Controller } from ".."
-import { StringRequest } from "../../../shared/proto/common"
+import { StringRequest, KeyValuePair } from "../../../shared/proto/common"
 
 /**
  * Executes a scroll to settings action
  * @param controller The controller instance
  * @param request The request containing the ID of the settings section to scroll to
- * @returns An object with action and value fields for the UI to process
+ * @returns KeyValuePair with action and value fields for the UI to process
  */
-export async function scrollToSettings(controller: Controller, request: StringRequest): Promise<Record<string, string>> {
-	return {
-		action: "scrollToSettings",
+export async function scrollToSettings(controller: Controller, request: StringRequest): Promise<KeyValuePair> {
+	return KeyValuePair.create({
+		key: "scrollToSettings",
 		value: request.value || "",
-	}
+	})
 }

src/core/controller/ui/subscribeToAccountButtonClicked.ts

@@ -0,0 +1,55 @@
+import { Controller } from "../index"
+import { Empty, EmptyRequest } from "@shared/proto/common"
+import { StreamingResponseHandler, getRequestRegistry } from "../grpc-handler"
+
+// Track subscriptions by controller ID
+const activeSubscriptions = new Map<string, StreamingResponseHandler>()
+
+/**
+ * Subscribe to account button clicked events
+ * @param controller The controller instance
+ * @param request The empty request
+ * @param responseStream The streaming response handler
+ * @param requestId The request ID for cleanup
+ */
+export async function subscribeToAccountButtonClicked(
+	controller: Controller,
+	_request: EmptyRequest,
+	responseStream: StreamingResponseHandler,
+	requestId?: string,
+): Promise<void> {
+	const controllerId = controller.id
+
+	// Store subscription with controller ID
+	activeSubscriptions.set(controllerId, responseStream)
+
+	// Register cleanup
+	const cleanup = () => {
+		activeSubscriptions.delete(controllerId)
+	}
+
+	if (requestId) {
+		getRequestRegistry().registerRequest(requestId, cleanup, { type: "account_button_subscription" }, responseStream)
+	}
+}
+
+/**
+ * Send account button clicked event to a specific controller
+ * @param controllerId The ID of the controller to send the event to
+ */
+export async function sendAccountButtonClickedEvent(controllerId: string): Promise<void> {
+	const responseStream = activeSubscriptions.get(controllerId)
+
+	if (!responseStream) {
+		console.log(`No active subscription for controller ${controllerId}`)
+		return
+	}
+
+	try {
+		const event: Empty = Empty.create({})
+		await responseStream(event, false)
+	} catch (error) {
+		console.error(`Error sending account button clicked event to controller ${controllerId}:`, error)
+		activeSubscriptions.delete(controllerId)
+	}
+}

src/core/controller/ui/subscribeToChatButtonClicked.ts

@@ -0,0 +1,63 @@
+import { Controller } from "../index"
+import { Empty } from "@shared/proto/common"
+import { EmptyRequest } from "@shared/proto/common"
+import { StreamingResponseHandler, getRequestRegistry } from "../grpc-handler"
+
+// Keep track of active chatButtonClicked subscriptions by controller ID
+const activeChatButtonClickedSubscriptions = new Map<string, StreamingResponseHandler>()
+
+/**
+ * Subscribe to chatButtonClicked events
+ * @param controller The controller instance
+ * @param request The empty request
+ * @param responseStream The streaming response handler
+ * @param requestId The ID of the request (passed by the gRPC handler)
+ */
+export async function subscribeToChatButtonClicked(
+	controller: Controller,
+	_request: EmptyRequest,
+	responseStream: StreamingResponseHandler,
+	requestId?: string,
+): Promise<void> {
+	const controllerId = controller.id
+	console.log(`[DEBUG] set up chatButtonClicked subscription for controller ${controllerId}`)
+
+	// Add this subscription to the active subscriptions with the controller ID
+	activeChatButtonClickedSubscriptions.set(controllerId, responseStream)
+
+	// Register cleanup when the connection is closed
+	const cleanup = () => {
+		activeChatButtonClickedSubscriptions.delete(controllerId)
+	}
+
+	// Register the cleanup function with the request registry if we have a requestId
+	if (requestId) {
+		getRequestRegistry().registerRequest(requestId, cleanup, { type: "chatButtonClicked_subscription" }, responseStream)
+	}
+}
+
+/**
+ * Send a chatButtonClicked event to a specific controller's subscription
+ * @param controllerId The ID of the controller to send the event to
+ */
+export async function sendChatButtonClickedEvent(controllerId: string): Promise<void> {
+	// Get the subscription for this specific controller
+	const responseStream = activeChatButtonClickedSubscriptions.get(controllerId)
+
+	if (!responseStream) {
+		console.log(`[DEBUG] No active subscription for controller ${controllerId}`)
+		return
+	}
+
+	try {
+		const event = Empty.create({})
+		await responseStream(
+			event,
+			false, // Not the last message
+		)
+	} catch (error) {
+		console.error(`Error sending chatButtonClicked event to controller ${controllerId}:`, error)
+		// Remove the subscription if there was an error
+		activeChatButtonClickedSubscriptions.delete(controllerId)
+	}
+}

