native handler instead of just openai

.changeset/dirty-roses-pretend.md

@@ -1,20 +0,0 @@
----
-"claude-dev": minor
----
-
-Focus Chain Feature
-• Context-aware todo list injection into system prompts based on task state, mode transitions, and reminder intervals
-• Dynamic prompt generation with conditional instructions for Plan/Act mode switching and user-edited lists
-• FocusChainManager class with file-based persistence, real-time watching, and enhanced TaskHeader UI with progress indicators
-• Strategic context inclusion logic: Plan mode transitions, user edits, reminder intervals, and first-time task creation
-
-Deep Planning Slash Command
-• New /deep-planning command for structured 4-step implementation planning workflow
-• Integration with Focus Chain for automatic progress tracking in created tasks
-• Comprehensive prompting system for silent investigation, discussion, plan creation, and task generation
-
-Telemetry
-• Focus Chain usage tracking and Deep planning workflow analytics
-
-Feature Flags
-• PostHog remote feature flag integration for Focus Chain gradual rollout

.clinerules/workflows/pr-review.md

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

.eslintrc.json

@@ -21,7 +21,6 @@
 		"semi": "off",
 		"react-hooks/exhaustive-deps": "off",
 		"eslint-rules/no-direct-vscode-api": "warn",
-		"eslint-rules/no-direct-vscode-state-api": "error",
 		"no-restricted-syntax": [
 			"error",
 			{
@@ -30,5 +29,5 @@
 			}
 		]
 	},
-	"ignorePatterns": ["out", "dist", "dist-standalone", "**/*.d.ts", "node_modules"]
+	"ignorePatterns": ["out", "dist", "**/*.d.ts"]
 }

.github/CODEOWNERS

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

.github/workflows/test.yml

@@ -96,15 +96,17 @@ 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
+            # FIX: Right now the tests are run when the PR is being reviewed, but if main is updated after that, the PR can merge without the tests being run on the latest version of main.
+            # - name: Unit Tests
+            #   run: npm run test:unit
 
             # Run extension tests with coverage
-            - name: Extension Integration Tests with Coverage
+            - name: Extension Tests with Coverage
               id: extension_coverage
               continue-on-error: true
               run: |
-                  node ./scripts/test-ci.js 2>&1 | tee extension_coverage.txt
+                  node ./scripts/test-ci.js > extension_coverage.txt 2>&1
                   # Default the encoding to UTF-8 - It's not the default on Windows
                   PYTHONUTF8=1 PYTHONPATH=.github/scripts python -m coverage_check extract-coverage extension_coverage.txt --type=extension --github-output --verbose
 
@@ -116,7 +118,7 @@ jobs:
                   cd webview-ui
                   # Ensure coverage dependency is installed
                   npm install --no-save @vitest/coverage-v8
-                  npm run test:coverage 2>&1 | tee webview_coverage.txt
+                  npm run test:coverage > webview_coverage.txt 2>&1
                   cd ..
                   # Default the encoding to UTF-8 - It's not the default on Windows
                   PYTHONUTF8=1 PYTHONPATH=.github/scripts python -m coverage_check extract-coverage webview-ui/webview_coverage.txt --type=webview --github-output --verbose
@@ -131,19 +133,21 @@ jobs:
                   path: |
                       extension_coverage.txt
                       webview-ui/webview_coverage.txt
+                  retention-period: workflow # Artifacts are automatically deleted when the workflow completes
 
             # Set the check as failed if any of the tests failed
-            - name: Check for test failures
+            - name: Print test results and check for failures
               run: |
+                  echo "Extension Tests Result: ${{ steps.extension_coverage.outcome }}"
+                  cat extension_coverage.txt
+
+                  echo "Webview Tests Result: ${{ steps.webview_coverage.outcome }}"
+                  cat webview-ui/webview_coverage.txt
+
                   # Check if any of the test steps failed
                   # https://docs.github.com/en/actions/writing-workflows/choosing-what-your-workflow-does/accessing-contextual-information-about-workflow-runs#steps-context
-                  if [ "${{ steps.extension_coverage.outcome }}" != "success" ]; then
-                      echo "Extension Integration Tests failed, see previous step for test output."
-                  fi
-                  if [ "${{ steps.webview_coverage.outcome }}" != "success" ]; then
-                      echo "Webview Tests failed, see previous step for test output."
-                  fi
                   if [ "${{ steps.extension_coverage.outcome }}" != "success" ] || [ "${{ steps.webview_coverage.outcome }}" != "success" ]; then
+                      echo "Tests failed."
                       exit 1
                   fi
 

.vscodeignore

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

CHANGELOG.md

@@ -1,61 +1,5 @@
 # Changelog
 
-## [3.24.0]
-
-- Add OpenAI GPT-5 Chat(gpt-5-chat-latest)
-- Add custom browser arguments setting to allow passing flags to the Chrome executable for better headless compatibility.
-- Add 1m context window model support for claude sonnet 4
-- Fis the API Keys URL for Requesty
-- Set gpt5 max tokens to 8_192 to fix 'context window exceeded' error
-- Fix issue where fallback request to retrieve cost was not using correct auth token
-- Add OpenAI context window exceeded error handling
-- Calibrate input token counts when using anthropic models of sap ai core provider
-
-## [3.23.0]
-
-- Add caching support for Bedrock inferences using SAP AI Core and minor refactor
-- Improve visibility for mode switch background color on different themes
-- Fix terminal commands putting webview in blocked state
-
-## [3.22.0]
-
-- Implemented a retry strategy for Cerebras to handle rate limit issues due to its generation speed
-- Add support for GPT-5 models to SAP AI Core Provider
-- Support sending context to active webview when editor panels are opened.
-- Fix bug where running out of credits on Cline accounts would show '402 empty body' response instead of 'buy credits' component
-- Fix LiteLLM Proxy Provider Cost Tracking
-
-## [3.21.0]
-
-- Add support for GPT-5 model family including GPT-5, GPT-5 Mini, and GPT-5 Nano with prompt caching support and set GPT-5 as the new default model
-- Add "Take a Tour" button for new users to easily access the VSCode walkthrough and improve onboarding experience
-- Enhance plan mode response handling with better exploration parameter support
-
-## [3.20.13]
-
-- Fix prompt caching support for Opus 4.1 on OpenRouter/Cline
-
-## [3.20.12]
-
-- Add Claude Opus 4.1 model support to AWS Bedrock provider (Thanks @omercelik!)
-- Fix prompt caching and extended thinking support for Claude Opus 4.1 in Anthropic provider
-
-## [3.20.11]
-
-Add gpt-oss-120b as a Cerebras model
-Add Opus 4.1 through Claude Code
-
-## [3.20.10]
-
-- Add OpenAI's new open-source models (GPT-OSS-120B and GPT-OSS-20B) to Hugging Face and Groq providers
-
-## [3.20.9]
-
-- Add support for Claude Opus 4.1 model in Anthropic provider
-- Add Baseten as a new API provider with support for DeepSeek, Llama, and Kimi K2 models (Thanks @AlexKer!)
-- Fix error messages not clearing from UI when retrying failed tasks
-- Fix chat input box positioning issues
-
 ## [3.20.8]
 
 - Add navbar tooltips on hover

docs/docs.json

@@ -79,8 +79,6 @@
 					"features/drag-and-drop",
 					"features/plan-and-act",
 					"features/slash-commands/workflows",
-					"features/focus-chain",
-					"features/auto-compact",
 					"features/editing-messages",
 					{
 						"group": "@ Mentions",
@@ -99,8 +97,7 @@
 							"features/slash-commands/new-task",
 							"features/slash-commands/new-rule",
 							"features/slash-commands/smol",
-							"features/slash-commands/report-bug",
-							"features/slash-commands/deep-planning"
+							"features/slash-commands/report-bug"
 						]
 					},
 					{
@@ -149,14 +146,9 @@
 				"pages": [
 					"provider-config/anthropic",
 					"provider-config/claude-code",
-					{
-						"group": "AWS Bedrock",
-						"pages": [
-							"provider-config/aws-bedrock/api-key",
-							"provider-config/aws-bedrock/iam-credentials",
-							"provider-config/aws-bedrock/cli-profile"
-						]
-					},
+					"provider-config/aws-bedrock-with-apikey-authentication",
+					"provider-config/aws-bedrock-with-credentials-authentication",
+					"provider-config/aws-bedrock-with-profile-authentication",
 					"provider-config/gcp-vertex-ai",
 					"provider-config/litellm-and-cline-using-codestral",
 					"provider-config/vscode-language-model-api",

docs/features/auto-compact.mdx

@@ -1,55 +0,0 @@
----
-title: "Automatic Context Summarization"
-sidebarTitle: "Auto Compact"
----
-
-When your conversation approaches the model's context window limit, Cline automatically summarizes it to free up space and keep working. 
-
-## How It Works
-
-Cline monitors token usage during your conversation. When you're getting close to the limit, he:
-
-1. Creates a comprehensive summary of everything that's happened
-2. Preserves all the technical details, code changes, and decisions
-3. Replaces the conversation history with the summary
-4. Continues exactly where he left off
-
-You'll see a summarization tool call when this happens, showing the total cost like any other api call in the chat view.
-
-## Why This Matters
-
-Previously, Cline would truncate older messages when hitting context limits. This meant losing important context from earlier in the conversation.
-
-Now with summarization:
-- All technical decisions and code patterns are preserved
-- File changes and project context remain intact
-- Cline remembers everything he's done
-- You can work on much larger projects without interruption
-
-<Tip>
-Context Summarization synergizes beautifully with [Focus Chain](/features/focus-chain). When Focus Chain is enabled, todo lists persist across summarizations. This means Cline can work on long-horizon tasks that span multiple context windows while staying on track with the todo list guiding him through each reset.
-</Tip>
-
-## Technical Details
-
-The summarization happens through your configured API provider using the same model you're already using. It leverages prompt caching to minimize costs.
-
-1. Cline uses a [summarization prompt](https://github.com/cline/cline/blob/main/src/core/prompts/contextManagement.ts) to request a summary of the conversation.
-
-2. Once the summary is generated, Cline replaces the conversation history with a [continuation prompt](https://github.com/cline/cline/blob/main/src/core/prompts/contextManagement.ts#L69) that asks Cline to keep working and provides the summary as context.
-
-Different models have different context window thresholds for when auto-summarization kicks in. You can see how thresholds are determined in [context-window-utils.ts](https://github.com/cline/cline/blob/main/src/core/context/context-management/context-window-utils.ts). 
-
-## Cost Considerations
-
-Summarization leverages your existing prompt cache from the conversation, so it costs about the same as any other tool call.
-
-Since most input tokens are already cached, you're primarily paying for the summary generation (output tokens), making it very cost-effective.
-
-## Restoring Context with Checkpoints
-
-You can use [checkpoints](/features/checkpoints) to restore your task state from before a summarization occurred. This means you never truly lose context - you can always roll back to previous versions of your conversation. 
-
-<Note>
-  Editing a message before a summarization tool call will work similarly to a checkpoint, allowing you to restore the conversation to that point.
-</Note>

docs/features/drag-and-drop.mdx

@@ -11,4 +11,4 @@ Dragging and dropping workspace files into Cline will automatically create a [fi
 
 ### Supported File Types
 
-Cline supports dragging external images, pdfs, csv, excel, and other text files from your file system, as well as files from your workspace.
+Cline supports dragging external images from your file system, as well as files from your workspace.

docs/features/focus-chain.mdx

@@ -1,296 +0,0 @@
----
-title: "Focus Chain"
-sidebarTitle: "Focus Chain"
----
-
-Focus Chain is a task management enhancement feature in Cline that provides automatic todo list management with real-time progress tracking throughout your tasks. 
-
-This enables Cline to work on long-horizon tasks, seamlessly managing the context sent to LLMs, and keeping Cline on track across many context window resets.
-
-<Tip>
-  Focus Chain works particularly well with Cline's [Deep Planning slash command](/features/slash-commands/deep-planning), providing seamless progress tracking for implementation tasks created through the [planning process](/features/plan-and-act).
-</Tip>
-
-## Key Features
-
-### Automatic Todo List Generation
-
-Cline analyzes your task and automatically creates a comprehensive todo list with:
-- Clear, actionable items in markdown checklist format
-- Logical breakdown of complex tasks into manageable steps
-- Real-time updates as work progresses
-
-### User-Editable Todo Lists
-
-Todo lists are stored as editable markdown files:
-- Direct editing through your preferred markdown editor
-- Automatic detection of changes you make
-- Seamless integration back into Cline's workflow
-- Quick access through the edit button in the task header
-
-### Visual Progress Tracking
-
-The task header displays clear progress indicators:
-- **Step counters** showing current progress (e.g., "3/8")
-- **Completed items** clearly marked with checkmarks
-- **Current work** highlighted with indicators
-- **Expandable view** to see the full todo list
-
-### Smart Reminder System
-
-Configurable reminders ensure todo lists stay current:
-- Default reminder every 6 messages (customizable 1-100)
-- Automatic prompts when switching from Plan Mode to Act Mode
-- User-triggered updates when todo lists are manually edited
-
-
-## Getting Started
-
-<Steps>
-  <Step title="Open Cline Settings">
-      - Click the gear icon in the Cline sidebar
-      - Navigate to the "Features" section
-  </Step>
-  <Step title="Enable Focus Chain">
-      - Check "Enable Focus Chain"
-      - Optionally adjust "Remind Cline Interval" (default: 6 messages)
-  </Step>
-  <Step title="Start a New Task">
-      - Begin a new task
-      - Cline will automatically start creating and managing todo lists
-  </Step>
-</Steps>
-
-| Setting | Default | Range | Description |
-|---------|---------|-------|-------------|
-| Enable Focus Chain | Disabled | On/Off | Enables enhanced task progress tracking |
-| Remind Cline Interval | 6 | 1-100 messages | How often Cline updates the todo list |
-
-## Usage Examples
-
-#### 1. Task Initiation
-
-When you start a new task with Focus Chain enabled:
-
-``` markdown User Request
-User: "Create a user authentication system for my React app"
-
-Cline: [Analyzes request and creates todo list]
-```
-
-#### 2. Todo List Created
-
-Cline creates a comprehensive plan for the task, stored in a markdown file:
-
-```markdown Todo List Created
-- [ ] Set up project structure
-- [ ] Install authentication dependencies
-- [ ] Create user registration component
-- [ ] Implement login functionality
-- [ ] Add password validation
-- [ ] Set up user database schema
-- [ ] Write authentication tests
-- [ ] Deploy to staging environment
-```
-
-#### 3. Progress Tracking
-
-As Cline works, the task header shows real-time progress:
-
-```markdown Todo List Header
-[3/8] Implement login functionality     ⌄
-```
-
-Click to expand and see the full list:
-
-```markdown Full Todo List
-✓ Set up project structure
-✓ Install authentication dependencies
-✓ Create user registration component
-○ Implement login functionality  ← Currently working
-○ Add password validation
-○ Set up user database schema
-○ Write authentication tests
-○ Deploy to staging environment
-```
-
-#### 4. User Editing
-
-Need to tweak the todo list? No problem.
-
-<Steps>
-  <Step title="Open the todo list">
-    Click the edit button in the expanded todo view
-  </Step>
-  <Step title="Edit the markdown file">
-    A markdown file opens in your editor:
-
-    ```markdown Editing Todo List
-    # Focus Chain Todo List for Task abc123
-
-    <!-- Edit this markdown file to update your focus chain todo list -->
-    <!-- Use - [ ] for incomplete items and - [x] for completed items -->
-
-    - [x] Set up project structure
-    - [x] Install authentication dependencies (e.g., Firebase Auth)
-    - [x] Create user registration component
-    - [ ] Implement login functionality
-    - [ ] Add password reset feature
-    - [ ] Set up protected routes
-    - [ ] Implement logout functionality
-    - [ ] Add user profile page
-    - [ ] Write authentication tests
-    - [ ] Deploy to staging environment
-
-    <!-- Save this file to update the task's todo list -->
-    ```
-  </Step>
-  <Step title="Make your changes">
-    Add, remove, or reorder items as needed
-  </Step>
-  <Step title="Save the file">
-    Cline automatically detects and uses your updates
-  </Step>
-</Steps>
-
-## File Structure
-
-### Todo List Storage
-
-Todo lists are stored as markdown files in your task directory:
-
-``` markdown
-<VSCode Global Storage>/
-  tasks/
-    <taskId>/
-      focus_chain_taskid_<taskId>.md
-      ... other task files
-```
-
-### Markdown Format
-
-Todo files use standard markdown checklist syntax:
-
-```markdown Example Todo Syntax
-# Focus Chain Todo List for Task abc123
-
-<!-- Edit this markdown file to update your focus chain todo list -->
-<!-- Use the format: - [ ] for incomplete items and - [x] for completed items -->
-
-- [x] Set up project structure
-- [x] Install authentication dependencies
-- [ ] Create user registration component
-- [ ] Implement login functionality
-- [ ] Add password validation
-- [ ] Set up user database schema
-- [ ] Write authentication tests
-- [ ] Deploy to staging environment
-
-<!-- Save this file and the todo list will be updated in the task -->
-```
-
-
-## Integration with Plan/Act Mode
-
-Focus Chain works seamlessly with Cline's [Plan/Act mode](/features/plan-and-act):
-
-- **Plan Mode**: Optional todo lists for presenting concrete steps
-- **Act Mode**: Automatic todo creation when switching from Plan Mode
-
-<Tip>
-  For complex projects, start in Plan Mode to discuss and refine your approach before switching to Act Mode for implementation.
-</Tip>
-
-## Best Practices
-
-<AccordionGroup>
-  <Accordion title="For Effective Todo Lists">
-    1. **Start with Clear Requests**
-      - Provide detailed initial task descriptions
-      - Include specific requirements and constraints
-      - Mention any preferred technologies or approaches
-
-    2. **Review Generated Lists**
-      - Check that Cline's breakdown aligns with your expectations
-      - Verify that all important steps are included
-      - Ensure the order makes sense for your project
-
-    3. **Edit When Needed**
-      - Add missing steps you identify
-      - Remove unnecessary items
-      - Reorder steps for better workflow
-      - Add more specific details to general items
-  </Accordion>
-  <Accordion title="For Complex Projects">
-    1. **Use Plan Mode First**
-      - Discuss the approach before implementation
-      - Refine requirements through conversation
-      - Switch to Act Mode when ready to begin work
-
-    2. **Break Down Large Tasks**
-      - Split complex projects into smaller, manageable tasks
-      - Create separate todo lists for different components
-      - Focus on one major area at a time
-
-    3. **Regular Reviews**
-      - Check progress periodically during long tasks
-      - Update todo lists as requirements evolve
-      - Communicate changes to Cline through edits
-  </Accordion>
-  <Accordion title="For Collaboration">
-    1. **Share Todo Files**
-      - Todo markdown files can be shared with team members
-      - Include in version control for project documentation
-      - Use as basis for project planning discussions
-
-    2. **Consistent Format**
-      - Follow the standard markdown checklist format
-      - Keep item descriptions clear and actionable
-      - Use consistent terminology across todo lists
-  </Accordion>
-</AccordionGroup>
-
-
-## Troubleshooting
-
-Having issues? Try these quick fixes:
-
-<AccordionGroup>
-  <Accordion title="Todo list not updating?">
-    - Check that Focus Chain is enabled in settings
-    - Focus Chain may not work as well with smaller, less capable models
-    - Ensure file permissions are correct in the task directory
-  </Accordion>
-  <Accordion title="Can't edit todo file?">
-    - Verify your editor supports markdown
-    - Check VSCode has write permissions for the directory
-  </Accordion>
-  <Accordion title="Progress not displaying?">
-    - Ensure todo items use correct syntax (`- [ ]` and `- [x]`)
-    - Verify the markdown file is properly formatted
-  </Accordion>
-</AccordionGroup>
-
-Still stuck? Use the [/reportbug](/features/slash-commands/report-bug) command in Cline to get help.
-
-## Technical Details (for the curious)
-
-<AccordionGroup>
-  <Accordion title="File Monitoring">
-    - Real-time file watching detects changes to todo markdown files
-    - Automatic synchronization between file edits and UI updates
-    - Graceful handling of file creation, modification, and deletion
-  </Accordion>
-  <Accordion title="Progress Calculation">
-    - Dynamic counting of completed vs. total todo items
-    - Support for both `- [x]` and `- [X]` completion syntax
-    - Unicode symbols (✓, ○) for enhanced visual display
-  </Accordion>
-  <Accordion title="Privacy Considerations">
-    - Todo lists stored locally in VSCode workspace
-    - No todo content transmitted to external services
-    - Usage telemetry (can be disabled in settings)
-  </Accordion>
-</AccordionGroup>
-
-Focus Chain turns Cline into your personal project manager, keeping you on track and your tasks organized. Give it a try on your next project!

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

@@ -1,153 +0,0 @@
----
-title: "Deep Planning Command"
-sidebarTitle: "/deep-planning"
----
-
-`/deep-planning` transforms Cline into a meticulous architect who investigates your codebase, asks clarifying questions, and creates a comprehensive implementation plan before writing a single line of code. 
-
-When you use `/deep-planning`, Cline follows a four-step process that mirrors how senior developers approach complex features: thorough investigation, discussion & clarification of requirements, detailed planning, and structured task creation with progress tracking.
-
-## The Four-Step Process
-
-### Step 1: Silent Investigation
-
-Cline becomes a detective, silently exploring your codebase to understand its structure, patterns, and constraints. He examines source files, analyzes import patterns, discovers class hierarchies, and identifies technical debt markers. No commentary, no narration - just focused research.
-
-During this phase, Cline runs commands like:
-- Finding all class and function definitions across your codebase
-- Analyzing import patterns to understand dependencies
-- Discovering project structure and file organization
-- Identifying TODOs and technical debt
-
-### Step 2: Discussion and Questions
-
-Once Cline understands your codebase, he asks targeted questions that will shape the implementation. These aren't generic questions - they're specific to your project and the feature you're building.
-
-Questions might cover:
-- Clarifying ambiguous requirements
-- Choosing between equally valid implementation approaches
-- Confirming assumptions about system behavior
-- Understanding preferences for technical decisions
-
-### Step 3: Implementation Plan Document
-
-Cline creates a structured markdown document (`implementation_plan.md`) that serves as your implementation blueprint. This isn't a vague outline - it's a detailed specification with exact file paths, function signatures, and implementation order.
-
-The plan includes eight comprehensive sections:
-- **Overview**: The goal and high-level approach
-- **Types**: Complete type definitions and data structures
-- **Files**: Exact files to create, modify, or delete
-- **Functions**: New and modified functions with signatures
-- **Classes**: Class modifications and inheritance details
-- **Dependencies**: Package requirements and versions
-- **Testing**: Validation strategies and test requirements
-- **Implementation Order**: Step-by-step execution sequence
-
-### Step 4: Implementation Task Creation
-
-Cline creates a new task that references the plan document and includes trackable implementation steps. The task comes with specific commands to read each section of the plan, ensuring the implementing agent (whether that's you or Cline in Act Mode) can navigate the blueprint efficiently.
-
-<Tip>
-  Deep Planning works beautifully with [Focus Chain](/features/focus-chain). The implementation steps automatically become a todo list with real-time progress tracking, keeping complex projects organized and on track.
-</Tip>
-
-## Using Deep Planning
-
-Start a deep planning session by typing `/deep-planning` followed by your feature description:
-
-```
-/deep-planning Add user authentication with JWT tokens and role-based access control
-```
-
-Cline will begin his investigation immediately. You'll see him reading files and running commands to understand your codebase. Once he's gathered enough context, he'll engage you in discussion before creating the plan.
-
-## Example Workflow
-
-Here's how I use `/deep-planning` for a real feature:
-
-<Steps>
-  <Step title="Initiate Planning">
-    I type `/deep-planning implement a caching layer for API responses`
-  </Step>
-  <Step title="Silent Investigation">
-    Cline explores my codebase, examining:
-    - Current API structure and endpoints
-    - Existing data flow patterns
-    - Database queries and performance bottlenecks
-    - Configuration and environment setup
-  </Step>
-  <Step title="Targeted Discussion">
-    Cline asks me:
-    - "Should we use Redis or in-memory caching?"
-    - "What's the acceptable cache staleness for user data?"
-    - "Do you need cache invalidation webhooks?"
-  </Step>
-  <Step title="Plan Creation">
-    Cline generates `implementation_plan.md` with:
-    - Cache service class specifications
-    - Redis connection configuration
-    - Modified API endpoints with caching logic
-    - Cache key generation strategies
-    - TTL configurations for different data types
-  </Step>
-  <Step title="Task Generation">
-    Cline creates a new task with:
-    - Reference to the implementation plan
-    - Commands to read specific sections
-    - Trackable todo items for each implementation step
-    - Request to switch to Act Mode for execution
-  </Step>
-</Steps>
-
-## Integration with Plan/Act Mode
-
-Deep Planning is designed to work seamlessly with [Plan/Act Mode](/features/plan-and-act):
-
-- Use `/deep-planning` in Plan Mode for the investigation and planning phases
-- The generated task requests switching to Act Mode for implementation
-- Focus Chain automatically tracks progress through the implementation steps
-
-This separation ensures planning stays focused on architecture while implementation stays focused on execution.
-
-## Best Practices
-
-### When to Use Deep Planning
-
-Use `/deep-planning` for:
-- Features touching multiple parts of your codebase
-- Architectural changes requiring careful coordination
-- Complex integrations with external services
-- Refactoring efforts that need systematic execution
-- Any feature where you'd normally spend time whiteboarding
-
-### Making the Most of Investigation
-
-Let Cline complete his investigation thoroughly. The quality of the plan directly correlates with how well he understands your codebase. If you have specific areas he should examine, mention them in your initial request.
-
-### Reviewing the Plan
-
-Always review `implementation_plan.md` before starting implementation. The plan is comprehensive but not immutable - you can edit it directly if needed. Think of it as a collaborative document between you and Cline.
-
-### Tracking Progress
-
-With Focus Chain enabled, your implementation progress displays in the task header. Each completed step gets checked off automatically as Cline works through the plan, giving you real-time visibility into complex implementations.
-
-## Inspiration
-
-I use `/deep-planning` whenever I'm about to build something that would normally require a design document. Recent examples from my workflow:
-
-- **Migrating authentication systems**: Deep Planning mapped every endpoint, identified all authentication touchpoints, and created a migration plan that avoided breaking changes.
-
-- **Adding real-time features**: The plan covered WebSocket integration, event handling, state synchronization, and fallback mechanisms for disconnections.
-
-- **Database schema refactoring**: Cline identified all affected queries, created migration scripts, and planned the rollout to minimize downtime.
-
-- **API versioning implementation**: The plan detailed route changes, backward compatibility layers, deprecation notices, and client migration paths.
-
-The power of `/deep-planning` is that it forces thoughtful architecture before implementation. It's like having a senior developer review your approach before you write code, except that developer has perfect knowledge of your entire codebase.
-
-<Note>
-  Deep Planning requires models with strong reasoning capabilities. It works best with the latest generation of models, like GPT-5, Claude 4, Gemini 2.5, or Grok 4. Smaller models may struggle with the comprehensive analysis required.
-</Note>
-
-For simpler tasks that don't require extensive planning, consider using [/newtask](/features/slash-commands/new-task) to create focused tasks with context, or jump straight into implementation if the path forward is clear.

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

@@ -55,8 +55,7 @@ Think of context like a whiteboard you and Cline share:
     -   Each model has a fixed size:
         -   Claude 3.5 Sonnet: 200,000 tokens
         -   DeepSeek: 64,000 tokens
-    -   When the whiteboard is full, Cline automatically summarizes the conversation to free up space
-        -   [Learn about Automatic Context Summarization](/features/automatic-context-summarization)
+    -   When the whiteboard is full, you need to erase (clear context) to write more
         -   [How Cline manages context under the hood](https://cline.bot/blog/understanding-the-new-context-window-progress-bar-in-cline)
 
 ⚠️ **Important**: Having a large context window (like Claude's 200k tokens) doesn't mean you should fill it completely. Just like a cluttered whiteboard, too much information can make it harder to focus on what's important.
@@ -86,28 +85,7 @@ Cline provides a visual way to monitor your context window usage through a progr
 -   Before starting complex tasks
 -   When Cline seems to lose context
 
-💡 **Tip**: With [Automatic Context Summarization](/features/automatic-context-summarization), Cline can now handle long conversations automatically. When combined with [Focus Chain](/features/focus-chain), you can work on complex projects that span multiple context windows without losing progress.
-
-## Automatic Context Management
-
-Cline now includes intelligent features to manage context automatically:
-
-### Automatic Context Summarization
-
-When your conversation approaches the context window limit, Cline automatically:
-- Creates a comprehensive summary of the conversation
-- Preserves all essential technical details and decisions
-- Seamlessly continues work without interruption
-- Maintains full task continuity
-
-This means you can work on larger projects without manually managing context. [Learn more about Automatic Context Summarization](/features/automatic-context-summarization).
-
-### Focus Chain Integration
-
-When combined with [Focus Chain](/features/focus-chain), Cline can:
-- Maintain todo lists across context resets
-- Track progress through multiple summarizations
-- Keep working on long-horizon tasks without losing direction
+💡 **Tip**: Consider starting a fresh session when usage reaches 70-80% to maintain optimal performance.
 
 ## Working with Context Files
 

docs/provider-config/anthropic.mdx

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

docs/provider-config/aws-bedrock-with-apikey-authentication.mdx

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

docs/provider-config/aws-bedrock-with-credentials-authentication.mdx

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

docs/provider-config/aws-bedrock-with-profile-authentication.mdx

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

docs/provider-config/claude-code.mdx

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

docs/provider-config/openai-compatible.mdx

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

docs/provider-config/openai.mdx

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

docs/provider-config/requesty.mdx

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

esbuild.mjs

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

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

@@ -1,171 +0,0 @@
-const { RuleTester: StateApiRuleTester } = require("eslint")
-const noDirectVscodeStateApiRule = require("../no-direct-vscode-state-api")
-
-const stateApiRuleTester = new StateApiRuleTester({
-	parser: require.resolve("@typescript-eslint/parser"),
-	parserOptions: {
-		ecmaVersion: 2020,
-		sourceType: "module",
-		ecmaFeatures: {
-			jsx: true,
-		},
-	},
-})
-
-stateApiRuleTester.run("no-direct-vscode-state-api", noDirectVscodeStateApiRule, {
-	valid: [
-		// Should allow state APIs in CacheService.ts
-		{
-			code: `await context.globalState.update("myKey", value);`,
-			filename: "CacheService.ts",
-		},
-		{
-			code: `const value = context.globalState.get("myKey");`,
-			filename: "/src/core/storage/CacheService.ts",
-		},
-		{
-			code: `await context.secrets.store("apiKey", value);`,
-			filename: "CacheService.ts",
-		},
-		// Should allow state APIs in state-helpers.ts
-		{
-			code: `const value = context.globalState.get("myKey");`,
-			filename: "state-helpers.ts",
-		},
-		{
-			code: `await context.secrets.get("apiKey");`,
-			filename: "/src/core/storage/utils/state-helpers.ts",
-		},
-		// Should allow state APIs in state-migrations.ts
-		{
-			code: `await context.globalState.update("myKey", value);`,
-			filename: "state-migrations.ts",
-		},
-		{
-			code: `const value = context.workspaceState.get("myKey");`,
-			filename: "/src/core/storage/state-migrations.ts",
-		},
-		// Should allow state APIs in extension.ts
-		{
-			code: `const distinctId = context.globalState.get<string>("cline.distinctId");`,
-			filename: "extension.ts",
-		},
-		{
-			code: `await context.globalState.update("clineVersion", currentVersion);`,
-			filename: "/src/extension.ts",
-		},
-		{
-			code: `const secret = await context.secrets.get("clineAccountId");`,
-			filename: "extension.ts",
-		},
-		// Should allow state APIs in test files
-		{
-			code: `context.globalState.get("testKey")`,
-			filename: "/foo/bar.test.ts",
-		},
-		// Should allow non-state API calls
-		{
-			code: `const value = someOtherObject.globalState.get("myKey");`,
-			filename: "some-file.ts",
-		},
-		{
-			code: `await myContext.secrets.store("key", "value");`,
-			filename: "some-file.ts",
-		},
-	],
-	invalid: [
-		// Should disallow context.globalState.get
-		{
-			code: `const value = context.globalState.get("myKey");`,
-			filename: "some-file.ts",
-			errors: [
-				{
-					messageId: "useCacheServiceGlobalGet",
-				},
-			],
-		},
-		// Should disallow context.globalState.update
-		{
-			code: `await context.globalState.update("myKey", "myValue");`,
-			filename: "some-file.ts",
-			errors: [
-				{
-					messageId: "useCacheServiceGlobalSet",
-				},
-			],
-		},
-		// Should disallow context.workspaceState.get
-		{
-			code: `const value = context.workspaceState.get("myKey");`,
-			filename: "some-file.ts",
-			errors: [
-				{
-					messageId: "useCacheServiceWorkspaceGet",
-				},
-			],
-		},
-		// Should disallow context.workspaceState.update
-		{
-			code: `await context.workspaceState.update("myKey", "myValue");`,
-			filename: "some-file.ts",
-			errors: [
-				{
-					messageId: "useCacheServiceWorkspaceSet",
-				},
-			],
-		},
-		// Should disallow context.secrets.get
-		{
-			code: `const secret = await context.secrets.get("apiKey");`,
-			filename: "some-file.ts",
-			errors: [
-				{
-					messageId: "useCacheServiceSecretsGet",
-				},
-			],
-		},
-		// Should disallow context.secrets.store
-		{
-			code: `await context.secrets.store("apiKey", "secret-value");`,
-			filename: "some-file.ts",
-			errors: [
-				{
-					messageId: "useCacheServiceSecretsSet",
-				},
-			],
-		},
-		// Should disallow context.secrets.delete
-		{
-			code: `await context.secrets.delete("apiKey");`,
-			filename: "some-file.ts",
-			errors: [
-				{
-					messageId: "useCacheServiceSecretsSet",
-				},
-			],
-		},
-		// Should disallow chained state API calls
-		{
-			code: `const value = await context.globalState.get("key") || "default";`,
-			filename: "some-file.ts",
-			errors: [
-				{
-					messageId: "useCacheServiceGlobalGet",
-				},
-			],
-		},
-		// Should disallow state API calls in Promise.all
-		{
-			code: `await Promise.all([context.secrets.get("key1"), context.secrets.get("key2")]);`,
-			filename: "some-file.ts",
-			errors: [
-				{
-					messageId: "useCacheServiceSecretsGet",
-				},
-				{
-					messageId: "useCacheServiceSecretsGet",
-				},
-			],
-		},
-	],
-})

eslint-rules/index.js

@@ -1,18 +1,15 @@
 // eslint-rules/index.js
 const noDirectVscodeApi = require("./no-direct-vscode-api")
-const noDirectVscodeStateApi = require("./no-direct-vscode-state-api")
 
 module.exports = {
 	rules: {
 		"no-direct-vscode-api": noDirectVscodeApi,
-		"no-direct-vscode-state-api": noDirectVscodeStateApi,
 	},
 	configs: {
 		recommended: {
 			plugins: ["local"],
 			rules: {
 				"local/no-direct-vscode-api": "warn",
-				"local/no-direct-vscode-state-api": "error",
 			},
 		},
 	},

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

@@ -1,144 +0,0 @@
-const { ESLintUtils } = require("@typescript-eslint/utils")
-const path = require("path")
-
-const createRule = ESLintUtils.RuleCreator((name) => `https://cline.bot/eslint-rules/${name}`)
-
-// Configuration for context-based state APIs
-const disallowedContextApis = {
-	"globalState.get": {
-		messageId: "useCacheServiceGlobalGet",
-	},
-	"globalState.update": {
-		messageId: "useCacheServiceGlobalSet",
-	},
-	"workspaceState.get": {
-		messageId: "useCacheServiceWorkspaceGet",
-	},
-	"workspaceState.update": {
-		messageId: "useCacheServiceWorkspaceSet",
-	},
-	"secrets.get": {
-		messageId: "useCacheServiceSecretsGet",
-	},
-	"secrets.store": {
-		messageId: "useCacheServiceSecretsSet",
-	},
-	"secrets.delete": {
-		messageId: "useCacheServiceSecretsSet",
-	},
-}
-
-module.exports = createRule({
-	name: "no-direct-vscode-state-api",
-	meta: {
-		type: "problem",
-		docs: {
-			description:
-				"Disallow direct VSCode state API usage (context.globalState, context.workspaceState, context.secrets) in favor of CacheService",
-			recommended: "error",
-		},
-		messages: {
-			useCacheServiceGlobalGet:
-				"Use CacheService.getGlobalStateKey() instead of context.globalState.get().\n" +
-				"The CacheService provides fast in-memory access with automatic persistence.\n" +
-				"Example: cacheService.getGlobalStateKey('myKey') instead of context.globalState.get('myKey').\n" +
-				"Found: {{code}}",
-			useCacheServiceGlobalSet:
-				"Use CacheService.setGlobalState() instead of context.globalState.update().\n" +
-				"The CacheService provides immediate updates with debounced persistence.\n" +
-				"Example: cacheService.setGlobalState('myKey', value) instead of context.globalState.update('myKey', value).\n" +
-				"Found: {{code}}",
-			useCacheServiceWorkspaceGet:
-				"Use CacheService.getWorkspaceStateKey() instead of context.workspaceState.get().\n" +
-				"The CacheService provides fast in-memory access with automatic persistence.\n" +
-				"Example: cacheService.getWorkspaceStateKey('myKey') instead of context.workspaceState.get('myKey').\n" +
-				"Found: {{code}}",
-			useCacheServiceWorkspaceSet:
-				"Use CacheService.setWorkspaceState() instead of context.workspaceState.update().\n" +
-				"The CacheService provides immediate updates with debounced persistence.\n" +
-				"Example: cacheService.setWorkspaceState('myKey', value) instead of context.workspaceState.update('myKey', value).\n" +
-				"Found: {{code}}",
-			useCacheServiceSecretsGet:
-				"Use CacheService.getSecretKey() instead of context.secrets.get().\n" +
-				"The CacheService provides fast in-memory access with automatic persistence.\n" +
-				"Example: cacheService.getSecretKey('mySecret') instead of context.secrets.get('mySecret').\n" +
-				"Found: {{code}}",
-			useCacheServiceSecretsSet:
-				"Use CacheService.setSecret() instead of context.secrets.store() or context.secrets.delete().\n" +
-				"The CacheService provides immediate updates with debounced persistence.\n" +
-				"Example: cacheService.setSecret('mySecret', value) instead of context.secrets.store('mySecret', value).\n" +
-				"For deletion, use: cacheService.setSecret('mySecret', undefined).\n" +
-				"Found: {{code}}",
-		},
-		schema: [],
-	},
-	defaultOptions: [],
-
-	create(context) {
-		function isExcluded(filename) {
-			// Skip checking test files
-			if (filename.endsWith(".test.ts")) {
-				return true
-			}
-			// Skip checking specific state-related files that need direct access
-			const basename = path.basename(filename)
-			if (
-				basename === "CacheService.ts" ||
-				basename === "state-helpers.ts" ||
-				basename === "state-migrations.ts" ||
-				basename === "extension.ts" ||
-				basename === "common.ts" // CI might report errors from this virtual file
-			) {
-				return true
-			}
-			return false
-		}
-
-		// Check for context-based state API calls
-		function checkContextStateApi(node) {
-			if (isExcluded(context.filename)) {
-				return
-			}
-
-			// Check if this is a member expression like context.globalState.get
-			if (
-				node.type === "MemberExpression" &&
-				node.object &&
-				node.object.type === "MemberExpression" &&
-				node.object.object &&
-				node.object.object.type === "Identifier" &&
-				node.object.object.name === "context"
-			) {
-				const stateType = node.object.property.name // e.g., "globalState", "workspaceState", "secrets"
-				const method = node.property.name // e.g., "get", "update", "store", "delete"
-				const apiPath = `${stateType}.${method}`
-
-				if (disallowedContextApis[apiPath]) {
-					// For method calls, get the whole call expression
-					let reportNode = node
-					let parentNode = context.sourceCode.getAncestors(node).pop()
-					if (parentNode && parentNode.type === "CallExpression" && parentNode.callee === node) {
-						reportNode = parentNode
-					}
-
-					const callText = context.sourceCode.getText(reportNode).trim()
-
-					context.report({
-						node: reportNode,
-						messageId: disallowedContextApis[apiPath].messageId,
-						data: {
-							code: callText,
-						},
-					})
-				}
-			}
-		}
-
-		return {
-			// Detect member expressions (e.g., context.globalState.get)
-			MemberExpression(node) {
-				checkContextStateApi(node)
-			},
-		}
-	},
-})

evals/diff-edits/ClineWrapper.ts

@@ -1,9 +1,12 @@
 import { OpenRouterHandler } from "../../src/api/providers/openrouter"
 import { OpenAiNativeHandler } from "../../src/api/providers/openai-native"
+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 constructNewFileContent_06_06_25 } from "./diff-apply/diff-06-06-25"
@@ -15,7 +18,9 @@ type ParseAssistantMessageFn = (message: string) => AssistantMessageContent[]
 type ConstructNewFileContentFn = (diff: string, original: string, strict: boolean) => Promise<string | any>
 
 const parsingFunctions: Record<string, ParseAssistantMessageFn> = {
+	parseAssistantMessageV1: parseAssistantMessageV1,
 	parseAssistantMessageV2: parseAssistantMessageV2,
+	parseAssistantMessageV3: parseAssistantMessageV3,
 }
 
 const diffEditingFunctions: Record<string, ConstructNewFileContentFn> = {

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

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

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": "3.24.0",
+	"version": "3.20.8",
 	"icon": "assets/icons/icon.png",
 	"engines": {
 		"vscode": "^1.84.0"
@@ -362,8 +362,8 @@
 		"test:unit": "TS_NODE_PROJECT='./tsconfig.unit-test.json' mocha",
 		"test:coverage": "vscode-test --coverage",
 		"e2e": "playwright test -c playwright.config.ts",
-		"test:e2e": "playwright install && vsce package --no-dependencies --allow-package-secrets sendgrid --out dist/e2e.vsix && node src/test/e2e/utils/build.mjs && playwright test",
-		"test:e2e:optimal": "vsce package --no-dependencies --allow-package-secrets sendgrid --out dist/e2e.vsix && node src/test/e2e/utils/build.mjs && playwright test",
+		"test:e2e": "playwright install && vsce package --no-dependencies --out dist/e2e.vsix && node src/test/e2e/utils/build.js && playwright test",
+		"test:e2e:optimal": "vsce package --no-dependencies --out dist/e2e.vsix && node src/test/e2e/utils/build.js && playwright test",
 		"install:all": "npm install && cd webview-ui && npm install",
 		"dev:webview": "cd webview-ui && npm run dev",
 		"build:webview": "cd webview-ui && npm run build",
@@ -447,7 +447,6 @@
 		"@playwright/test": "^1.53.2",
 		"@sentry/browser": "^9.12.0",
 		"@streamparser/json": "^0.0.22",
-		"@types/uuid": "^10.0.0",
 		"@vscode/codicons": "^0.0.36",
 		"archiver": "^7.0.1",
 		"axios": "^1.8.2",
@@ -492,8 +491,6 @@
 		"tree-sitter-wasms": "^0.1.11",
 		"ts-morph": "^25.0.1",
 		"turndown": "^7.2.0",
-		"ulid": "^2.4.0",
-		"uuid": "^11.1.0",
 		"vscode-uri": "^3.1.0",
 		"web-tree-sitter": "^0.22.6",
 		"zod": "^3.24.2"