src/core/controller/ui/subscribeToDidBecomeVisible.ts

@@ -0,0 +1,63 @@
+import { Controller } from "../index"
+import { Empty } from "@shared/proto/common"
+import { EmptyRequest } from "@shared/proto/common"
+import { StreamingResponseHandler, getRequestRegistry } from "../grpc-handler"
+
+// Keep track of active didBecomeVisible subscriptions by controller ID
+const activeDidBecomeVisibleSubscriptions = new Map<string, StreamingResponseHandler>()
+
+/**
+ * Subscribe to didBecomeVisible events
+ * @param controller The controller instance
+ * @param request The empty request
+ * @param responseStream The streaming response handler
+ * @param requestId The ID of the request (passed by the gRPC handler)
+ */
+export async function subscribeToDidBecomeVisible(
+	controller: Controller,
+	request: EmptyRequest,
+	responseStream: StreamingResponseHandler,
+	requestId?: string,
+): Promise<void> {
+	const controllerId = controller.id
+	console.log(`[DEBUG] set up didBecomeVisible subscription for controller ${controllerId}`)
+
+	// Add this subscription to the active subscriptions with the controller ID
+	activeDidBecomeVisibleSubscriptions.set(controllerId, responseStream)
+
+	// Register cleanup when the connection is closed
+	const cleanup = () => {
+		activeDidBecomeVisibleSubscriptions.delete(controllerId)
+	}
+
+	// Register the cleanup function with the request registry if we have a requestId
+	if (requestId) {
+		getRequestRegistry().registerRequest(requestId, cleanup, { type: "didBecomeVisible_subscription" }, responseStream)
+	}
+}
+
+/**
+ * Send a didBecomeVisible event to a specific controller's subscription
+ * @param controllerId The ID of the controller to send the event to
+ */
+export async function sendDidBecomeVisibleEvent(controllerId: string): Promise<void> {
+	// Get the subscription for this specific controller
+	const responseStream = activeDidBecomeVisibleSubscriptions.get(controllerId)
+
+	if (!responseStream) {
+		console.log(`[DEBUG] No active subscription for controller ${controllerId}`)
+		return
+	}
+
+	try {
+		const event: Empty = Empty.create({})
+		await responseStream(
+			event,
+			false, // Not the last message
+		)
+	} catch (error) {
+		console.error(`Error sending didBecomeVisible event to controller ${controllerId}:`, error)
+		// Remove the subscription if there was an error
+		activeDidBecomeVisibleSubscriptions.delete(controllerId)
+	}
+}

src/core/controller/ui/subscribeToFocusChatInput.ts

@@ -0,0 +1,62 @@
+import { StringRequest, Empty } from "@shared/proto/common"
+import { StreamingResponseHandler, getRequestRegistry } from "../grpc-handler"
+import type { Controller } from "../index"
+
+// Map client IDs to their subscription handlers
+const focusChatInputSubscriptions = new Map<string, StreamingResponseHandler>()
+
+/**
+ * Subscribe to focus chat input events
+ * @param controller The controller instance
+ * @param request The request containing the client ID
+ * @param responseStream The streaming response handler
+ * @param requestId The ID of the request
+ */
+export async function subscribeToFocusChatInput(
+	controller: Controller,
+	request: StringRequest,
+	responseStream: StreamingResponseHandler,
+	requestId?: string,
+): Promise<void> {
+	const clientId = request.value
+	if (!clientId) {
+		throw new Error("Client ID is required for focusChatInput subscription")
+	}
+
+	// Store this subscription with its client ID
+	focusChatInputSubscriptions.set(clientId, responseStream)
+
+	// Register cleanup when the connection is closed
+	const cleanup = () => {
+		focusChatInputSubscriptions.delete(clientId)
+	}
+
+	// Register the cleanup function with the request registry if we have a requestId
+	if (requestId) {
+		getRequestRegistry().registerRequest(requestId, cleanup, { type: "focus_chat_input_subscription" }, responseStream)
+	}
+}
+
+/**
+ * Send a focus chat input event to a specific webview by client ID
+ * @param clientId The ID of the client to send the event to
+ */
+export async function sendFocusChatInputEvent(clientId: string): Promise<void> {
+	const responseStream = focusChatInputSubscriptions.get(clientId)
+	if (!responseStream) {
+		console.warn(`No subscription found for client ID: ${clientId}`)
+		return
+	}
+
+	try {
+		const event = Empty.create({})
+		await responseStream(
+			event,
+			false, // Not the last message
+		)
+	} catch (error) {
+		console.error(`Error sending focus chat input event to client ${clientId}:`, error)
+		// Remove the subscription if there was an error
+		focusChatInputSubscriptions.delete(clientId)
+	}
+}

src/core/controller/ui/subscribeToHistoryButtonClicked.ts

@@ -14,7 +14,7 @@ const activeHistoryButtonClickedSubscriptions = new Map<StreamingResponseHandler
  * @param requestId The ID of the request (passed by the gRPC handler)
  */
 export async function subscribeToHistoryButtonClicked(
-	controller: Controller,
+	_controller: Controller,
 	request: WebviewProviderTypeRequest,
 	responseStream: StreamingResponseHandler,
 	requestId?: string,