playwright.config.ts

@@ -6,18 +6,13 @@ const isWindow = process?.platform?.startsWith("win")
 export default defineConfig({
 	workers: 1,
 	retries: 1,
-	forbidOnly: isCI,
 	testDir: "src/test/e2e",
-	testMatch: /.*\.test\.ts/,
 	timeout: isCI || isWindow ? 40000 : 20000,
 	expect: {
 		timeout: isCI || isWindow ? 5000 : 2000,
 	},
 	fullyParallel: true,
 	reporter: isCI ? [["github"], ["list"]] : [["list"]],
-	use: {
-		video: "retain-on-failure",
-	},
 	projects: [
 		{
 			name: "setup test environment",
@@ -25,7 +20,13 @@ export default defineConfig({
 		},
 		{
 			name: "e2e tests",
+			testMatch: /.*\.test\.ts/,
 			dependencies: ["setup test environment"],
 		},
+		{
+			name: "cleanup test environment",
+			testMatch: /global\.teardown\.ts/,
+			dependencies: ["e2e tests"],
+		},
 	],
 })

proto/cline/account.proto

@@ -36,8 +36,6 @@ service AccountService {
   rpc getUserOrganizations(EmptyRequest) returns (UserOrganizationsResponse);
 
   rpc setUserOrganization(UserOrganizationUpdateRequest) returns (Empty);
-
-  rpc openrouterAuthClicked(EmptyRequest) returns (Empty);
 }
 
 message AuthStateChangedRequest {

proto/cline/browser.proto

@@ -42,7 +42,6 @@ message BrowserSettings {
   optional bool remote_browser_enabled = 3;
   optional string chrome_executable_path = 4;
   optional bool disable_tool_use = 5;
-  optional string custom_args = 6;
 }
 
 message UpdateBrowserSettingsRequest {
@@ -52,5 +51,4 @@ message UpdateBrowserSettingsRequest {
   optional bool remote_browser_enabled = 4;
   optional string chrome_executable_path = 5;
   optional bool disable_tool_use = 6;
-  optional string custom_args = 7;
 }

proto/cline/common.proto

@@ -55,11 +55,6 @@ message Boolean {
   bool value = 1;
 }
 
-// the same as Boolean, but avoiding name conflicts
-message BooleanResponse {
-  bool value = 1;
-}
-
 message StringArray {
   repeated string values = 1;
 }
@@ -73,32 +68,3 @@ message KeyValuePair {
   string key = 1;
   string value = 2;
 }
-
-message FileDiagnostics {
-  string file_path = 1;
-  repeated Diagnostic diagnostics = 2;
-}
-
-message Diagnostic {
-  string message = 1;
-  DiagnosticRange range = 2;
-  DiagnosticSeverity severity = 3;
-  optional string source = 4;
-}
-
-message DiagnosticRange {
-  DiagnosticPosition start = 1;
-  DiagnosticPosition end = 2;
-}
-
-message DiagnosticPosition {
-  int32 line = 1;
-  int32 character = 2;
-}
-
-enum DiagnosticSeverity {
-  DIAGNOSTIC_ERROR = 0;
-  DIAGNOSTIC_WARNING = 1;
-  DIAGNOSTIC_INFORMATION = 2;
-  DIAGNOSTIC_HINT = 3;
-}

proto/cline/file.proto

@@ -55,14 +55,8 @@ service FileService {
   // Toggles a workflow on or off
   rpc toggleWorkflow(ToggleWorkflowRequest) returns (ClineRulesToggles);
 
-  // Check if file exists in the project
-  rpc ifFileExistsRelativePath(StringRequest) returns (BooleanResponse);
-
-  // Open a file in editor by a relative path
-  rpc openFileRelativePath(StringRequest) returns (Empty);
-  
-  // Opens or creates a focus chain checklist markdown file for editing
-  rpc openFocusChainFile(StringRequest) returns (Empty);
+  // Subscribe to workspace file updates
+  rpc subscribeToWorkspaceUpdates(EmptyRequest) returns (stream StringArray);
 }
 
 // Response for refreshRules operation
@@ -93,19 +87,12 @@ message RelativePaths {
   repeated string paths = 1;
 }
 
-// Enum for file search type filtering
-enum FileSearchType {
-  FILE = 0;
-  FOLDER = 1;
-}
-
 // Request for file search operations
 message FileSearchRequest {
   Metadata metadata = 1;
   string query = 2;                  // Search query string
   optional string mentions_request_id = 3;  // Optional request ID for tracking requests
   optional int32 limit = 4;          // Optional limit for results (default: 20)
-  optional FileSearchType selected_type = 5;  // Optional selected type filter
 }
 
 // Result for file search operations

proto/cline/models.proto

@@ -25,10 +25,8 @@ service ModelsService {
   rpc subscribeToOpenRouterModels(EmptyRequest) returns (stream OpenRouterCompatibleModelInfo);
   // Updates API configuration
   rpc updateApiConfigurationProto(UpdateApiConfigurationRequest) returns (Empty);
-  // Refreshes and returns Groq models
+   // Refreshes and returns Groq models
   rpc refreshGroqModels(EmptyRequest) returns (OpenRouterCompatibleModelInfo);
-  // Refreshes and returns Baseten models
-  rpc refreshBasetenModels(EmptyRequest) returns (OpenRouterCompatibleModelInfo);
 }
 
 // List of VS Code LM models
@@ -132,7 +130,6 @@ enum ApiProvider {
   MOONSHOT = 27;
   HUGGINGFACE = 28;
   HUAWEI_CLOUD_MAAS = 29;
-  BASETEN = 30;
 }
 
 // Model info for OpenAI-compatible models
@@ -175,7 +172,7 @@ message ModelsApiConfiguration {
   // Global configuration fields (not mode-specific)
   optional string api_key = 1;
   optional string cline_api_key = 2;
-  optional string ulid = 3;
+  optional string task_id = 3;
   optional string lite_llm_base_url = 4;
   optional string lite_llm_api_key = 5;
   optional bool lite_llm_use_prompt_cache = 6;
@@ -234,8 +231,6 @@ message ModelsApiConfiguration {
   optional string groq_api_key = 59;
   optional string hugging_face_api_key = 60;
   optional string huawei_cloud_maas_api_key = 61;
-  optional string baseten_api_key = 62;
-  optional string ollama_api_key = 63;
 
   // Plan mode configurations
   optional ApiProvider plan_mode_api_provider = 100;
@@ -264,8 +259,6 @@ message ModelsApiConfiguration {
   optional OpenRouterModelInfo plan_mode_hugging_face_model_info = 123;
   optional string plan_mode_huawei_cloud_maas_model_id = 124;
   optional OpenRouterModelInfo plan_mode_huawei_cloud_maas_model_info = 125; 
-  optional string plan_mode_baseten_model_id = 126;
-  optional OpenRouterModelInfo plan_mode_baseten_model_info = 127;
 
   // Act mode configurations
   optional ApiProvider act_mode_api_provider = 200;
@@ -294,8 +287,6 @@ message ModelsApiConfiguration {
   optional OpenRouterModelInfo act_mode_hugging_face_model_info = 223;
   optional string act_mode_huawei_cloud_maas_model_id = 224;
   optional OpenRouterModelInfo act_mode_huawei_cloud_maas_model_info = 225;
-  optional string act_mode_baseten_model_id = 226;
-  optional OpenRouterModelInfo act_mode_baseten_model_info = 227;
 
   repeated string favorited_model_ids = 300;
 }

proto/cline/state.proto

@@ -52,18 +52,6 @@ enum PlanActMode {
   ACT = 1;
 }
 
-enum OpenaiReasoningEffort {
-  LOW = 0;
-  MEDIUM = 1;
-  HIGH = 2;
-}
-
-enum McpDisplayMode {
-  RICH = 0;
-  PLAIN = 1;
-  MARKDOWN = 2;
-}
-
 message ChatContent {
   optional string message = 1;
   repeated string images = 2;
@@ -117,13 +105,11 @@ message UpdateSettingsRequest {
   optional int32 shell_integration_timeout = 8;
   optional bool terminal_reuse_enabled = 9;
   optional bool mcp_responses_collapsed = 10;
-  optional McpDisplayMode mcp_display_mode = 11;
+  optional string mcp_display_mode = 11;
   optional int32 terminal_output_line_limit = 12;
   optional PlanActMode mode = 13;
   optional string preferred_language = 14;
-  optional OpenaiReasoningEffort openai_reasoning_effort = 15;
-  optional bool strict_plan_mode_enabled = 16;
-  optional FocusChainSettings focus_chain_settings = 17;
+  optional string openai_reasoning_effort = 15;
 }
 
 // Complete API Configuration message
@@ -131,7 +117,7 @@ message ApiConfiguration {
   // Global configuration fields (not mode-specific)
   optional string api_key = 1; // anthropic
   optional string cline_api_key = 2;
-  optional string ulid = 3;
+  optional string task_id = 3;
   optional string lite_llm_base_url = 4;
   optional string lite_llm_api_key = 5;
   optional bool lite_llm_use_prompt_cache = 6;
@@ -185,7 +171,6 @@ message ApiConfiguration {
   optional string moonshot_api_key = 54;
   optional string moonshot_api_line = 55;
   optional string huawei_cloud_maas_api_key = 56;
-  optional string ollama_api_key = 57;
   
   // Plan mode configurations
   optional string plan_mode_api_provider = 100;
@@ -249,11 +234,6 @@ message UpdateTerminalConnectionTimeoutRequest {
   optional int32 timeout_ms = 1;
 }
 
-message FocusChainSettings {
-  bool enabled = 1;
-  int32 remind_cline_interval = 2;
-}
-
 message UpdateTerminalConnectionTimeoutResponse {
   optional int32 timeout_ms = 1;
 }

proto/cline/ui.proto

@@ -41,7 +41,6 @@ enum ClineAsk {
   NEW_TASK = 13;
   CONDENSE = 14;
   REPORT_BUG = 15;
-  SUMMARIZE_TASK = 16;
 }
 
 // Enum for ClineSay types
@@ -73,7 +72,6 @@ enum ClineSay {
   CHECKPOINT_CREATED = 24;
   LOAD_MCP_DOCUMENTATION = 25;
   INFO = 26;
-  TASK_PROGRESS = 27;
 }
 
 // Enum for ClineSayTool tool types
@@ -229,7 +227,7 @@ service UiService {
   rpc onDidShowAnnouncement(EmptyRequest) returns (Boolean);
   
   // Subscribe to addToInput events (when user adds content via context menu)
-  rpc subscribeToAddToInput(StringRequest) returns (stream String);
+  rpc subscribeToAddToInput(EmptyRequest) returns (stream String);
   
   // Subscribe to MCP button clicked events
   rpc subscribeToMcpButtonClicked(WebviewProviderTypeRequest) returns (stream Empty);
@@ -269,7 +267,4 @@ service UiService {
   
   // Opens a URL in the default browser
   rpc openUrl(StringRequest) returns (Empty);
-  
-  // Opens the Cline walkthrough
-  rpc openWalkthrough(EmptyRequest) returns (Empty);
 }

proto/host/diff.proto

@@ -10,28 +10,17 @@ import "cline/common.proto";
 service DiffService {
   // Open the diff view/editor.
   rpc openDiff(OpenDiffRequest) returns (OpenDiffResponse);
-
   // Get the contents of the diff view.
   rpc getDocumentText(GetDocumentTextRequest) returns (GetDocumentTextResponse);
-
   // Replace a text selection in the diff.
   rpc replaceText(ReplaceTextRequest) returns (ReplaceTextResponse);
-
   rpc scrollDiff(ScrollDiffRequest) returns (ScrollDiffResponse);
-
   // Truncate the diff document.
   rpc truncateDocument(TruncateDocumentRequest) returns (TruncateDocumentResponse);
-
   // Save the diff document.
   rpc saveDocument(SaveDocumentRequest) returns (SaveDocumentResponse);
-
-  // Close all the diff editor windows/tabs.
-  // Any diff editors with unsaved content should not be closed.
-  rpc closeAllDiffs(CloseAllDiffsRequest) returns (CloseAllDiffsResponse);
-
-  // Display a diff view comparing before/after states for multiple files.
-  // Content is passed as in-memory data, not read from the file system.
-  rpc openMultiFileDiff(OpenMultiFileDiffRequest) returns (OpenMultiFileDiffResponse);
+  // Close the diff editor UI.
+  rpc closeDiff(CloseDiffRequest) returns (CloseDiffResponse);
 }
 
 message OpenDiffRequest {
@@ -81,9 +70,12 @@ message TruncateDocumentRequest {
 
 message TruncateDocumentResponse {}
 
-message CloseAllDiffsRequest {}
+message CloseDiffRequest {
+  optional cline.Metadata metadata = 1;
+  optional string diff_id = 2;
+}
 
-message CloseAllDiffsResponse {}
+message CloseDiffResponse {}
 
 message SaveDocumentRequest {
   optional cline.Metadata metadata = 1;
@@ -91,17 +83,3 @@ message SaveDocumentRequest {
 }
 
 message SaveDocumentResponse {}
-
-message OpenMultiFileDiffRequest {
-  optional string title = 1;
-  repeated ContentDiff diffs = 2;
-}
-
-message ContentDiff {
-  // The absolute file path.
-  optional string file_path = 1;
-  optional string left_content = 2;
-  optional string right_content = 3;
-}
-
-message OpenMultiFileDiffResponse {}

proto/host/testing.proto

@@ -1,17 +0,0 @@
-syntax = "proto3";
-
-package host;
-option java_package = "bot.cline.host.proto";
-option java_multiple_files = true;
-
-// This is for use in integration tests to get the contents of the webview.
-service TestingService {
-  rpc getWebviewHtml(GetWebviewHtmlRequest) returns (GetWebviewHtmlResponse);
-}
-
-message GetWebviewHtmlRequest {
-}
-
-message GetWebviewHtmlResponse {
-  optional string html = 1;
-}

proto/host/window.proto

@@ -84,8 +84,6 @@ message ShowSaveDialogRequest {
 
 message ShowSaveDialogOptions {
   optional string default_path = 1;
-  // A map of file types to extensions, e.g
-  // "Text Files": { "extensions": ["txt", "md"] }
   map<string, FileExtensionList> filters = 2;
 }
 
@@ -94,7 +92,6 @@ message FileExtensionList {
 }
 
 message ShowSaveDialogResponse {
-  // If the user cancelled the dialog, this will be empty.
   optional string selected_path = 1;
 }
 
@@ -132,4 +129,4 @@ message GetVisibleTabsRequest {
 
 message GetVisibleTabsResponse {
   repeated string paths = 1;
-}
+}

proto/host/workspace.proto

@@ -10,12 +10,8 @@ import "cline/common.proto";
 service WorkspaceService {
   // Returns a list of the top level directories of the workspace.
   rpc getWorkspacePaths(GetWorkspacePathsRequest) returns (GetWorkspacePathsResponse);
-  // Saves an open document if it's open in the editor and has unsaved changes. 
-  // Returns true if the document was saved, returns false if the document was not found, or did not
-  // need to be saved.
-  rpc saveOpenDocumentIfDirty(SaveOpenDocumentIfDirtyRequest) returns (SaveOpenDocumentIfDirtyResponse);
-  // Get diagnostics from the workspace.
-  rpc getDiagnostics(GetDiagnosticsRequest) returns (GetDiagnosticsResponse);
+  // Saves an open document if it's dirty
+  rpc saveOpenDocumentIfDirty(SaveOpenDocumentIfDirtyRequest) returns (cline.Empty);
 }
 
 message GetWorkspacePathsRequest {
@@ -32,17 +28,6 @@ message GetWorkspacePathsResponse {
 }
 
 message SaveOpenDocumentIfDirtyRequest {
-  optional string file_path = 2;
-}
-message SaveOpenDocumentIfDirtyResponse {
-  // Returns true if the document was saved. 
-  optional bool was_saved = 1;
-}
-
-message GetDiagnosticsRequest {
-  optional cline.Metadata metadata = 1;
-}
-
-message GetDiagnosticsResponse {
-  repeated cline.FileDiagnostics file_diagnostics = 1;
+  cline.Metadata metadata = 1;
+  string file_path = 2;
 }

scripts/generate-protobus-setup.mjs

@@ -40,11 +40,11 @@ async function generateWebviewProtobusClients(protobusServices) {
 			}
 			if (!rpc.responseStream) {
 				rpcs.push(`    static async ${rpcName}(request: ${requestType}): Promise<${responseType}> {
-		return this.makeUnaryRequest("${rpcName}", request, ${requestType}.toJSON, ${responseType}.fromJSON)
+		return this.makeRequest("${rpcName}", request)
 	}`)
 			} else {
 				rpcs.push(`    static ${rpcName}(request: ${requestType}, callbacks: Callbacks<${responseType}>): ()=>void {
-		return this.makeStreamingRequest("${rpcName}", request, ${requestType}.toJSON, ${responseType}.fromJSON, callbacks)
+		return this.makeStreamingRequest("${rpcName}", request, callbacks)
 	}`)
 			}
 		}

src/api/index.ts

@@ -32,7 +32,6 @@ import { GroqHandler } from "./providers/groq"
 import { Mode } from "@shared/storage/types"
 import { HuggingFaceHandler } from "./providers/huggingface"
 import { HuaweiCloudMaaSHandler } from "./providers/huawei-cloud-maas"
-import { BasetenHandler } from "./providers/baseten"
 
 export interface ApiHandler {
 	createMessage(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]): ApiStream
@@ -98,7 +97,7 @@ function createHandlerForProvider(
 					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
 				geminiApiKey: options.geminiApiKey,
 				geminiBaseUrl: options.geminiBaseUrl,
-				ulid: options.ulid,
+				taskId: options.taskId,
 			})
 		case "openai":
 			return new OpenAiHandler({
@@ -113,7 +112,6 @@ function createHandlerForProvider(
 		case "ollama":
 			return new OllamaHandler({
 				ollamaBaseUrl: options.ollamaBaseUrl,
-				ollamaApiKey: options.ollamaApiKey,
 				ollamaModelId: mode === "plan" ? options.planModeOllamaModelId : options.actModeOllamaModelId,
 				ollamaApiOptionsCtxNum: options.ollamaApiOptionsCtxNum,
 				requestTimeoutMs: options.requestTimeoutMs,
@@ -132,7 +130,7 @@ function createHandlerForProvider(
 				thinkingBudgetTokens:
 					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
 				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
-				ulid: options.ulid,
+				taskId: options.taskId,
 			})
 		case "openai-native":
 			return new OpenAiNativeHandler({
@@ -193,7 +191,7 @@ function createHandlerForProvider(
 		case "cline":
 			return new ClineHandler({
 				clineAccountId: options.clineAccountId,
-				ulid: options.ulid,
+				taskId: options.taskId,
 				reasoningEffort: mode === "plan" ? options.planModeReasoningEffort : options.actModeReasoningEffort,
 				thinkingBudgetTokens:
 					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
@@ -210,7 +208,7 @@ function createHandlerForProvider(
 				thinkingBudgetTokens:
 					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
 				liteLlmUsePromptCache: options.liteLlmUsePromptCache,
-				ulid: options.ulid,
+				taskId: options.taskId,
 			})
 		case "moonshot":
 			return new MoonshotHandler({
@@ -259,13 +257,6 @@ function createHandlerForProvider(
 				groqModelInfo: mode === "plan" ? options.planModeGroqModelInfo : options.actModeGroqModelInfo,
 				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
 			})
-		case "baseten":
-			return new BasetenHandler({
-				basetenApiKey: options.basetenApiKey,
-				basetenModelId: mode === "plan" ? options.planModeBasetenModelId : options.actModeBasetenModelId,
-				basetenModelInfo: mode === "plan" ? options.planModeBasetenModelInfo : options.actModeBasetenModelInfo,
-				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
-			})
 		case "sapaicore":
 			return new SapAiCoreHandler({
 				sapAiCoreClientId: options.sapAiCoreClientId,
@@ -274,8 +265,6 @@ function createHandlerForProvider(
 				sapAiResourceGroup: options.sapAiResourceGroup,
 				sapAiCoreBaseUrl: options.sapAiCoreBaseUrl,
 				apiModelId: mode === "plan" ? options.planModeApiModelId : options.actModeApiModelId,
-				thinkingBudgetTokens:
-					mode === "plan" ? options.planModeThinkingBudgetTokens : options.actModeThinkingBudgetTokens,
 			})
 		case "claude-code":
 			return new ClaudeCodeHandler({

src/api/providers/anthropic.ts

@@ -1,8 +1,8 @@
 import { Anthropic } from "@anthropic-ai/sdk"
 import { Stream as AnthropicStream } from "@anthropic-ai/sdk/streaming"
-import { anthropicDefaultModelId, AnthropicModelId, anthropicModels, CLAUDE_SONNET_4_1M_SUFFIX, ModelInfo } from "@shared/api"
-import { ApiHandler } from "../index"
 import { withRetry } from "../retry"
+import { anthropicDefaultModelId, AnthropicModelId, anthropicModels, ApiHandlerOptions, ModelInfo } from "@shared/api"
+import { ApiHandler } from "../index"
 import { ApiStream } from "../transform/stream"
 
 interface AnthropicHandlerOptions {
@@ -43,11 +43,7 @@ export class AnthropicHandler implements ApiHandler {
 
 		const model = this.getModel()
 		let stream: AnthropicStream<Anthropic.RawMessageStreamEvent>
-
-		const modelId = model.id.endsWith(CLAUDE_SONNET_4_1M_SUFFIX)
-			? model.id.slice(0, -CLAUDE_SONNET_4_1M_SUFFIX.length)
-			: model.id
-		const enable1mContextWindow = model.id.endsWith(CLAUDE_SONNET_4_1M_SUFFIX)
+		const modelId = model.id
 
 		const budget_tokens = this.options.thinkingBudgetTokens || 0
 		const reasoningOn = (modelId.includes("3-7") || modelId.includes("4-")) && budget_tokens !== 0 ? true : false
@@ -59,7 +55,6 @@ export class AnthropicHandler implements ApiHandler {
 			case "claude-3-5-sonnet-20241022":
 			case "claude-3-5-haiku-20241022":
 			case "claude-opus-4-20250514":
-			case "claude-opus-4-1-20250805":
 			case "claude-3-opus-20240229":
 			case "claude-3-haiku-20240307": {
 				/*
@@ -121,15 +116,24 @@ export class AnthropicHandler implements ApiHandler {
 						stream: true,
 					},
 					(() => {
-						// 1m context window beta header
-						if (enable1mContextWindow) {
-							return {
-								headers: {
-									"anthropic-beta": "context-1m-2025-08-07",
-								},
-							}
-						} else {
-							return undefined
+						// prompt caching: https://x.com/alexalbert__/status/1823751995901272068
+						// https://github.com/anthropics/anthropic-sdk-typescript?tab=readme-ov-file#default-headers
+						// https://github.com/anthropics/anthropic-sdk-typescript/commit/c920b77fc67bd839bfeb6716ceab9d7c9bbe7393
+						switch (modelId) {
+							case "claude-sonnet-4-20250514":
+							case "claude-opus-4-20250514":
+							case "claude-3-7-sonnet-20250219":
+							case "claude-3-5-sonnet-20241022":
+							case "claude-3-5-haiku-20241022":
+							case "claude-3-opus-20240229":
+							case "claude-3-haiku-20240307":
+								return {
+									headers: {
+										"anthropic-beta": "prompt-caching-2024-07-31",
+									},
+								}
+							default:
+								return undefined
 						}
 					})(),
 				)

src/api/providers/baseten.ts

@@ -1,165 +0,0 @@
-import { Anthropic } from "@anthropic-ai/sdk"
-import OpenAI from "openai"
-import { withRetry } from "../retry"
-import { ApiHandler } from "../"
-import { BasetenModelId, ModelInfo, basetenDefaultModelId, basetenModels } from "@shared/api"
-import { calculateApiCostOpenAI } from "../../utils/cost"
-import { convertToOpenAiMessages } from "../transform/openai-format"
-import { ApiStream } from "../transform/stream"
-
-interface BasetenHandlerOptions {
-	basetenApiKey?: string
-	basetenModelId?: string
-	basetenModelInfo?: ModelInfo
-	apiModelId?: string // For backward compatibility
-}
-
-export class BasetenHandler implements ApiHandler {
-	private options: BasetenHandlerOptions
-	private client: OpenAI | undefined
-
-	constructor(options: BasetenHandlerOptions) {
-		this.options = options
-	}
-
-	private ensureClient(): OpenAI {
-		if (!this.client) {
-			if (!this.options.basetenApiKey) {
-				throw new Error("Baseten API key is required")
-			}
-			try {
-				this.client = new OpenAI({
-					baseURL: "https://inference.baseten.co/v1",
-					apiKey: this.options.basetenApiKey,
-				})
-			} catch (error) {
-				throw new Error(`Error creating Baseten client: ${error.message}`)
-			}
-		}
-		return this.client
-	}
-
-	/**
-	 * Gets the optimal max_tokens based on model capabilities
-	 */
-	private getOptimalMaxTokens(model: { id: BasetenModelId; info: ModelInfo }): number {
-		// Use model-specific max tokens if available
-		if (model.info.maxTokens && model.info.maxTokens > 0) {
-			return model.info.maxTokens
-		}
-
-		// Default fallback
-		return 8192
-	}
-
-	getModel(): { id: BasetenModelId; info: ModelInfo } {
-		// First priority: basetenModelId and basetenModelInfo
-		const basetenModelId = this.options.basetenModelId
-		const basetenModelInfo = this.options.basetenModelInfo
-		if (basetenModelId && basetenModelInfo) {
-			return { id: basetenModelId as BasetenModelId, info: basetenModelInfo }
-		}
-
-		// Second priority: basetenModelId with static model info
-		if (basetenModelId && basetenModelId in basetenModels) {
-			const id = basetenModelId as BasetenModelId
-			return { id, info: basetenModels[id] }
-		}
-
-		// Third priority: apiModelId (for backward compatibility)
-		const apiModelId = this.options.apiModelId
-		if (apiModelId && apiModelId in basetenModels) {
-			const id = apiModelId as BasetenModelId
-			return { id, info: basetenModels[id] }
-		}
-
-		// Default fallback
-		return {
-			id: basetenDefaultModelId,
-			info: basetenModels[basetenDefaultModelId],
-		}
-	}
-
-	private async *yieldUsage(modelInfo: ModelInfo, usage: any): ApiStream {
-		if (usage.prompt_tokens || usage.completion_tokens) {
-			const cost = calculateApiCostOpenAI(modelInfo, usage.prompt_tokens || 0, usage.completion_tokens || 0)
-
-			yield {
-				type: "usage",
-				inputTokens: usage.prompt_tokens || 0,
-				outputTokens: usage.completion_tokens || 0,
-				cacheWriteTokens: 0,
-				cacheReadTokens: 0,
-				totalCost: cost,
-			}
-		}
-	}
-
-	@withRetry()
-	async *createMessage(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]): ApiStream {
-		const client = this.ensureClient()
-		const model = this.getModel()
-		const maxTokens = this.getOptimalMaxTokens(model)
-
-		const openAiMessages: OpenAI.Chat.ChatCompletionMessageParam[] = [
-			{ role: "system", content: systemPrompt },
-			...convertToOpenAiMessages(messages),
-		]
-
-		const stream = await client.chat.completions.create({
-			model: model.id,
-			max_tokens: maxTokens,
-			messages: openAiMessages,
-			stream: true,
-			stream_options: { include_usage: true },
-			temperature: 0,
-		})
-
-		let didOutputUsage = false
-
-		for await (const chunk of stream) {
-			const delta = chunk.choices[0]?.delta
-
-			// Handle reasoning field if present (for reasoning models with parsed output)
-			if ((delta as any)?.reasoning) {
-				const reasoningContent = (delta as any).reasoning as string
-				yield {
-					type: "reasoning",
-					reasoning: reasoningContent,
-				}
-				continue
-			}
-
-			// Handle content field
-			if (delta?.content) {
-				yield {
-					type: "text",
-					text: delta.content,
-				}
-			}
-
-			// Handle usage information - only output once
-			if (!didOutputUsage && chunk.usage) {
-				yield* this.yieldUsage(model.info, chunk.usage)
-				didOutputUsage = true
-			}
-		}
-	}
-
-	/**
-	 * Checks if the current model supports vision/images
-	 */
-	supportsImages(): boolean {
-		const model = this.getModel()
-		return model.info.supportsImages === true
-	}
-
-	/**
-	 * Checks if the current model supports tools
-	 */
-	supportsTools(): boolean {
-		const model = this.getModel()
-		// Baseten models support tools via OpenAI-compatible API
-		return true
-	}
-}

src/api/providers/bedrock.ts

@@ -2,7 +2,7 @@ import { Anthropic } from "@anthropic-ai/sdk"
 import { withRetry } from "../retry"
 import { ApiHandler } from "../"
 import { convertToR1Format } from "../transform/r1-format"
-import { bedrockDefaultModelId, BedrockModelId, bedrockModels, CLAUDE_SONNET_4_1M_SUFFIX, ModelInfo } from "@shared/api"
+import { bedrockDefaultModelId, BedrockModelId, bedrockModels, ModelInfo } from "@shared/api"
 import { calculateApiCostOpenAI } from "../../utils/cost"
 import { ApiStream } from "../transform/stream"
 import { fromNodeProviderChain } from "@aws-sdk/credential-providers"
@@ -117,14 +117,7 @@ export class AwsBedrockHandler implements ApiHandler {
 	@withRetry({ maxRetries: 4 })
 	async *createMessage(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]): ApiStream {
 		// cross region inference requires prefixing the model id with the region
-		const rawModelId = await this.getModelId()
-
-		const modelId = rawModelId.endsWith(CLAUDE_SONNET_4_1M_SUFFIX)
-			? rawModelId.slice(0, -CLAUDE_SONNET_4_1M_SUFFIX.length)
-			: rawModelId
-
-		const enable1mContextWindow = rawModelId.endsWith(CLAUDE_SONNET_4_1M_SUFFIX)
-
+		const modelId = await this.getModelId()
 		const model = this.getModel()
 
 		// This baseModelId is used to indicate the capabilities of the model.
@@ -146,7 +139,7 @@ export class AwsBedrockHandler implements ApiHandler {
 		}
 
 		// Default: Use Anthropic Converse API for all Anthropic models
-		yield* this.createAnthropicMessage(systemPrompt, messages, modelId, model, enable1mContextWindow)
+		yield* this.createAnthropicMessage(systemPrompt, messages, modelId, model)
 	}
 
 	getModel(): { id: string; info: ModelInfo } {
@@ -750,7 +743,6 @@ export class AwsBedrockHandler implements ApiHandler {
 		messages: Anthropic.Messages.MessageParam[],
 		modelId: string,
 		model: { id: string; info: ModelInfo },
-		enable1mContextWindow: boolean,
 	): ApiStream {
 		// Format messages for Anthropic model using unified formatter
 		const formattedMessages = this.formatMessagesForConverseAPI(messages)
@@ -781,18 +773,15 @@ export class AwsBedrockHandler implements ApiHandler {
 			messages: messagesWithCache,
 			system: systemMessages,
 			inferenceConfig: this.getInferenceConfig(model.info, "anthropic"),
-			additionalModelRequestFields: {
-				// Add thinking configuration as per LangChain documentation
-				...(reasoningOn && {
-					thinking: {
-						type: "enabled",
-						budget_tokens: budget_tokens,
-					},
-				}),
-				...(enable1mContextWindow && {
-					anthropic_beta: ["context-1m-2025-08-07"],
-				}),
-			},
+			// Add thinking configuration as per LangChain documentation
+			additionalModelRequestFields: reasoningOn
+				? {
+						thinking: {
+							type: "enabled",
+							budget_tokens: budget_tokens,
+						},
+					}
+				: undefined,
 		})
 
 		// Execute the streaming request using unified handler

src/api/providers/cerebras.ts

@@ -39,11 +39,7 @@ export class CerebrasHandler implements ApiHandler {
 		return this.client
 	}
 
-	@withRetry({
-		maxRetries: 6, // More retries to be patient with rate limits
-		baseDelay: 5000, // Start with 5 second delay
-		maxDelay: 60000, // Allow up to 60 second delays to respect rate limits
-	})
+	@withRetry()
 	async *createMessage(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]): ApiStream {
 		const client = this.ensureClient()
 
@@ -174,25 +170,7 @@ export class CerebrasHandler implements ApiHandler {
 					}
 				}
 			}
-		} catch (error: any) {
-			// Enhanced error handling for Cerebras API
-			if (error?.status === 429 || error?.code === "rate_limit_exceeded") {
-				// Rate limit error - will be handled by retry decorator with patient backoff
-				const limits = this.getRateLimits()
-				throw new Error(`Cerebras API rate limit exceeded.`)
-			} else if (error?.status === 401) {
-				throw new Error("Cerebras API authentication failed. Please check your API key.")
-			} else if (error?.status === 403) {
-				throw new Error("Cerebras API access denied. Please check your API key permissions.")
-			} else if (error?.status >= 500) {
-				// Server errors - retryable
-				throw new Error(`Cerebras API server error (${error.status}): ${error.message || "Unknown server error"}`)
-			} else if (error?.status === 400) {
-				// Client errors - not retryable
-				throw new Error(`Cerebras API bad request: ${error.message || "Invalid request parameters"}`)
-			}
-
-			// Re-throw original error for other cases
+		} catch (error) {
 			throw error
 		}
 	}
@@ -215,35 +193,6 @@ export class CerebrasHandler implements ApiHandler {
 		}
 	}
 
-	/**
-	 * Get rate limit information for the current model
-	 *
-	 * These limits are used for informational purposes and to calculate appropriate
-	 * retry delays. Since Cerebras inference is extremely fast, users hit these limits
-	 * quickly, so we need to be patient with retries to maximize usage efficiency.
-	 *
-	 * @returns Rate limit configuration for the model
-	 */
-	private getRateLimits(): { requestsPerMinute: number; tokensPerMinute: number } {
-		const modelId = this.getModel().id
-
-		switch (modelId) {
-			case "qwen-3-coder-480b":
-			case "qwen-3-coder-480b-free":
-				return { requestsPerMinute: 10, tokensPerMinute: 150_000 }
-			case "qwen-3-235b-a22b-instruct-2507":
-			case "qwen-3-235b-a22b-thinking-2507":
-				return { requestsPerMinute: 30, tokensPerMinute: 60_000 }
-			case "llama-3.3-70b":
-			case "gpt-oss-120b":
-			case "qwen-3-32b":
-				return { requestsPerMinute: 30, tokensPerMinute: 64_000 }
-			default:
-				// Default rate limits for unknown models
-				return { requestsPerMinute: 30, tokensPerMinute: 60_000 }
-		}
-	}
-
 	private calculateCost({ inputTokens, outputTokens }: { inputTokens: number; outputTokens: number }): number {
 		const model = this.getModel()
 		const inputPrice = model.info.inputPrice || 0

src/api/providers/cline.ts

@@ -15,7 +15,7 @@ import { CLINE_ACCOUNT_AUTH_ERROR_MESSAGE } from "@/shared/ClineAccount"
 import { clineEnvConfig } from "@/config"
 
 interface ClineHandlerOptions {
-	ulid?: string
+	taskId?: string
 	reasoningEffort?: string
 	thinkingBudgetTokens?: number
 	openRouterProviderSorting?: string
@@ -51,7 +51,7 @@ export class ClineHandler implements ApiHandler {
 					defaultHeaders: {
 						"HTTP-Referer": "https://cline.bot",
 						"X-Title": "Cline",
-						"X-Task-ID": this.options.ulid || "",
+						"X-Task-ID": this.options.taskId || "",
 						"X-Cline-Version": extensionVersion,
 					},
 				})
@@ -133,6 +133,7 @@ export class ClineHandler implements ApiHandler {
 				if (!didOutputUsage && chunk.usage) {
 					// @ts-ignore-next-line
 					let totalCost = (chunk.usage.cost || 0) + (chunk.usage.cost_details?.upstream_inference_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)
@@ -140,14 +141,27 @@ export class ClineHandler implements ApiHandler {
 					// 	totalCost = 0
 					// }
 
-					yield {
-						type: "usage",
-						cacheWriteTokens: 0,
-						cacheReadTokens: chunk.usage.prompt_tokens_details?.cached_tokens || 0,
-						inputTokens: (chunk.usage.prompt_tokens || 0) - (chunk.usage.prompt_tokens_details?.cached_tokens || 0),
-						outputTokens: chunk.usage.completion_tokens || 0,
-						// @ts-ignore-next-line
-						totalCost: totalCost,
+					if (modelId.includes("gemini")) {
+						yield {
+							type: "usage",
+							cacheWriteTokens: 0,
+							cacheReadTokens: chunk.usage.prompt_tokens_details?.cached_tokens || 0,
+							inputTokens:
+								(chunk.usage.prompt_tokens || 0) - (chunk.usage.prompt_tokens_details?.cached_tokens || 0),
+							outputTokens: chunk.usage.completion_tokens || 0,
+							// @ts-ignore-next-line
+							totalCost,
+						}
+					} else {
+						yield {
+							type: "usage",
+							cacheWriteTokens: 0,
+							cacheReadTokens: chunk.usage.prompt_tokens_details?.cached_tokens || 0,
+							inputTokens: chunk.usage.prompt_tokens || 0,
+							outputTokens: chunk.usage.completion_tokens || 0,
+							// @ts-ignore-next-line
+							totalCost,
+						}
 					}
 					didOutputUsage = true
 				}
@@ -172,26 +186,36 @@ export class ClineHandler implements ApiHandler {
 			try {
 				// TODO: replace this with firebase auth
 				// TODO: use global API Host
-				const clineAccountAuthToken = await this._authService.getAuthToken()
-				if (!clineAccountAuthToken) {
-					throw new Error(CLINE_ACCOUNT_AUTH_ERROR_MESSAGE)
-				}
+
 				const response = await axios.get(`${this.clineAccountService.baseUrl}/generation?id=${this.lastGenerationId}`, {
 					headers: {
-						Authorization: `Bearer ${clineAccountAuthToken}`,
+						Authorization: `Bearer ${this.options.clineAccountId}`,
 					},
 					timeout: 15_000, // this request hangs sometimes
 				})
 
 				const generation = response.data
-				return {
-					type: "usage",
-					cacheWriteTokens: 0,
-					cacheReadTokens: generation?.native_tokens_cached || 0,
-					// openrouter generation endpoint fails often
-					inputTokens: (generation?.native_tokens_prompt || 0) - (generation?.native_tokens_cached || 0),
-					outputTokens: generation?.native_tokens_completion || 0,
-					totalCost: generation?.total_cost || 0,
+				let modelId = this.options.openRouterModelId
+				if (modelId && modelId.includes("gemini")) {
+					return {
+						type: "usage",
+						cacheWriteTokens: 0,
+						cacheReadTokens: generation?.native_tokens_cached || 0,
+						// openrouter generation endpoint fails often
+						inputTokens: (generation?.native_tokens_prompt || 0) - (generation?.native_tokens_cached || 0),
+						outputTokens: generation?.native_tokens_completion || 0,
+						totalCost: generation?.total_cost || 0,
+					}
+				} else {
+					return {
+						type: "usage",
+						cacheWriteTokens: 0,
+						cacheReadTokens: generation?.native_tokens_cached || 0,
+						// openrouter generation endpoint fails often
+						inputTokens: generation?.native_tokens_prompt || 0,
+						outputTokens: generation?.native_tokens_completion || 0,
+						totalCost: generation?.total_cost || 0,
+					}
 				}
 			} catch (error) {
 				// ignore if fails

src/api/providers/gemini.ts

@@ -7,7 +7,7 @@ import { ApiHandler } from "../"
 import { ApiHandlerOptions, geminiDefaultModelId, GeminiModelId, geminiModels, ModelInfo } from "@shared/api"
 import { convertAnthropicMessageToGemini } from "../transform/gemini-format"
 import { ApiStream } from "../transform/stream"
-import { telemetryService } from "@services/posthog/PostHogClientProvider"
+import { telemetryService } from "@services/posthog/telemetry/TelemetryService"
 
 // Define a default TTL for the cache (e.g., 15 minutes in seconds)
 const DEFAULT_CACHE_TTL_SECONDS = 900
@@ -20,7 +20,7 @@ interface GeminiHandlerOptions {
 	geminiBaseUrl?: string
 	thinkingBudgetTokens?: number
 	apiModelId?: string
-	ulid?: string
+	taskId?: string
 }
 
 /**
@@ -28,7 +28,7 @@ interface GeminiHandlerOptions {
  *
  * Key features:
  * - One cache per task: Creates a single cache per task and reuses it for subsequent turns
- * - Stable cache keys: Uses ulid as a stable identifier for caches
+ * - Stable cache keys: Uses taskId as a stable identifier for caches
  * - Efficient cache updates: Only updates caches when there's new content to add
  * - Split cost accounting: Separates immediate costs from ongoing cache storage costs
  *
@@ -255,8 +255,8 @@ export class GeminiHandler implements ApiHandler {
 			const throughputTokensPerSecSdk =
 				totalDurationSdkMs > 0 && outputTokens > 0 ? outputTokens / (totalDurationSdkMs / 1000) : undefined
 
-			if (this.options.ulid) {
-				telemetryService.captureGeminiApiPerformance(this.options.ulid, modelId, {
+			if (this.options.taskId) {
+				telemetryService.captureGeminiApiPerformance(this.options.taskId, modelId, {
 					ttftSec: ttftSdkMs !== undefined ? ttftSdkMs / 1000 : undefined,
 					totalDurationSec: totalDurationSdkMs / 1000,
 					promptTokens,
@@ -269,7 +269,7 @@ export class GeminiHandler implements ApiHandler {
 					throughputTokensPerSec: throughputTokensPerSecSdk,
 				})
 			} else {
-				console.warn("GeminiHandler: ulid not available for telemetry in createMessage.")
+				console.warn("GeminiHandler: taskId not available for telemetry in createMessage.")
 			}
 		}
 	}

src/api/providers/litellm.ts

@@ -13,32 +13,12 @@ interface LiteLlmHandlerOptions {
 	liteLlmModelInfo?: LiteLLMModelInfo
 	thinkingBudgetTokens?: number
 	liteLlmUsePromptCache?: boolean
-	ulid?: string
-}
-
-interface LiteLlmModelInfoResponse {
-	data: Array<{
-		model_name: string
-		litellm_params: {
-			model: string
-			[key: string]: any
-		}
-		model_info: {
-			input_cost_per_token: number
-			output_cost_per_token: number
-			cache_creation_input_token_cost?: number
-			cache_read_input_token_cost?: number
-			[key: string]: any
-		}
-	}>
+	taskId?: string
 }
 
 export class LiteLlmHandler implements ApiHandler {
 	private options: LiteLlmHandlerOptions
 	private client: OpenAI | undefined
-	private modelInfoCache: LiteLlmModelInfoResponse | undefined
-	private modelInfoCacheTimestamp: number = 0
-	private readonly modelInfoCacheTTL = 5 * 60 * 1000 // 5 minutes
 
 	constructor(options: LiteLlmHandlerOptions) {
 		this.options = options
@@ -61,112 +41,35 @@ export class LiteLlmHandler implements ApiHandler {
 		return this.client
 	}
 
-	private async fetchModelInfo(): Promise<LiteLlmModelInfoResponse | undefined> {
-		// Check if cache is still valid
-		const now = Date.now()
-		if (this.modelInfoCache && now - this.modelInfoCacheTimestamp < this.modelInfoCacheTTL) {
-			return this.modelInfoCache
-		}
-
+	async calculateCost(prompt_tokens: number, completion_tokens: number): Promise<number | undefined> {
+		// Reference: https://github.com/BerriAI/litellm/blob/122ee634f434014267af104814022af1d9a0882f/litellm/proxy/spend_tracking/spend_management_endpoints.py#L1473
 		const client = this.ensureClient()
-		// Handle base URLs that already include /v1 to avoid double /v1/v1/
-		const baseUrl = client.baseURL.endsWith("/v1") ? client.baseURL : `${client.baseURL}/v1`
-		const url = `${baseUrl}/model/info`
-
+		const modelId = this.options.liteLlmModelId || liteLlmDefaultModelId
 		try {
-			const response = await fetch(url, {
-				method: "GET",
+			const response = await fetch(`${client.baseURL}/spend/calculate`, {
+				method: "POST",
 				headers: {
-					accept: "application/json",
-					"x-litellm-api-key": this.options.liteLlmApiKey || "",
+					"Content-Type": "application/json",
+					Authorization: `Bearer ${this.options.liteLlmApiKey}`,
 				},
+				body: JSON.stringify({
+					completion_response: {
+						model: modelId,
+						usage: {
+							prompt_tokens,
+							completion_tokens,
+						},
+					},
+				}),
 			})
 
 			if (response.ok) {
-				const data: LiteLlmModelInfoResponse = await response.json()
-				this.modelInfoCache = data
-				this.modelInfoCacheTimestamp = now
-				return data
+				const data: { cost: number } = await response.json()
+				return data.cost
 			} else {
-				console.warn("Failed to fetch LiteLLM model info:", response.statusText)
-				// Try with Authorization header instead
-				const retryResponse = await fetch(url, {
-					method: "GET",
-					headers: {
-						accept: "application/json",
-						Authorization: `Bearer ${this.options.liteLlmApiKey || ""}`,
-					},
-				})
-
-				if (retryResponse.ok) {
-					const data: LiteLlmModelInfoResponse = await retryResponse.json()
-					this.modelInfoCache = data
-					this.modelInfoCacheTimestamp = now
-					return data
-				} else {
-					console.warn("Failed to fetch LiteLLM model info with Authorization header:", retryResponse.statusText)
-					return undefined
-				}
+				console.error("Error calculating spend:", response.statusText)
+				return undefined
 			}
-		} catch (error) {
-			console.warn("Error fetching LiteLLM model info:", error)
-			return undefined
-		}
-	}
-
-	private async getModelCostInfo(publicModelName: string): Promise<{
-		inputCostPerToken: number
-		outputCostPerToken: number
-		cacheCreationCostPerToken?: number
-		cacheReadCostPerToken?: number
-	}> {
-		try {
-			const modelInfo = await this.fetchModelInfo()
-
-			if (modelInfo?.data) {
-				// Find the model by public name
-				const matchingModel = modelInfo.data.find((model) => model.model_name === publicModelName)
-
-				if (matchingModel?.model_info) {
-					return {
-						inputCostPerToken: matchingModel.model_info.input_cost_per_token || 0,
-						outputCostPerToken: matchingModel.model_info.output_cost_per_token || 0,
-						cacheCreationCostPerToken: matchingModel.model_info.cache_creation_input_token_cost,
-						cacheReadCostPerToken: matchingModel.model_info.cache_read_input_token_cost,
-					}
-				}
-			}
-		} catch (error) {
-			console.warn("Error getting LiteLLM model cost info:", error)
-		}
-
-		// Fallback to zero costs if we can't get the information
-		return {
-			inputCostPerToken: 0,
-			outputCostPerToken: 0,
-		}
-	}
-
-	async calculateCost(
-		prompt_tokens: number,
-		completion_tokens: number,
-		cache_creation_tokens?: number,
-		cache_read_tokens?: number,
-	): Promise<number | undefined> {
-		const publicModelId = this.options.liteLlmModelId || liteLlmDefaultModelId
-
-		try {
-			const costInfo = await this.getModelCostInfo(publicModelId)
-
-			// Calculate costs for different token types
-			const inputCost = Math.max(0, prompt_tokens - (cache_read_tokens || 0)) * costInfo.inputCostPerToken
-			const outputCost = completion_tokens * costInfo.outputCostPerToken
-			const cacheCreationCost = (cache_creation_tokens || 0) * (costInfo.cacheCreationCostPerToken || 0)
-			const cacheReadCost = (cache_read_tokens || 0) * (costInfo.cacheReadCostPerToken || 0)
-
-			const totalCost = inputCost + outputCost + cacheCreationCost + cacheReadCost
-
-			return totalCost
 		} catch (error) {
 			console.error("Error calculating spend:", error)
 			return undefined
@@ -230,9 +133,12 @@ export class LiteLlmHandler implements ApiHandler {
 			stream: true,
 			stream_options: { include_usage: true },
 			...(thinkingConfig && { thinking: thinkingConfig }), // Add thinking configuration when applicable
-			...(this.options.ulid && { litellm_session_id: `cline-${this.options.ulid}` }), // Add session ID for LiteLLM tracking
+			...(this.options.taskId && { litellm_session_id: `cline-${this.options.taskId}` }), // Add session ID for LiteLLM tracking
 		})
 
+		const inputCost = (await this.calculateCost(1e6, 0)) || 0
+		const outputCost = (await this.calculateCost(0, 1e6)) || 0
+
 		for await (const chunk of stream) {
 			const delta = chunk.choices[0]?.delta
 
@@ -259,6 +165,9 @@ export class LiteLlmHandler implements ApiHandler {
 
 			// Handle token usage information
 			if (chunk.usage) {
+				const totalCost =
+					(inputCost * chunk.usage.prompt_tokens) / 1e6 + (outputCost * chunk.usage.completion_tokens) / 1e6
+
 				// Extract cache-related information if available
 				// Need to use type assertion since these properties are not in the standard OpenAI types
 				const usage = chunk.usage as {
@@ -273,15 +182,6 @@ export class LiteLlmHandler implements ApiHandler {
 				const cacheWriteTokens = usage.cache_creation_input_tokens || usage.prompt_cache_miss_tokens || 0
 				const cacheReadTokens = usage.cache_read_input_tokens || usage.prompt_cache_hit_tokens || 0
 
-				// Calculate cost using the actual token usage including cache tokens
-				const totalCost =
-					(await this.calculateCost(
-						usage.prompt_tokens || 0,
-						usage.completion_tokens || 0,
-						cacheWriteTokens > 0 ? cacheWriteTokens : undefined,
-						cacheReadTokens > 0 ? cacheReadTokens : undefined,
-					)) || 0
-
 				yield {
 					type: "usage",
 					inputTokens: usage.prompt_tokens || 0,

src/api/providers/ollama.ts

@@ -1,5 +1,5 @@
 import { Anthropic } from "@anthropic-ai/sdk"
-import { Message, Ollama, Config } from "ollama"
+import { Message, Ollama } from "ollama"
 import { ApiHandler } from "../"
 import { ApiHandlerOptions, ModelInfo, openAiModelInfoSaneDefaults } from "../../shared/api"
 import { convertToOllamaMessages } from "../transform/ollama-format"
@@ -8,7 +8,6 @@ import { withRetry } from "../retry"
 
 interface OllamaHandlerOptions {
 	ollamaBaseUrl?: string
-	ollamaApiKey?: string
 	ollamaModelId?: string
 	ollamaApiOptionsCtxNum?: string
 	requestTimeoutMs?: number
@@ -25,18 +24,7 @@ export class OllamaHandler implements ApiHandler {
 	private ensureClient(): Ollama {
 		if (!this.client) {
 			try {
-				const clientOptions: Partial<Config> = {
-					host: this.options.ollamaBaseUrl || "http://localhost:11434",
-				}
-
-				// Add API key if provided (for Ollama cloud or authenticated instances)
-				if (this.options.ollamaApiKey) {
-					clientOptions.headers = {
-						Authorization: `Bearer ${this.options.ollamaApiKey}`,
-					}
-				}
-
-				this.client = new Ollama(clientOptions)
+				this.client = new Ollama({ host: this.options.ollamaBaseUrl || "http://localhost:11434" })
 			} catch (error) {
 				throw new Error(`Error creating Ollama client: ${error.message}`)
 			}

src/api/providers/openai-native.ts

@@ -104,33 +104,6 @@ export class OpenAiNativeHandler implements ApiHandler {
 				}
 				break
 			}
-			case "nectarine-alpha-new-reasoning-effort-2025-07-25":
-			case "gpt-5-2025-08-07":
-			case "gpt-5-mini-2025-08-07":
-			case "gpt-5-nano-2025-08-07":
-				const stream = await client.chat.completions.create({
-					model: model.id,
-					temperature: 1,
-					messages: [{ role: "developer", content: systemPrompt }, ...convertToOpenAiMessages(messages)],
-					stream: true,
-					stream_options: { include_usage: true },
-					reasoning_effort: (this.options.reasoningEffort as ChatCompletionReasoningEffort) || "medium",
-				})
-
-				for await (const chunk of stream) {
-					const delta = chunk.choices[0]?.delta
-					if (delta?.content) {
-						yield {
-							type: "text",
-							text: delta.content,
-						}
-					}
-					if (chunk.usage) {
-						// Only last chunk contains usage
-						yield* this.yieldUsage(model.info, chunk.usage)
-					}
-				}
-				break
 			default: {
 				const stream = await client.chat.completions.create({
 					model: model.id,

src/api/providers/openrouter.ts

@@ -132,14 +132,27 @@ export class OpenRouterHandler implements ApiHandler {
 			}
 
 			if (!didOutputUsage && chunk.usage) {
-				yield {
-					type: "usage",
-					cacheWriteTokens: 0,
-					cacheReadTokens: chunk.usage.prompt_tokens_details?.cached_tokens || 0,
-					inputTokens: (chunk.usage.prompt_tokens || 0) - (chunk.usage.prompt_tokens_details?.cached_tokens || 0),
-					outputTokens: chunk.usage.completion_tokens || 0,
-					// @ts-ignore-next-line
-					totalCost: (chunk.usage.cost || 0) + (chunk.usage.cost_details?.upstream_inference_cost || 0),
+				let modelId = this.options.openRouterModelId
+				if (modelId && modelId.includes("gemini")) {
+					yield {
+						type: "usage",
+						cacheWriteTokens: 0,
+						cacheReadTokens: chunk.usage.prompt_tokens_details?.cached_tokens || 0,
+						inputTokens: (chunk.usage.prompt_tokens || 0) - (chunk.usage.prompt_tokens_details?.cached_tokens || 0),
+						outputTokens: chunk.usage.completion_tokens || 0,
+						// @ts-ignore-next-line
+						totalCost: (chunk.usage.cost || 0) + (chunk.usage.cost_details?.upstream_inference_cost || 0),
+					}
+				} else {
+					yield {
+						type: "usage",
+						cacheWriteTokens: 0,
+						cacheReadTokens: chunk.usage.prompt_tokens_details?.cached_tokens || 0,
+						inputTokens: chunk.usage.prompt_tokens || 0,
+						outputTokens: chunk.usage.completion_tokens || 0,
+						// @ts-ignore-next-line
+						totalCost: (chunk.usage.cost || 0) + (chunk.usage.cost_details?.upstream_inference_cost || 0),
+					}
 				}
 				didOutputUsage = true
 			}
@@ -161,14 +174,27 @@ export class OpenRouterHandler implements ApiHandler {
 				const generationIterator = this.fetchGenerationDetails(this.lastGenerationId)
 				const generation = (await generationIterator.next()).value
 				// console.log("OpenRouter generation details:", generation)
-				return {
-					type: "usage",
-					cacheWriteTokens: 0,
-					cacheReadTokens: generation?.native_tokens_cached || 0,
-					// openrouter generation endpoint fails often
-					inputTokens: (generation?.native_tokens_prompt || 0) - (generation?.native_tokens_cached || 0),
-					outputTokens: generation?.native_tokens_completion || 0,
-					totalCost: generation?.total_cost || 0,
+				let modelId = this.options.openRouterModelId
+				if (modelId && modelId.includes("gemini")) {
+					return {
+						type: "usage",
+						cacheWriteTokens: 0,
+						cacheReadTokens: generation?.native_tokens_cached || 0,
+						// openrouter generation endpoint fails often
+						inputTokens: (generation?.native_tokens_prompt || 0) - (generation?.native_tokens_cached || 0),
+						outputTokens: generation?.native_tokens_completion || 0,
+						totalCost: generation?.total_cost || 0,
+					}
+				} else {
+					return {
+						type: "usage",
+						cacheWriteTokens: 0,
+						cacheReadTokens: generation?.native_tokens_cached || 0,
+						// openrouter generation endpoint fails often
+						inputTokens: generation?.native_tokens_prompt || 0,
+						outputTokens: generation?.native_tokens_completion || 0,
+						totalCost: generation?.total_cost || 0,
+					}
 				}
 			} catch (error) {
 				// ignore if fails

src/api/providers/requesty.ts

@@ -74,10 +74,7 @@ export class RequestyHandler implements ApiHandler {
 				? { thinking: { type: "enabled", budget_tokens: thinkingBudget } }
 				: { thinking: { type: "disabled" } }
 		const thinkingArgs =
-			model.id.includes("claude-3-7-sonnet") ||
-			model.id.includes("claude-sonnet-4") ||
-			model.id.includes("claude-opus-4") ||
-			model.id.includes("claude-opus-4-1")
+			model.id.includes("claude-3-7-sonnet") || model.id.includes("claude-sonnet-4") || model.id.includes("claude-opus-4")
 				? thinking
 				: {}
 

src/api/providers/sapaicore.ts

@@ -5,11 +5,6 @@ import { ApiHandler } from "../"
 import { ModelInfo, sapAiCoreDefaultModelId, SapAiCoreModelId, sapAiCoreModels } from "../../shared/api"
 import { convertToOpenAiMessages } from "../transform/openai-format"
 import { ApiStream } from "../transform/stream"
-import {
-	type Message as BedrockMessage,
-	type ContentBlock as BedrockContentBlock,
-	ConversationRole as BedrockConversationRole,
-} from "@aws-sdk/client-bedrock-runtime"
 
 interface SapAiCoreHandlerOptions {
 	sapAiCoreClientId?: string
@@ -18,7 +13,6 @@ interface SapAiCoreHandlerOptions {
 	sapAiResourceGroup?: string
 	sapAiCoreBaseUrl?: string
 	apiModelId?: string
-	thinkingBudgetTokens?: number
 }
 
 interface Deployment {
@@ -33,307 +27,6 @@ interface Token {
 	token_type: string
 	expires_at: number
 }
-
-// Bedrock namespace containing caching-related functions
-namespace Bedrock {
-	// Define cache point type for AWS Bedrock
-	interface CachePointContentBlock {
-		cachePoint: {
-			type: "default"
-		}
-	}
-
-	// 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
-		}
-	}
-
-	/**
-	 * Prepares system messages with optional caching support
-	 */
-	export function prepareSystemMessages(systemPrompt: string, enableCaching: boolean): any[] | undefined {
-		if (!systemPrompt) {
-			return undefined
-		}
-
-		if (enableCaching) {
-			return [{ text: systemPrompt }, { cachePoint: { type: "default" } }]
-		}
-
-		return [{ text: systemPrompt }]
-	}
-
-	/**
-	 * 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
-	 */
-	export function applyCacheControlToMessages(
-		messages: BedrockMessage[],
-		lastUserMsgIndex: number,
-		secondLastMsgUserIndex: number,
-	): BedrockMessage[] {
-		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
-		})
-	}
-
-	/**
-	 * Formats messages for models using the Converse API specification
-	 * Used by both Anthropic and Nova models to avoid code duplication
-	 */
-	export function formatMessagesForConverseAPI(messages: Anthropic.Messages.MessageParam[]): BedrockMessage[] {
-		return messages.map((message) => {
-			// Determine role (user or assistant)
-			const role = message.role === "user" ? BedrockConversationRole.USER : BedrockConversationRole.ASSISTANT
-
-			// Process content based on type
-			let content: BedrockContentBlock[] = []
-
-			if (typeof message.content === "string") {
-				// Simple text content
-				content = [{ text: message.content }]
-			} else if (Array.isArray(message.content)) {
-				// Convert Anthropic content format to Converse API content format
-				const processedContent = message.content
-					.map((item) => {
-						// Text content
-						if (item.type === "text") {
-							return { text: item.text }
-						}
-
-						// Image content
-						if (item.type === "image") {
-							return processImageContent(item)
-						}
-
-						// Log unsupported content types for debugging
-						console.warn(`Unsupported content type: ${(item as ContentItem).type}`)
-						return null
-					})
-					.filter((item): item is BedrockContentBlock => item !== null)
-
-				content = processedContent
-			}
-
-			// Return formatted message
-			return {
-				role,
-				content,
-			}
-		})
-	}
-
-	/**
-	 * Processes image content with proper error handling and user notification
-	 */
-	function processImageContent(item: any): BedrockContentBlock | 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"}]`,
-			}
-		}
-	}
-}
-
-// Gemini namespace containing caching-related functions and types
-namespace Gemini {
-	/**
-	 * Process Gemini streaming response with enhanced thinking content support and caching awareness
-	 */
-	export function processStreamChunk(data: any): {
-		text?: string
-		reasoning?: string
-		usageMetadata?: {
-			promptTokenCount?: number
-			candidatesTokenCount?: number
-			thoughtsTokenCount?: number
-			cachedContentTokenCount?: number
-		}
-	} {
-		const result: ReturnType<typeof processStreamChunk> = {}
-
-		// Handle thinking content from Gemini's response
-		const candidateForThoughts = data?.candidates?.[0]
-		const partsForThoughts = candidateForThoughts?.content?.parts
-		let thoughts = ""
-
-		if (partsForThoughts) {
-			for (const part of partsForThoughts) {
-				const { thought, text } = part
-				if (thought && text) {
-					thoughts += text + "\n"
-				}
-			}
-		}
-
-		if (thoughts.trim() !== "") {
-			result.reasoning = thoughts.trim()
-		}
-
-		// Handle regular text content
-		if (data.text) {
-			result.text = data.text
-		}
-
-		// Handle content parts for non-thought text
-		if (data.candidates && data.candidates[0]?.content?.parts) {
-			let nonThoughtText = ""
-			for (const part of data.candidates[0].content.parts) {
-				if (part.text && !part.thought) {
-					nonThoughtText += part.text
-				}
-			}
-			if (nonThoughtText && !result.text) {
-				result.text = nonThoughtText
-			}
-		}
-
-		// Handle usage metadata with caching support
-		if (data.usageMetadata) {
-			result.usageMetadata = {
-				promptTokenCount: data.usageMetadata.promptTokenCount,
-				candidatesTokenCount: data.usageMetadata.candidatesTokenCount,
-				thoughtsTokenCount: data.usageMetadata.thoughtsTokenCount,
-				cachedContentTokenCount: data.usageMetadata.cachedContentTokenCount,
-			}
-		}
-
-		return result
-	}
-
-	function convertAnthropicMessageToGemini(message: Anthropic.Messages.MessageParam) {
-		const role = message.role === "assistant" ? "model" : "user"
-		const parts = []
-
-		if (typeof message.content === "string") {
-			parts.push({ text: message.content })
-		} else if (Array.isArray(message.content)) {
-			for (const block of message.content) {
-				if (block.type === "text") {
-					parts.push({ text: block.text })
-				} else if (block.type === "image") {
-					parts.push({
-						inlineData: {
-							mimeType: block.source.media_type,
-							data: block.source.data,
-						},
-					})
-				}
-			}
-		}
-
-		return { role, parts }
-	}
-
-	/**
-	 * Prepare Gemini request payload with thinking configuration and implicit caching support
-	 */
-	export function prepareRequestPayload(
-		systemPrompt: string,
-		messages: Anthropic.Messages.MessageParam[],
-		model: { id: SapAiCoreModelId; info: ModelInfo },
-		thinkingBudgetTokens?: number,
-	): any {
-		const contents = messages.map(convertAnthropicMessageToGemini)
-
-		const payload = {
-			contents,
-			systemInstruction: {
-				parts: [
-					{
-						text: systemPrompt,
-					},
-				],
-			},
-			generationConfig: {
-				maxOutputTokens: model.info.maxTokens,
-				temperature: 0.0,
-			},
-		}
-
-		// Add thinking config if the model supports it and budget is provided
-		const thinkingBudget = thinkingBudgetTokens ?? 0
-		const maxBudget = model.info.thinkingConfig?.maxBudget ?? 0
-
-		if (thinkingBudget > 0 && model.info.thinkingConfig) {
-			// Add thinking configuration to the payload
-			;(payload as any).thinkingConfig = {
-				thinkingBudget: thinkingBudget,
-				includeThoughts: true,
-			}
-		}
-
-		return payload
-	}
-}
-
 export class SapAiCoreHandler implements ApiHandler {
 	private options: SapAiCoreHandlerOptions
 	private token?: Token
@@ -449,20 +142,7 @@ export class SapAiCoreHandler implements ApiHandler {
 			"anthropic--claude-3-opus",
 		]
 
-		const openAIModels = [
-			"gpt-4o",
-			"gpt-4",
-			"gpt-4o-mini",
-			"o1",
-			"gpt-4.1",
-			"gpt-4.1-nano",
-			"gpt-5",
-			"gpt-5-nano",
-			"gpt-5-mini",
-			"o3-mini",
-			"o3",
-			"o4-mini",
-		]
+		const openAIModels = ["gpt-4o", "gpt-4", "gpt-4o-mini", "o1", "gpt-4.1", "gpt-4.1-nano", "o3-mini", "o3", "o4-mini"]
 
 		const geminiModels = ["gemini-2.5-flash", "gemini-2.5-pro"]
 
@@ -471,47 +151,21 @@ export class SapAiCoreHandler implements ApiHandler {
 		if (anthropicModels.includes(model.id)) {
 			url = `${this.options.sapAiCoreBaseUrl}/v2/inference/deployments/${deploymentId}/invoke-with-response-stream`
 
-			// Format messages for Converse API. Note that the Invoke API has
-			// the same format for messages as the Converse API.
-			const formattedMessages = Bedrock.formatMessagesForConverseAPI(messages)
-
-			// Get 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
-
 			if (
 				model.id === "anthropic--claude-4-sonnet" ||
 				model.id === "anthropic--claude-4-opus" ||
 				model.id === "anthropic--claude-3.7-sonnet"
 			) {
-				// Use converse-stream endpoint with caching support
 				url = `${this.options.sapAiCoreBaseUrl}/v2/inference/deployments/${deploymentId}/converse-stream`
-
-				// Apply caching controls to messages (enabled by default)
-				const messagesWithCache = Bedrock.applyCacheControlToMessages(
-					formattedMessages,
-					lastUserMsgIndex,
-					secondLastMsgUserIndex,
-				)
-
-				// Prepare system message with caching support (enabled by default)
-				const systemMessages = Bedrock.prepareSystemMessages(systemPrompt, true)
-
 				payload = {
 					inferenceConfig: {
 						maxTokens: model.info.maxTokens,
 						temperature: 0.0,
 					},
-					system: systemMessages,
-					messages: messagesWithCache,
+					system: systemPrompt ? [{ text: systemPrompt }] : undefined,
+					messages: this.formatAnthropicMessages(messages),
 				}
 			} else {
-				// Use invoke-with-response-stream endpoint
-				// TODO: add caching support using Anthropic-native cache_control blocks
 				payload = {
 					max_tokens: model.info.maxTokens,
 					system: systemPrompt,
@@ -537,7 +191,7 @@ export class SapAiCoreHandler implements ApiHandler {
 				stream_options: { include_usage: true },
 			}
 
-			if (["o1", "o3-mini", "o3", "o4-mini", "gpt-5", "gpt-5-nano", "gpt-5-mini"].includes(model.id)) {
+			if (["o1", "o3-mini", "o3", "o4-mini"].includes(model.id)) {
 				delete payload.max_tokens
 				delete payload.temperature
 			}
@@ -548,7 +202,7 @@ export class SapAiCoreHandler implements ApiHandler {
 			}
 		} else if (geminiModels.includes(model.id)) {
 			url = `${this.options.sapAiCoreBaseUrl}/v2/inference/deployments/${deploymentId}/models/${model.id}:streamGenerateContent`
-			payload = Gemini.prepareRequestPayload(systemPrompt, messages, model, this.options.thinkingBudgetTokens)
+			payload = this.convertToGeminiFormat(systemPrompt, messages)
 		} else {
 			throw new Error(`Unsupported model: ${model.id}`)
 		}
@@ -705,17 +359,9 @@ export class SapAiCoreHandler implements ApiHandler {
 
 							// Handle metadata (token usage)
 							if (data.metadata?.usage) {
-								let inputTokens = data.metadata.usage.inputTokens || 0
+								const inputTokens = data.metadata.usage.inputTokens || 0
 								const outputTokens = data.metadata.usage.outputTokens || 0
 
-								// calibrate input token
-								const totalTokens = data.metadata.usage.totalTokens || 0
-								const cacheReadInputTokens = data.metadata.usage.cacheReadInputTokens || 0
-								const cacheWriteOutputTokens = data.metadata.usage.cacheWriteOutputTokens || 0
-								if (inputTokens + outputTokens + cacheReadInputTokens + cacheWriteOutputTokens !== totalTokens) {
-									inputTokens = totalTokens - outputTokens - cacheReadInputTokens - cacheWriteOutputTokens
-								}
-
 								yield {
 									type: "usage",
 									inputTokens,
@@ -847,31 +493,50 @@ export class SapAiCoreHandler implements ApiHandler {
 						const jsonData = line.slice(6)
 						try {
 							const data = JSON.parse(jsonData)
+							const candidateForThoughts = data?.candidates?.[0]
+							const partsForThoughts = candidateForThoughts?.content?.parts
+							let thoughts = ""
 
-							// Use Gemini namespace to process the chunk
-							const processed = Gemini.processStreamChunk(data)
+							if (partsForThoughts) {
+								for (const part of partsForThoughts) {
+									const { thought, text } = part
+									if (thought && text) {
+										thoughts += text + "\n"
+									}
+								}
+							}
 
-							// Yield reasoning if present
-							if (processed.reasoning) {
+							if (thoughts.trim() !== "") {
 								yield {
 									type: "reasoning",
-									reasoning: processed.reasoning,
+									reasoning: thoughts.trim(),
 								}
 							}
 
-							// Yield text if present
-							if (processed.text) {
+							if (data.text) {
 								yield {
 									type: "text",
-									text: processed.text,
+									text: data.text,
 								}
 							}
 
-							if (processed.usageMetadata) {
-								promptTokens = processed.usageMetadata.promptTokenCount ?? promptTokens
-								outputTokens = processed.usageMetadata.candidatesTokenCount ?? outputTokens
-								thoughtsTokenCount = processed.usageMetadata.thoughtsTokenCount ?? thoughtsTokenCount
-								cacheReadTokens = processed.usageMetadata.cachedContentTokenCount ?? cacheReadTokens
+							if (data.candidates && data.candidates[0]?.content?.parts) {
+								for (const part of data.candidates[0].content.parts) {
+									if (part.text && !part.thought) {
+										// Only non-thought text
+										yield {
+											type: "text",
+											text: part.text,
+										}
+									}
+								}
+							}
+
+							if (data.usageMetadata) {
+								promptTokens = data.usageMetadata.promptTokenCount ?? promptTokens
+								outputTokens = data.usageMetadata.candidatesTokenCount ?? outputTokens
+								thoughtsTokenCount = data.usageMetadata.thoughtsTokenCount ?? thoughtsTokenCount
+								cacheReadTokens = data.usageMetadata.cachedContentTokenCount ?? cacheReadTokens
 
 								yield {
 									type: "usage",
@@ -879,7 +544,6 @@ export class SapAiCoreHandler implements ApiHandler {
 									outputTokens,
 									thoughtsTokenCount,
 									cacheReadTokens,
-									cacheWriteTokens: 0,
 								}
 							}
 						} catch (error) {
@@ -917,4 +581,111 @@ export class SapAiCoreHandler implements ApiHandler {
 		}
 		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 convertToGeminiFormat(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]) {
+		const contents = messages.map(this.convertAnthropicMessageToGemini)
+
+		const payload = {
+			contents,
+			systemInstruction: {
+				parts: [
+					{
+						text: systemPrompt,
+					},
+				],
+			},
+			generationConfig: {
+				maxOutputTokens: this.getModel().info.maxTokens,
+				temperature: 0.0,
+			},
+		}
+
+		return payload
+	}
+
+	private convertAnthropicMessageToGemini(message: Anthropic.Messages.MessageParam) {
+		const role = message.role === "assistant" ? "model" : "user"
+		const parts = []
+
+		if (typeof message.content === "string") {
+			parts.push({ text: message.content })
+		} else if (Array.isArray(message.content)) {
+			for (const block of message.content) {
+				if (block.type === "text") {
+					parts.push({ text: block.text })
+				} else if (block.type === "image") {
+					parts.push({
+						inlineData: {
+							mimeType: block.source.media_type,
+							data: block.source.data,
+						},
+					})
+				}
+			}
+		}
+
+		return { role, parts }
+	}
+	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/providers/vertex.ts

@@ -13,7 +13,7 @@ interface VertexHandlerOptions {
 	thinkingBudgetTokens?: number
 	geminiApiKey?: string
 	geminiBaseUrl?: string
-	ulid?: string
+	taskId?: string
 }
 
 export class VertexHandler implements ApiHandler {
@@ -86,7 +86,6 @@ export class VertexHandler implements ApiHandler {
 
 		switch (modelId) {
 			case "claude-sonnet-4@20250514":
-			case "claude-opus-4-1@20250805":
 			case "claude-opus-4@20250514":
 			case "claude-3-7-sonnet@20250219":
 			case "claude-3-5-sonnet-v2@20241022":

src/api/providers/vscode-lm.ts

@@ -252,19 +252,77 @@ export class VsCodeLmHandler implements ApiHandler, SingleCompletionHandler {
 	}
 
 	private async countTokens(text: string | vscode.LanguageModelChatMessage): Promise<number> {
-		/**
-		 * NOTE (intentional trade-off):
-		 * We use a coarse chars/4 heuristic here instead of a real tokenizer (e.g., js-tiktoken with o200k_base).
-		 * Rationale:
-		 *  - Avoid pulling multi‑MB rank files and increasing the extension install/download size.
-		 *  - Eliminate encoder lifecycle/memory concerns in long-running sessions.
-		 * Consequences:
-		 *  - This is not model-accurate and can under/over-estimate tokens, especially with tool/function calls.
-		 *  - It is “good enough” for budgeting/context checks, and we accept the inaccuracy by design.
-		 * If precise accounting becomes a requirement, reintroduce a tokenizer behind a feature flag or backend-only path.
-		 */
-		const textContent = typeof text === "string" ? text : this.extractTextFromMessage(text)
-		return Math.ceil((textContent || "").length / 4)
+		// For Claude models, use character-to-token ratio instead of VSCode LM's inaccurate counting
+		if (this.isClaudeModel()) {
+			const textContent = typeof text === "string" ? text : this.extractTextFromMessage(text)
+			// Use 4 character-to-token ratio for Claude models
+			return Math.ceil(textContent.length / 4)
+		}
+
+		// Check for required dependencies
+		if (!this.client) {
+			console.warn("Cline <Language Model API>: No client available for token counting")
+			return 0
+		}
+
+		if (!this.currentRequestCancellation) {
+			console.warn("Cline <Language Model API>: No cancellation token available for token counting")
+			return 0
+		}
+
+		// Validate input
+		if (!text) {
+			console.debug("Cline <Language Model API>: Empty text provided for token counting")
+			return 0
+		}
+
+		try {
+			// Handle different input types
+			let tokenCount: number
+
+			if (typeof text === "string") {
+				tokenCount = await this.client.countTokens(text, this.currentRequestCancellation.token)
+			} else if (text instanceof vscode.LanguageModelChatMessage) {
+				// For chat messages, ensure we have content
+				if (!text.content || (Array.isArray(text.content) && text.content.length === 0)) {
+					console.debug("Cline <Language Model API>: Empty chat message content")
+					return 0
+				}
+				tokenCount = await this.client.countTokens(text, this.currentRequestCancellation.token)
+			} else {
+				console.warn("Cline <Language Model API>: Invalid input type for token counting")
+				return 0
+			}
+
+			// Validate the result
+			if (typeof tokenCount !== "number") {
+				console.warn("Cline <Language Model API>: Non-numeric token count received:", tokenCount)
+				return 0
+			}
+
+			if (tokenCount < 0) {
+				console.warn("Cline <Language Model API>: Negative token count received:", tokenCount)
+				return 0
+			}
+
+			return tokenCount
+		} catch (error) {
+			// Handle specific error types
+			if (error instanceof vscode.CancellationError) {
+				console.debug("Cline <Language Model API>: Token counting cancelled by user")
+				return 0
+			}
+
+			const errorMessage = error instanceof Error ? error.message : "Unknown error"
+			console.warn("Cline <Language Model API>: Token counting failed:", errorMessage)
+
+			// Log additional error details if available
+			if (error instanceof Error && error.stack) {
+				console.debug("Token counting error stack:", error.stack)
+			}
+
+			return 0 // Fallback to prevent stream interruption
+		}
 	}
 
 	private async calculateTotalInputTokens(vsCodeLmMessages: vscode.LanguageModelChatMessage[]): Promise<number> {

src/api/transform/openrouter-stream.ts

@@ -24,7 +24,6 @@ export async function createOpenRouterStream(
 	// handles direct model.id match logic
 	switch (model.id) {
 		case "anthropic/claude-sonnet-4":
-		case "anthropic/claude-opus-4.1":
 		case "anthropic/claude-opus-4":
 		case "anthropic/claude-3.7-sonnet":
 		case "anthropic/claude-3.7-sonnet:beta":
@@ -83,7 +82,6 @@ export async function createOpenRouterStream(
 	let maxTokens: number | undefined
 	switch (model.id) {
 		case "anthropic/claude-sonnet-4":
-		case "anthropic/claude-opus-4.1":
 		case "anthropic/claude-opus-4":
 		case "anthropic/claude-3.7-sonnet":
 		case "anthropic/claude-3.7-sonnet:beta":
@@ -119,7 +117,6 @@ export async function createOpenRouterStream(
 	let reasoning: { max_tokens: number } | undefined = undefined
 	switch (model.id) {
 		case "anthropic/claude-sonnet-4":
-		case "anthropic/claude-opus-4.1":
 		case "anthropic/claude-opus-4":
 		case "anthropic/claude-3.7-sonnet":
 		case "anthropic/claude-3.7-sonnet:beta":
@@ -146,9 +143,6 @@ export async function createOpenRouterStream(
 	const isKimiK2 = model.id === "moonshotai/kimi-k2"
 	openRouterProviderSorting = isKimiK2 ? undefined : openRouterProviderSorting
 
-	// Force 1m context window for Claude Sonnet 4
-	const isClaudeSonnet4 = model.id === "anthropic/claude-sonnet-4"
-
 	// @ts-ignore-next-line
 	const stream = await client.chat.completions.create({
 		model: model.id,
@@ -167,8 +161,6 @@ export async function createOpenRouterStream(
 		...(isKimiK2
 			? { provider: { order: ["groq", "together", "baseten", "parasail", "novita", "deepinfra"], allow_fallbacks: false } }
 			: {}),
-		// limit providers to only those that support the 1m context window
-		...(isClaudeSonnet4 ? { provider: { order: ["anthropic", "amazon-bedrock"], allow_fallbacks: false } } : {}),
 	})
 
 	return stream

src/common.ts

@@ -1,92 +0,0 @@
-import * as vscode from "vscode"
-import {
-	migrateCustomInstructionsToGlobalRules,
-	migrateWelcomeViewCompleted,
-	migrateWorkspaceToGlobalStorage,
-} from "./core/storage/state-migrations"
-import { WebviewProvider } from "./core/webview"
-import { Logger } from "./services/logging/Logger"
-import { PostHogClientProvider } from "./services/posthog/PostHogClientProvider"
-import { WebviewProviderType } from "./shared/webview/types"
-import "./utils/path" // necessary to have access to String.prototype.toPosix
-
-import { HostProvider } from "@/hosts/host-provider"
-import { FileContextTracker } from "./core/context/context-tracking/FileContextTracker"
-import { telemetryService } from "./services/posthog/PostHogClientProvider"
-import { ShowMessageType } from "./shared/proto/host/window"
-import { getLatestAnnouncementId } from "./utils/announcements"
-/**
- * Performs intialization for Cline that is common to all platforms.
- *
- * @param context
- * @returns The webview provider
- */
-export async function initialize(context: vscode.ExtensionContext): Promise<WebviewProvider> {
-	// Initialize PostHog client provider
-	const distinctId = context.globalState.get<string>("cline.distinctId")
-	PostHogClientProvider.getInstance(distinctId)
-
-	// Migrate custom instructions to global Cline rules (one-time cleanup)
-	await migrateCustomInstructionsToGlobalRules(context)
-
-	// Migrate welcomeViewCompleted setting based on existing API keys (one-time cleanup)
-	await migrateWelcomeViewCompleted(context)
-
-	// Migrate workspace storage values back to global storage (reverting previous migration)
-	await migrateWorkspaceToGlobalStorage(context)
-
-	// Clean up orphaned file context warnings (startup cleanup)
-	await FileContextTracker.cleanupOrphanedWarnings(context)
-
-	const sidebarWebview = HostProvider.get().createWebviewProvider(WebviewProviderType.SIDEBAR)
-
-	await showVersionUpdateAnnouncement(context)
-
-	telemetryService.captureExtensionActivated()
-
-	return sidebarWebview
-}
-
-async function showVersionUpdateAnnouncement(context: vscode.ExtensionContext) {
-	// Version checking for autoupdate notification
-	const currentVersion = context.extension.packageJSON.version
-	const previousVersion = context.globalState.get<string>("clineVersion")
-	// Perform post-update actions if necessary
-	try {
-		if (!previousVersion || currentVersion !== previousVersion) {
-			Logger.log(`Cline version changed: ${previousVersion} -> ${currentVersion}. First run or update detected.`)
-
-			// Use the same condition as announcements: focus when there's a new announcement to show
-			const lastShownAnnouncementId = context.globalState.get<string>("lastShownAnnouncementId")
-			const latestAnnouncementId = getLatestAnnouncementId(context)
-
-			if (lastShownAnnouncementId !== latestAnnouncementId) {
-				// Focus Cline when there's a new announcement to show (major/minor updates or fresh installs)
-				const message = previousVersion
-					? `Cline has been updated to v${currentVersion}`
-					: `Welcome to Cline v${currentVersion}`
-				await vscode.commands.executeCommand("claude-dev.SidebarProvider.focus")
-				await new Promise((resolve) => setTimeout(resolve, 200))
-				HostProvider.window.showMessage({
-					type: ShowMessageType.INFORMATION,
-					message,
-				})
-			}
-			// Always update the main version tracker for the next launch.
-			await context.globalState.update("clineVersion", currentVersion)
-		}
-	} catch (error) {
-		const errorMessage = error instanceof Error ? error.message : String(error)
-		console.error(`Error during post-update actions: ${errorMessage}, Stack trace: ${error.stack}`)
-	}
-}
-
-/**
- * Performs cleanup when Cline is deactivated that is common to all platforms.
- */
-export async function tearDown(): Promise<void> {
-	PostHogClientProvider.getInstance().dispose()
-
-	// Dispose all webview instances
-	await WebviewProvider.disposeAllInstances()
-}

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

@@ -0,0 +1,145 @@
+import { JSONParser } from "@streamparser/json"
+import * as fs from "fs"
+import * as path from "path"
+import * as os from "os"
+
+// Fallback type definition based on the error message: "Property 'value' is optional in type 'ParsedElementInfo'"
+type ParsedElementInfo = {
+	value?: any
+	key?: string | number
+	parent?: any
+	stack?: any[]
+}
+
+export interface ReplacementItem {
+	old_string: string
+	new_string: string
+}
+
+export interface ChangeLocation {
+	startLine: number
+	endLine: number
+	startChar: number
+	endChar: number
+}
+
+export class StreamingJsonReplacer {
+	private currentFileContent: string
+	private parser: JSONParser
+	private onContentUpdated: (newContent: string, isFinalItem: boolean, changeLocation?: ChangeLocation) => void
+	private onErrorCallback: (error: Error) => void
+	private itemsProcessed: number = 0
+	private successfullyParsedItems: ReplacementItem[] = []
+
+	constructor(
+		initialContent: string,
+		onContentUpdatedCallback: (newContent: string, isFinalItem: boolean, changeLocation?: ChangeLocation) => void,
+		onErrorCallback: (error: Error) => void,
+	) {
+		// Initialize log file path
+		const timestamp = new Date().toISOString().replace(/[:.]/g, "-")
+
+		this.currentFileContent = initialContent
+		this.onContentUpdated = onContentUpdatedCallback
+		this.onErrorCallback = onErrorCallback
+
+		this.parser = new JSONParser({ paths: ["$.*"] })
+
+		this.parser.onValue = (parsedElementInfo: ParsedElementInfo) => {
+			const { value } = parsedElementInfo // Destructure to get value, which might be undefined
+
+			// This callback is triggered for each item matched by '$.replacements.*'
+			if (value && typeof value === "object" && "old_string" in value && "new_string" in value) {
+				const item = value as ReplacementItem // Value here is confirmed to be an object
+				if (typeof item.old_string === "string" && typeof item.new_string === "string") {
+					this.successfullyParsedItems.push(item) // Store the structurally valid item
+
+					if (this.currentFileContent.includes(item.old_string)) {
+						// Calculate the change location before making the replacement
+						const changeLocation = this.calculateChangeLocation(item.old_string, item.new_string)
+
+						const beforeLength = this.currentFileContent.length
+						this.currentFileContent = this.currentFileContent.replace(item.old_string, item.new_string)
+						const afterLength = this.currentFileContent.length
+
+						this.itemsProcessed++
+
+						// Notify that an item has been processed. The `isFinalItem` argument here is tricky
+						// as we don't know from the parser alone if this is the *absolute* last item
+						// until the stream ends. The caller (Task.ts) will manage the final update.
+						// For now, we'll pass `false` and let Task.ts handle the final diff view update.
+						this.onContentUpdated(this.currentFileContent, false, changeLocation)
+					} else {
+						const snippet = item.old_string.length > 50 ? item.old_string.substring(0, 47) + "..." : item.old_string
+						const error = new Error(`Streaming Replacement failed: 'old_string' not found. Snippet: "${snippet}"`)
+						this.onErrorCallback(error) // Call our own error callback
+					}
+				} else {
+					const error = new Error(`Invalid item structure in replacements stream: ${JSON.stringify(item)}`)
+					this.onErrorCallback(error) // Call our own error callback
+				}
+			}
+		}
+
+		this.parser.onError = (err: Error) => {
+			// Propagate the error to the caller via the callback
+			this.onErrorCallback(err)
+			// Note: The @streamparser/json library might throw synchronously on write if onError is not set,
+			// or if it re-throws. We'll ensure Task.ts wraps write/end in try-catch.
+		}
+	}
+
+	public write(jsonChunk: string): void {
+		try {
+			// Errors during write will be caught by the parser's onError or thrown.
+			this.parser.write(jsonChunk)
+		} catch (error) {
+			throw error
+		}
+	}
+
+	public getCurrentContent(): string {
+		return this.currentFileContent
+	}
+
+	public getSuccessfullyParsedItems(): ReplacementItem[] {
+		return [...this.successfullyParsedItems] // Return a copy
+	}
+
+	private calculateChangeLocation(oldStr: string, newStr: string): ChangeLocation {
+		// Find the index where the old string starts
+		const startIndex = this.currentFileContent.indexOf(oldStr)
+
+		if (startIndex === -1) {
+			// This shouldn't happen since we already checked includes(), but just in case
+			return { startLine: 0, endLine: 0, startChar: 0, endChar: 0 }
+		}
+
+		// Calculate line numbers by counting newlines before the start index
+		const contentBeforeStart = this.currentFileContent.substring(0, startIndex)
+
+		const startLine = (contentBeforeStart.match(/\n/g) || []).length
+		// Calculate the end index after replacement
+		const endIndex = startIndex + oldStr.length
+
+		const contentBeforeEnd = this.currentFileContent.substring(0, endIndex)
+
+		const endLine = (contentBeforeEnd.match(/\n/g) || []).length
+		// Calculate character positions within their respective lines
+		const lastNewlineBeforeStart = contentBeforeStart.lastIndexOf("\n")
+		const startChar = lastNewlineBeforeStart === -1 ? startIndex : startIndex - lastNewlineBeforeStart - 1
+
+		const lastNewlineBeforeEnd = contentBeforeEnd.lastIndexOf("\n")
+
+		const endChar = lastNewlineBeforeEnd === -1 ? endIndex : endIndex - lastNewlineBeforeEnd - 1
+
+		const result = {
+			startLine,
+			endLine,
+			startChar,
+			endChar,
+		}
+
+		return result
+	}
+}

src/core/assistant-message/index.ts

@@ -1,6 +1,6 @@
 export type AssistantMessageContent = TextContent | ToolUse
 
-export { parseAssistantMessageV2 } from "./parse-assistant-message"
+export { parseAssistantMessageV1, parseAssistantMessageV2, parseAssistantMessageV3 } from "./parse-assistant-message"
 
 export interface TextContent {
 	type: "text"
@@ -25,7 +25,6 @@ export const toolUseNames = [
 	"attempt_completion",
 	"new_task",
 	"condense",
-	"summarize_task",
 	"report_bug",
 	"new_rule",
 	"web_fetch",
@@ -61,8 +60,6 @@ export const toolParamNames = [
 	"steps_to_reproduce",
 	"api_request_output",
 	"additional_context",
-	"needs_more_exploration",
-	"task_progress",
 ] as const
 
 export type ToolParamName = (typeof toolParamNames)[number]

src/core/assistant-message/parse-assistant-message.ts

@@ -1,6 +1,245 @@
 import { AssistantMessageContent, TextContent, ToolUse, ToolParamName, toolParamNames, toolUseNames, ToolUseName } from "." // Assuming types are defined in index.ts or a similar file
 
-// parseAssistantmessageV1 removed in https://github.com/cline/cline/pull/5425
+/**
+ * @description **Version 1**
+ * Parses an assistant message string potentially containing mixed text and tool usage blocks
+ * marked with XML-like tags into an array of structured content objects.
+ *
+ * This version iterates through the message character by character, building an accumulator string.
+ * It maintains state to track whether it's currently parsing text, a tool use block, or a specific tool parameter.
+ * It detects the start and end of tool uses and parameters by checking if the accumulator ends with
+ * the corresponding opening or closing tags.
+ * Special handling is included for `write_to_file` and `new_rule` tool uses to correctly parse
+ * the `content` parameter, which might contain the closing tag itself, by looking for the *last*
+ * occurrence of the closing tag.
+ * If the input string ends mid-tag or mid-content, the last block (text or tool use) is marked as partial.
+ *
+ * @param assistantMessage The raw string output from the assistant.
+ * @returns An array of `AssistantMessageContent` objects, which can be `TextContent` or `ToolUse`.
+ *          Blocks that were not fully closed by the end of the input string will have their `partial` flag set to `true`.
+ */
+export function parseAssistantMessageV1(assistantMessage: string): AssistantMessageContent[] {
+	const contentBlocks: AssistantMessageContent[] = []
+	let currentTextContent: TextContent | undefined = undefined
+	let currentTextContentStartIndex = 0
+	let currentToolUse: ToolUse | undefined = undefined
+	let currentToolUseStartIndex = 0
+	let currentParamName: ToolParamName | undefined = undefined
+	let currentParamValueStartIndex = 0
+	let accumulator = ""
+
+	for (let i = 0; i < assistantMessage.length; i++) {
+		const char = assistantMessage[i]
+		accumulator += char
+
+		// --- State: Parsing a Tool Parameter ---
+		// there should not be a param without a tool use
+		if (currentToolUse && currentParamName) {
+			const currentParamValue = accumulator.slice(currentParamValueStartIndex)
+			const paramClosingTag = `</${currentParamName}>`
+			if (currentParamValue.endsWith(paramClosingTag)) {
+				// End of param value found
+				currentToolUse.params[currentParamName] = currentParamValue.slice(0, -paramClosingTag.length).trim()
+				currentParamName = undefined // Go back to parsing tool content or looking for next param
+				continue // Move to next character
+			} else {
+				// Partial param value is accumulating
+				continue // Move to next character
+			}
+		}
+
+		// --- State: Parsing a Tool Use (but not a specific parameter) ---
+		// no currentParamName
+		if (currentToolUse) {
+			const currentToolValue = accumulator.slice(currentToolUseStartIndex)
+			const toolUseClosingTag = `</${currentToolUse.name}>`
+
+			if (currentToolValue.endsWith(toolUseClosingTag)) {
+				// End of a tool use found
+				currentToolUse.partial = false
+				contentBlocks.push(currentToolUse)
+				currentToolUse = undefined // Go back to parsing text or looking for next tool
+				// Reset text start index in case text follows immediately
+				currentTextContentStartIndex = i + 1
+				continue // Move to next character
+			} else {
+				// Check if starting a new parameter within the current tool use
+				const possibleParamOpeningTags = toolParamNames.map((name) => `<${name}>`)
+				let foundParamStart = false
+				for (const paramOpeningTag of possibleParamOpeningTags) {
+					if (accumulator.endsWith(paramOpeningTag)) {
+						// Start of a new parameter found
+						currentParamName = paramOpeningTag.slice(1, -1) as ToolParamName
+						currentParamValueStartIndex = accumulator.length
+						foundParamStart = true
+						break
+					}
+				}
+				if (foundParamStart) {
+					continue // Move to next character
+				}
+
+				// Special case for write_to_file/new_rule content param allowing nested tags
+				// Check if a </content> tag appears, potentially indicating the end of the content param
+				// even if the main tool closing tag hasn't been seen yet.
+				const contentParamName: ToolParamName = "content"
+				if (
+					(currentToolUse.name === "write_to_file" || currentToolUse.name === "new_rule") &&
+					accumulator.endsWith(`</${contentParamName}>`)
+				) {
+					const toolContent = accumulator.slice(currentToolUseStartIndex)
+					const contentStartTag = `<${contentParamName}>`
+					const contentEndTag = `</${contentParamName}>`
+					const contentStartIndex = toolContent.indexOf(contentStartTag) + contentStartTag.length
+					// Use lastIndexOf to handle cases where </content> might appear within the content itself
+					const contentEndIndex = toolContent.lastIndexOf(contentEndTag)
+
+					// Ensure we found valid start/end tags and end is after start
+					if (
+						contentStartIndex !== -1 &&
+						contentEndIndex !== -1 &&
+						contentEndIndex > contentStartIndex - contentStartTag.length // Ensure end tag is after start tag begins
+					) {
+						// Check if this content param was already being parsed. If so, update it.
+						// If not, and we just found the closing tag, assign it.
+						// This handles cases where the </content> detection might fire before
+						// the <content> tag detection logic, or if the content is very short.
+						if (currentParamName === contentParamName) {
+							// Already parsing content, now we found the end tag
+							currentToolUse.params[contentParamName] = toolContent.slice(contentStartIndex, contentEndIndex).trim()
+							currentParamName = undefined // Finished with this param
+						} else if (currentParamName === undefined) {
+							// Not parsing a param, but found </content>. Assume it closes the content block.
+							currentToolUse.params[contentParamName] = toolContent.slice(contentStartIndex, contentEndIndex).trim()
+							// We stay in the "parsing tool use" state, looking for more params or the tool end tag.
+						}
+					}
+				}
+
+				// If none of the above, partial tool value is accumulating
+				continue // Move to next character
+			}
+		}
+
+		// --- State: Parsing Text (or looking for start of a tool use) ---
+		// no currentToolUse
+		let didStartToolUse = false
+		const possibleToolUseOpeningTags = toolUseNames.map((name) => `<${name}>`)
+		for (const toolUseOpeningTag of possibleToolUseOpeningTags) {
+			if (accumulator.endsWith(toolUseOpeningTag)) {
+				// Start of a new tool use found
+				const toolName = toolUseOpeningTag.slice(1, -1) as ToolUseName
+				currentToolUse = {
+					type: "tool_use",
+					name: toolName,
+					params: {},
+					partial: true,
+				}
+				currentToolUseStartIndex = accumulator.length
+
+				// This also indicates the end of the current text content block (if any)
+				if (currentTextContent) {
+					currentTextContent.partial = false
+					// Extract text content, removing the part that formed the tool opening tag
+					const textEndIndex = accumulator.length - toolUseOpeningTag.length
+					currentTextContent.content = accumulator.slice(currentTextContentStartIndex, textEndIndex).trim()
+					// Only add if there's actual content
+					if (currentTextContent.content.length > 0) {
+						contentBlocks.push(currentTextContent)
+					}
+					currentTextContent = undefined
+				} else {
+					// Check if there was text before this tool use started
+					const textEndIndex = accumulator.length - toolUseOpeningTag.length
+					const potentialText = accumulator.slice(currentTextContentStartIndex, textEndIndex).trim()
+					if (potentialText.length > 0) {
+						contentBlocks.push({
+							type: "text",
+							content: potentialText,
+							partial: false, // Ended because tool use started
+						})
+					}
+				}
+
+				didStartToolUse = true
+				break // Found tool start, stop checking for others
+			}
+		}
+
+		if (!didStartToolUse) {
+			// No tool use started, so it must be text content accumulating
+			// (or continuing after a closed tool use)
+			if (currentTextContent === undefined) {
+				// Start of a new text block
+				currentTextContentStartIndex = i - (accumulator.length - currentTextContentStartIndex - 1) // Adjust start index based on how much we've accumulated since the last block ended or the beginning
+				// If accumulator starts from 0, start index is i
+				if (contentBlocks.length === 0 && currentToolUse === undefined) {
+					currentTextContentStartIndex = accumulator.length - 1 // i
+				} else {
+					// Re-calculate based on the actual start of the current text segment
+					// Find the end of the last block
+					let lastBlockEndIndex = 0
+					if (contentBlocks.length > 0) {
+						const lastBlock = contentBlocks[contentBlocks.length - 1]
+						// Approximation: find where the accumulator matches the end of the message string representation of the last block. This is complex.
+						// Simpler: Assume text starts right after the last block ended implicitly at index i.
+						lastBlockEndIndex = i // Where the loop *was* when the last block finished processing
+						// Need a more robust way to track the end index of the *raw string* corresponding to the last block.
+						// Let's stick to the accumulator slice approach for simplicity in this version.
+						// The start index should be where the current *unmatched* text began.
+						let lastProcessedIndex = -1
+						if (contentBlocks.length > 0) {
+							// This requires knowing the raw string length of the previous block, which V1 doesn't explicitly track easily.
+							// We'll approximate based on the current accumulator and start index logic.
+							// The issue arises if a tool tag was just closed. accumulator contains everything up to i.
+							// lastBlockEndIndex should point to the character *after* the closing tag of the last block.
+						}
+						// Reset start index to the beginning of the *current* potential text block
+						currentTextContentStartIndex = accumulator.length - 1 // Start accumulating from the current character `i`
+					}
+
+					// If we just closed a tool, text starts *after* its closing tag
+					// The logic needs refinement here for accurate start index after a tool closure.
+					// Let's assume for now the start index logic inside the loop handles it via slicing.
+				}
+
+				currentTextContent = {
+					type: "text",
+					content: "", // Content will be filled by slicing accumulator
+					partial: true,
+				}
+			}
+			// Update text content based on the accumulator from its start index
+			currentTextContent.content = accumulator.slice(currentTextContentStartIndex).trimStart() // Trim start to avoid leading space if text follows tool
+		}
+	} // End of loop
+
+	// --- Finalization after loop ---
+
+	// If a tool use was open at the end
+	if (currentToolUse) {
+		// If a parameter was open within that tool use
+		if (currentParamName) {
+			// The remaining accumulator content belongs to this partial parameter
+			currentToolUse.params[currentParamName] = accumulator.slice(currentParamValueStartIndex).trim()
+		}
+		// Add the potentially partial tool use block
+		contentBlocks.push(currentToolUse)
+	}
+	// If text content was being accumulated at the end
+	// Note: Only one of currentToolUse or currentTextContent can be defined here,
+	// as starting a tool use finalizes the preceding text block.
+	else if (currentTextContent) {
+		// Update content one last time
+		currentTextContent.content = accumulator.slice(currentTextContentStartIndex).trim()
+		// Add the potentially partial text block only if it contains content
+		if (currentTextContent.content.length > 0) {
+			contentBlocks.push(currentTextContent)
+		}
+	}
+
+	return contentBlocks
+}
 
 /**
  * @description **Version 2**
@@ -234,3 +473,621 @@ export function parseAssistantMessageV2(assistantMessage: string): AssistantMess
 
 	return contentBlocks
 }
+
+export function parseAssistantMessageV3(assistantMessage: string): AssistantMessageContent[] {
+	const contentBlocks: AssistantMessageContent[] = []
+	let currentTextContentStart = 0 // Index where the current text block started
+	let currentTextContent: TextContent | undefined = undefined
+	let currentToolUseStart = 0 // Index *after* the opening tag of the current tool use
+	let currentToolUse: ToolUse | undefined = undefined
+	let currentParamValueStart = 0 // Index *after* the opening tag of the current param
+	let currentParamName: ToolParamName | undefined = undefined
+
+	// Precompute tags for faster lookups
+	const toolUseOpenTags = new Map<string, ToolUseName>()
+	const toolParamOpenTags = new Map<string, ToolParamName>()
+	for (const name of toolUseNames) {
+		toolUseOpenTags.set(`<${name}>`, name)
+	}
+	for (const name of toolParamNames) {
+		toolParamOpenTags.set(`<${name}>`, name)
+	}
+
+	// Function calls format detection
+	const isFunctionCallsOpen = "<function_calls>"
+	const isFunctionCallsClose = "</function_calls>"
+	const isInvokeStart = '<invoke name="'
+	const isInvokeEnd = '">'
+	const isInvokeClose = "</invoke>"
+	const isParameterStart = '<parameter name="'
+	const isParameterNameEnd = '">'
+	const isParameterClose = "</parameter>"
+
+	// Variables for function calls parsing
+	let inFunctionCalls = false
+	let currentInvokeName = ""
+	let currentParameterName = ""
+
+	const len = assistantMessage.length
+	for (let i = 0; i < len; i++) {
+		const currentCharIndex = i
+
+		// --- State: Parsing Function Calls ---
+		// Check for opening function_calls tag
+		if (
+			!inFunctionCalls &&
+			currentCharIndex >= isFunctionCallsOpen.length - 1 &&
+			assistantMessage.startsWith(isFunctionCallsOpen, currentCharIndex - isFunctionCallsOpen.length + 1)
+		) {
+			// End current text block if one was active
+			if (currentTextContent) {
+				currentTextContent.content = assistantMessage
+					.slice(currentTextContentStart, currentCharIndex - isFunctionCallsOpen.length + 1)
+					.trim()
+				currentTextContent.partial = false
+				if (currentTextContent.content.length > 0) {
+					contentBlocks.push(currentTextContent)
+				}
+				currentTextContent = undefined
+			}
+
+			inFunctionCalls = true
+			continue
+		}
+
+		// Check for invoke start within function_calls
+		if (
+			inFunctionCalls &&
+			currentInvokeName === "" &&
+			!currentToolUse && // Don't create a new tool if we already have one
+			currentCharIndex >= isInvokeStart.length - 1 &&
+			assistantMessage.startsWith(isInvokeStart, currentCharIndex - isInvokeStart.length + 1)
+		) {
+			// Find the end of the invoke name
+			const nameEndPos = assistantMessage.indexOf(isInvokeEnd, currentCharIndex + 1)
+			if (nameEndPos !== -1) {
+				// Extract the invoke name
+				currentInvokeName = assistantMessage.slice(currentCharIndex + 1, nameEndPos)
+				i = nameEndPos + isInvokeEnd.length - 1 // Skip to after the '">
+
+				// If this is an LS invoke, create a list_files tool
+				if (currentInvokeName === "LS") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "list_files",
+						params: {},
+						partial: true,
+					}
+				}
+
+				// If this is a Grep invoke, create a search_files tool
+				if (currentInvokeName === "Grep") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "search_files",
+						params: {},
+						partial: true,
+					}
+				}
+
+				if (currentInvokeName === "Bash") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "execute_command",
+						params: {},
+						partial: true,
+					}
+				}
+
+				if (currentInvokeName === "Read") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "read_file",
+						params: {},
+						partial: true,
+					}
+				}
+
+				if (currentInvokeName === "Write") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "write_to_file",
+						params: {},
+						partial: true,
+					}
+				}
+
+				if (currentInvokeName === "WebFetch") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "web_fetch",
+						params: {},
+						partial: true,
+					}
+				}
+
+				if (currentInvokeName === "AskQuestion") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "ask_followup_question",
+						params: {},
+						partial: true,
+					}
+				}
+
+				if (currentInvokeName === "UseMCPTool") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "use_mcp_tool",
+						params: {},
+						partial: true,
+					}
+				}
+
+				if (currentInvokeName === "AccessMCPResource") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "access_mcp_resource",
+						params: {},
+						partial: true,
+					}
+				}
+
+				if (currentInvokeName === "ListCodeDefinitionNames") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "list_code_definition_names",
+						params: {},
+						partial: true,
+					}
+				}
+
+				if (currentInvokeName === "PlanModeRespond") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "plan_mode_respond",
+						params: {},
+						partial: true,
+					}
+				}
+
+				if (currentInvokeName === "LoadMcpDocumentation") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "load_mcp_documentation",
+						params: {},
+						partial: true,
+					}
+				}
+
+				if (currentInvokeName === "AttemptCompletion") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "attempt_completion",
+						params: {},
+						partial: true,
+					}
+				}
+
+				if (currentInvokeName === "BrowserAction") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "browser_action",
+						params: {},
+						partial: true,
+					}
+				}
+
+				if (currentInvokeName === "NewTask") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "new_task",
+						params: {},
+						partial: true,
+					}
+				}
+
+				// If this is a MultiEdit invoke, create a replace_in_file tool
+				if (currentInvokeName === "MultiEdit") {
+					currentToolUse = {
+						type: "tool_use",
+						name: "replace_in_file",
+						params: {},
+						partial: true,
+					}
+				}
+
+				continue
+			}
+		}
+
+		// Check for parameter start within invoke
+		if (
+			inFunctionCalls &&
+			currentInvokeName !== "" &&
+			currentParameterName === "" &&
+			currentCharIndex >= isParameterStart.length - 1 &&
+			assistantMessage.startsWith(isParameterStart, currentCharIndex - isParameterStart.length + 1)
+		) {
+			// Find the end of the parameter name
+			const nameEndPos = assistantMessage.indexOf(isParameterNameEnd, currentCharIndex + 1)
+			if (nameEndPos !== -1) {
+				// Extract the parameter name
+				currentParameterName = assistantMessage.slice(currentCharIndex + 1, nameEndPos)
+				currentParamValueStart = nameEndPos + isParameterNameEnd.length
+				i = nameEndPos + isParameterNameEnd.length - 1 // Skip to after the '">'
+				continue
+			}
+		}
+
+		// Check for parameter end
+		if (
+			inFunctionCalls &&
+			currentInvokeName !== "" &&
+			currentParameterName !== "" &&
+			currentCharIndex >= isParameterClose.length - 1 &&
+			assistantMessage.startsWith(isParameterClose, currentCharIndex - isParameterClose.length + 1)
+		) {
+			// Extract parameter value
+			const value = assistantMessage.slice(currentParamValueStart, currentCharIndex - isParameterClose.length + 1).trim()
+
+			// Map parameter to tool params
+			if (currentToolUse && currentInvokeName === "LS" && currentParameterName === "path") {
+				currentToolUse.params["path"] = value
+				// Default recursive to false - only show top level
+				currentToolUse.params["recursive"] = "false"
+			}
+
+			if (currentToolUse && currentInvokeName === "Read" && currentParameterName === "file_path") {
+				currentToolUse.params["path"] = value
+			}
+
+			if (currentToolUse && currentInvokeName === "PlanModeRespond" && currentParameterName === "response") {
+				currentToolUse.params["response"] = value
+			}
+
+			if (currentToolUse && currentInvokeName === "WebFetch" && currentParameterName === "url") {
+				currentToolUse.params["url"] = value
+			}
+
+			if (currentToolUse && currentInvokeName === "ListCodeDefinitionNames" && currentParameterName === "path") {
+				currentToolUse.params["path"] = value
+			}
+
+			if (currentToolUse && currentInvokeName === "NewTask" && currentParameterName === "context") {
+				currentToolUse.params["context"] = value
+			}
+
+			// Map parameter to tool params for Grep
+			if (currentToolUse && currentInvokeName === "Grep") {
+				if (currentParameterName === "pattern") {
+					currentToolUse.params["regex"] = value
+				} else if (currentParameterName === "path") {
+					currentToolUse.params["path"] = value
+				} else if (currentParameterName === "include") {
+					currentToolUse.params["file_pattern"] = value
+				}
+			}
+
+			if (currentToolUse && currentInvokeName === "Bash") {
+				if (currentParameterName === "command") {
+					currentToolUse.params["command"] = value
+				} else if (currentParameterName === "requires_approval") {
+					currentToolUse.params["requires_approval"] = value === "true" ? "true" : "false"
+				}
+			}
+
+			if (currentToolUse && currentInvokeName === "Write") {
+				if (currentParameterName === "file_path") {
+					currentToolUse.params["path"] = value
+				} else if (currentParameterName === "content") {
+					currentToolUse.params["content"] = value
+				}
+			}
+
+			if (currentToolUse && currentInvokeName === "AskQuestion") {
+				if (currentParameterName === "question") {
+					currentToolUse.params["question"] = value
+				} else if (currentParameterName === "options") {
+					currentToolUse.params["options"] = value
+				}
+			}
+
+			if (currentToolUse && currentInvokeName === "UseMCPTool") {
+				if (currentParameterName === "server_name") {
+					currentToolUse.params["server_name"] = value
+				} else if (currentParameterName === "tool_name") {
+					currentToolUse.params["tool_name"] = value
+				} else if (currentParameterName === "arguments") {
+					currentToolUse.params["arguments"] = value
+				}
+			}
+
+			if (currentToolUse && currentInvokeName === "AccessMCPResource") {
+				if (currentParameterName === "server_name") {
+					currentToolUse.params["server_name"] = value
+				} else if (currentParameterName === "uri") {
+					currentToolUse.params["uri"] = value
+				}
+			}
+
+			if (currentToolUse && currentInvokeName === "AttemptCompletion") {
+				if (currentParameterName === "result") {
+					currentToolUse.params["result"] = value
+				}
+				if (currentParameterName === "command") {
+					currentToolUse.params["command"] = value
+				}
+			}
+
+			if (currentToolUse && currentInvokeName === "BrowserAction") {
+				if (currentParameterName === "action") {
+					currentToolUse.params["action"] = value
+				} else if (currentParameterName === "url") {
+					currentToolUse.params["url"] = value
+				} else if (currentParameterName === "coordinate") {
+					currentToolUse.params["coordinate"] = value
+				} else if (currentParameterName === "text") {
+					currentToolUse.params["text"] = value
+				}
+			}
+
+			// Map parameter to tool params for MultiEdit
+			if (currentToolUse && currentInvokeName === "MultiEdit") {
+				if (currentParameterName === "file_path") {
+					currentToolUse.params["path"] = value
+				} else if (currentParameterName === "edits") {
+					// Save the value to the diff parameter for replace_in_file
+					currentToolUse.params["diff"] = value
+				}
+			}
+
+			currentParameterName = ""
+			continue
+		}
+
+		// Check for invoke end
+		if (
+			inFunctionCalls &&
+			currentInvokeName !== "" &&
+			currentCharIndex >= isInvokeClose.length - 1 &&
+			assistantMessage.startsWith(isInvokeClose, currentCharIndex - isInvokeClose.length + 1)
+		) {
+			// If we have a tool use from this invoke, finalize it
+			if (
+				currentToolUse &&
+				(currentInvokeName === "LS" ||
+					currentInvokeName === "Grep" ||
+					currentInvokeName === "Bash" ||
+					currentInvokeName === "Read" ||
+					currentInvokeName === "Write" ||
+					currentInvokeName === "WebFetch" ||
+					currentInvokeName === "AskQuestion" ||
+					currentInvokeName === "UseMCPTool" ||
+					currentInvokeName === "AccessMCPResource" ||
+					currentInvokeName === "ListCodeDefinitionNames" ||
+					currentInvokeName === "PlanModeRespond" ||
+					currentInvokeName === "LoadMcpDocumentation" ||
+					currentInvokeName === "AttemptCompletion" ||
+					currentInvokeName === "BrowserAction" ||
+					currentInvokeName === "NewTask" ||
+					currentInvokeName === "MultiEdit")
+			) {
+				currentToolUse.partial = false
+				contentBlocks.push(currentToolUse)
+				currentToolUse = undefined
+			}
+			currentInvokeName = ""
+			continue
+		}
+
+		// Check for function_calls end
+		if (
+			inFunctionCalls &&
+			currentCharIndex >= isFunctionCallsClose.length - 1 &&
+			assistantMessage.startsWith(isFunctionCallsClose, currentCharIndex - isFunctionCallsClose.length + 1)
+		) {
+			inFunctionCalls = false
+			currentTextContentStart = currentCharIndex + 1
+			// Start a new text content block for any text after function_calls
+			currentTextContent = {
+				type: "text",
+				content: "",
+				partial: true,
+			}
+			continue
+		}
+
+		// Skip normal parsing when inside function_calls
+		if (inFunctionCalls) {
+			continue
+		}
+
+		// --- State: Parsing a Tool Parameter ---
+		if (currentToolUse && currentParamName) {
+			const closeTag = `</${currentParamName}>`
+			// Check if the string *ending* at index `i` matches the closing tag
+			if (
+				currentCharIndex >= closeTag.length - 1 &&
+				assistantMessage.startsWith(
+					closeTag,
+					currentCharIndex - closeTag.length + 1, // Start checking from potential start of tag
+				)
+			) {
+				// Found the closing tag for the parameter
+				const value = assistantMessage
+					.slice(
+						currentParamValueStart, // Start after the opening tag
+						currentCharIndex - closeTag.length + 1, // End before the closing tag
+					)
+					.trim()
+				currentToolUse.params[currentParamName] = value
+				currentParamName = undefined // Go back to parsing tool content
+				// We don't continue loop here, need to check for tool close or other params at index i
+			} else {
+				continue // Still inside param value, move to next char
+			}
+		}
+
+		// --- State: Parsing a Tool Use (but not a specific parameter) ---
+		if (currentToolUse && !currentParamName) {
+			// Ensure we are not inside a parameter already
+			// Check if starting a new parameter
+			let startedNewParam = false
+			for (const [tag, paramName] of toolParamOpenTags.entries()) {
+				if (currentCharIndex >= tag.length - 1 && assistantMessage.startsWith(tag, currentCharIndex - tag.length + 1)) {
+					currentParamName = paramName
+					currentParamValueStart = currentCharIndex + 1 // Value starts after the tag
+					startedNewParam = true
+					break
+				}
+			}
+			if (startedNewParam) {
+				continue // Handled start of param, move to next char
+			}
+
+			// Check if closing the current tool use
+			const toolCloseTag = `</${currentToolUse.name}>`
+			if (
+				currentCharIndex >= toolCloseTag.length - 1 &&
+				assistantMessage.startsWith(toolCloseTag, currentCharIndex - toolCloseTag.length + 1)
+			) {
+				// End of the tool use found
+				// Special handling for content params *before* finalizing the tool
+				const toolContentSlice = assistantMessage.slice(
+					currentToolUseStart, // From after the tool opening tag
+					currentCharIndex - toolCloseTag.length + 1, // To before the tool closing tag
+				)
+
+				// Check if content parameter needs special handling (write_to_file/new_rule)
+				// This check is important if the closing </content> tag was missed by the parameter parsing logic
+				// (e.g., if content is empty or parsing logic prioritizes tool close)
+				const contentParamName: ToolParamName = "content"
+				if (
+					currentToolUse.name === "write_to_file" /* || currentToolUse.name === "new_rule" */ &&
+					toolContentSlice.includes(`<${contentParamName}>`)
+				) {
+					const contentStartTag = `<${contentParamName}>`
+					const contentEndTag = `</${contentParamName}>`
+					const contentStart = toolContentSlice.indexOf(contentStartTag)
+					// Use lastIndexOf for robustness against nested tags
+					const contentEnd = toolContentSlice.lastIndexOf(contentEndTag)
+
+					if (contentStart !== -1 && contentEnd !== -1 && contentEnd > contentStart) {
+						const contentValue = toolContentSlice.slice(contentStart + contentStartTag.length, contentEnd).trim()
+						currentToolUse.params[contentParamName] = contentValue
+					}
+				}
+
+				currentToolUse.partial = false // Mark as complete
+				contentBlocks.push(currentToolUse)
+				currentToolUse = undefined // Reset state
+				currentTextContentStart = currentCharIndex + 1 // Potential text starts after this tag
+				continue // Move to next char
+			}
+			// If not starting a param and not closing the tool, continue accumulating tool content implicitly
+			continue
+		}
+
+		// --- State: Parsing Text / Looking for Tool Start ---
+		if (!currentToolUse) {
+			// Check if starting a new tool use
+			let startedNewTool = false
+			for (const [tag, toolName] of toolUseOpenTags.entries()) {
+				if (currentCharIndex >= tag.length - 1 && assistantMessage.startsWith(tag, currentCharIndex - tag.length + 1)) {
+					// End current text block if one was active
+					if (currentTextContent) {
+						currentTextContent.content = assistantMessage
+							.slice(
+								currentTextContentStart, // From where text started
+								currentCharIndex - tag.length + 1, // To before the tool tag starts
+							)
+							.trim()
+						currentTextContent.partial = false // Ended because tool started
+						if (currentTextContent.content.length > 0) {
+							contentBlocks.push(currentTextContent)
+						}
+						currentTextContent = undefined
+					} else {
+						// Check for any text between the last block and this tag
+						const potentialText = assistantMessage
+							.slice(
+								currentTextContentStart, // From where text *might* have started
+								currentCharIndex - tag.length + 1, // To before the tool tag starts
+							)
+							.trim()
+						if (potentialText.length > 0) {
+							contentBlocks.push({
+								type: "text",
+								content: potentialText,
+								partial: false,
+							})
+						}
+					}
+
+					// Start the new tool use
+					currentToolUse = {
+						type: "tool_use",
+						name: toolName,
+						params: {},
+						partial: true, // Assume partial until closing tag is found
+					}
+					currentToolUseStart = currentCharIndex + 1 // Tool content starts after the opening tag
+					startedNewTool = true
+					break
+				}
+			}
+
+			if (startedNewTool) {
+				continue // Handled start of tool, move to next char
+			}
+
+			// If not starting a tool, it must be text content
+			if (!currentTextContent) {
+				// Start a new text block if we aren't already in one
+				currentTextContentStart = currentCharIndex // Text starts at the current character
+				// Check if the current char is the start of potential text *immediately* after a tag
+				// This needs the previous state - simpler to let slicing handle it later.
+				// Resetting start index accurately is key.
+				// It should be the index *after* the last processed tag.
+				// The logic managing currentTextContentStart after closing tags handles this.
+
+				currentTextContent = {
+					type: "text",
+					content: "", // Will be determined by slicing at the end or when a tool starts
+					partial: true,
+				}
+			}
+			// Continue accumulating text implicitly; content is extracted later.
+		}
+	} // End of loop
+
+	// --- Finalization after loop ---
+
+	// Finalize any open parameter within an open tool use
+	if (currentToolUse && currentParamName) {
+		currentToolUse.params[currentParamName] = assistantMessage
+			.slice(currentParamValueStart) // From param start to end of string
+			.trim()
+		// Tool use remains partial
+	}
+
+	// Finalize any open tool use (which might contain the finalized partial param)
+	if (currentToolUse) {
+		// Tool use is partial because the loop finished before its closing tag
+		contentBlocks.push(currentToolUse)
+	}
+	// Finalize any trailing text content
+	// Only possible if a tool use wasn't open at the very end
+	else if (currentTextContent) {
+		currentTextContent.content = assistantMessage
+			.slice(currentTextContentStart) // From text start to end of string
+			.trim()
+		// Text is partial because the loop finished
+		if (currentTextContent.content.length > 0) {
+			contentBlocks.push(currentTextContent)
+		}
+	}
+
+	return contentBlocks
+}

src/core/context/context-management/ContextManager.ts

@@ -106,24 +106,7 @@ export class ContextManager {
 	}
 
 	/**
-	 * Determine whether we should compact context window, based on token counts
-	 */
-	shouldCompactContextWindow(clineMessages: ClineMessage[], api: ApiHandler, previousApiReqIndex: number): boolean {
-		if (previousApiReqIndex >= 0) {
-			const previousRequest = clineMessages[previousApiReqIndex]
-			if (previousRequest && previousRequest.text) {
-				const { tokensIn, tokensOut, cacheWrites, cacheReads }: ClineApiReqInfo = JSON.parse(previousRequest.text)
-				const totalTokens = (tokensIn || 0) + (tokensOut || 0) + (cacheWrites || 0) + (cacheReads || 0)
-
-				const { maxAllowedSize } = getContextWindowInfo(api)
-				return totalTokens >= maxAllowedSize
-			}
-		}
-		return false
-	}
-
-	/**
-	 * primary entry point for getting up to date context
+	 * primary entry point for getting up to date context & truncating when required
 	 */
 	async getNewContextMessagesAndMetadata(
 		apiConversationHistory: Anthropic.Messages.MessageParam[],
@@ -135,6 +118,63 @@ export class ContextManager {
 	) {
 		let updatedConversationHistoryDeletedRange = false
 
+		// If the previous API request's total token usage is close to the context window, truncate the conversation history to free up space for the new request
+		if (previousApiReqIndex >= 0) {
+			const previousRequest = clineMessages[previousApiReqIndex]
+			if (previousRequest && previousRequest.text) {
+				const timestamp = previousRequest.ts
+				const { tokensIn, tokensOut, cacheWrites, cacheReads }: ClineApiReqInfo = JSON.parse(previousRequest.text)
+				const totalTokens = (tokensIn || 0) + (tokensOut || 0) + (cacheWrites || 0) + (cacheReads || 0)
+				const { maxAllowedSize } = getContextWindowInfo(api)
+
+				// This is the most reliable way to know when we're close to hitting the context window.
+				if (totalTokens >= maxAllowedSize) {
+					// Since the user may switch between models with different context windows, truncating half may not be enough (ie if switching from claude 200k to deepseek 64k, half truncation will only remove 100k tokens, but we need to remove much more)
+					// So if totalTokens/2 is greater than maxAllowedSize, we truncate 3/4 instead of 1/2
+					const keep = totalTokens / 2 > maxAllowedSize ? "quarter" : "half"
+
+					// we later check how many chars we trim to determine if we should still truncate history
+					let [anyContextUpdates, uniqueFileReadIndices] = this.applyContextOptimizations(
+						apiConversationHistory,
+						conversationHistoryDeletedRange ? conversationHistoryDeletedRange[1] + 1 : 2,
+						timestamp,
+					)
+
+					let needToTruncate = true
+					if (anyContextUpdates) {
+						// determine whether we've saved enough chars to not truncate
+						const charactersSavedPercentage = this.calculateContextOptimizationMetrics(
+							apiConversationHistory,
+							conversationHistoryDeletedRange,
+							uniqueFileReadIndices,
+						)
+						if (charactersSavedPercentage >= 0.3) {
+							needToTruncate = false
+						}
+					}
+
+					if (needToTruncate) {
+						// go ahead with truncation
+						anyContextUpdates = this.applyStandardContextTruncationNoticeChange(timestamp) || anyContextUpdates
+
+						// NOTE: it's okay that we overwriteConversationHistory in resume task since we're only ever removing the last user message and not anything in the middle which would affect this range
+						conversationHistoryDeletedRange = this.getNextTruncationRange(
+							apiConversationHistory,
+							conversationHistoryDeletedRange,
+							keep,
+						)
+
+						updatedConversationHistoryDeletedRange = true
+					}
+
+					// if we alter the context history, save the updated version to disk
+					if (anyContextUpdates) {
+						await this.saveContextHistory(taskDirectory)
+					}
+				}
+			}
+		}
+
 		const truncatedConversationHistory = this.getAndAlterTruncatedMessages(
 			apiConversationHistory,
 			conversationHistoryDeletedRange,

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

@@ -1,68 +1,15 @@
-import LengthFinishReasonError, { APIError } from "openai"
-
-export function checkContextWindowExceededError(error: unknown): boolean {
-	return (
-		checkIsOpenAIContextWindowError(error) ||
-		checkIsOpenRouterContextWindowError(error) ||
-		checkIsAnthropicContextWindowError(error) ||
-		checkIsCerebrasContextWindowError(error)
-	)
-}
-
-function checkIsOpenRouterContextWindowError(error: any): boolean {
+export function checkIsOpenRouterContextWindowError(error: any): boolean {
 	try {
-		const status = error?.status ?? error?.code ?? error?.error?.status ?? error?.response?.status
-		const message: string = String(error?.message || error?.error?.message || "")
-
-		// Known OpenAI/OpenRouter-style signal (code 400 and message includes "context length")
-		const CONTEXT_ERROR_PATTERNS = [
-			/\bcontext\s*(?:length|window)\b/i,
-			/\bmaximum\s*context\b/i,
-			/\b(?:input\s*)?tokens?\s*exceed/i,
-			/\btoo\s*many\s*tokens?\b/i,
-		] as const
-
-		return String(status) === "400" && CONTEXT_ERROR_PATTERNS.some((pattern) => pattern.test(message))
-	} catch {
+		return error.code === 400 && error.message?.includes("context length")
+	} catch (e: unknown) {
 		return false
 	}
 }
 
-// Docs: https://platform.openai.com/docs/guides/error-codes/api-errors
-function checkIsOpenAIContextWindowError(error: unknown): boolean {
-	try {
-		if (error instanceof LengthFinishReasonError) {
-			return true
-		}
-
-		const KNOWN_CONTEXT_ERROR_SUBSTRINGS = ["token", "context length"] as const
-
-		return (
-			Boolean(error) &&
-			error instanceof APIError &&
-			error.code?.toString() === "400" &&
-			KNOWN_CONTEXT_ERROR_SUBSTRINGS.some((substring) => error.message.includes(substring))
-		)
-	} catch {
-		return false
-	}
-}
-
-function checkIsAnthropicContextWindowError(response: any): boolean {
+export function checkIsAnthropicContextWindowError(response: any): boolean {
 	try {
 		return response?.error?.error?.type === "invalid_request_error"
-	} catch {
-		return false
-	}
-}
-
-function checkIsCerebrasContextWindowError(response: any): boolean {
-	try {
-		const status = response?.status ?? response?.code ?? response?.error?.status ?? response?.response?.status
-		const message: string = String(response?.message || response?.error?.message || "")
-
-		return String(status) === "400" && message.includes("Please reduce the length of the messages or completion")
-	} catch {
+	} catch (e: unknown) {
 		return false
 	}
 }

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

@@ -1,22 +1,20 @@
-import { HostProvider } from "@/hosts/host-provider"
-import { setVscodeHostProviderMock } from "@/test/host-provider-test-utils"
-import * as diskModule from "@core/storage/disk"
+import { describe, it, beforeEach, afterEach } from "mocha"
 import { expect } from "chai"
-import { afterEach, beforeEach, describe, it } from "mocha"
-import * as path from "path"
 import * as sinon from "sinon"
 import * as vscode from "vscode"
-import chokidar from "chokidar"
-import type { FileMetadataEntry, TaskMetadata } from "./ContextTrackerTypes"
+import * as path from "path"
 import { FileContextTracker } from "./FileContextTracker"
-import { Controller } from "@/core/controller"
+import * as diskModule from "@core/storage/disk"
+import type { TaskMetadata, FileMetadataEntry } from "./ContextTrackerTypes"
+import type { DiffViewProviderCreator, WebviewProviderCreator } from "@/hosts/host-provider"
+import { HostProvider } from "@/hosts/host-provider"
+import { vscodeHostBridgeClient } from "@/hosts/vscode/hostbridge/client/host-grpc-client"
 
 describe("FileContextTracker", () => {
 	let sandbox: sinon.SinonSandbox
-	let mockController: Controller
+	let mockContext: vscode.ExtensionContext
 	let mockWorkspace: sinon.SinonStub
 	let mockFileSystemWatcher: any
-	let chokidarWatchStub: sinon.SinonStub
 	let tracker: FileContextTracker
 	let taskId: string
 	let mockTaskMetadata: TaskMetadata
@@ -35,32 +33,39 @@ describe("FileContextTracker", () => {
 			} as vscode.WorkspaceFolder,
 		])
 
-		// Mock chokidar file watcher
+		// Mock file system watcher
 		mockFileSystemWatcher = {
-			close: sandbox.stub().resolves(),
-			on: sandbox.stub(),
+			dispose: sandbox.stub(),
+			onDidChange: sandbox.stub().returns({ dispose: () => {} }),
 		}
-		// Return the watcher itself for chaining
-		mockFileSystemWatcher.on.returns(mockFileSystemWatcher)
 
-		// Stub chokidar.watch to return our mock watcher
-		chokidarWatchStub = sandbox.stub(chokidar, "watch").returns(mockFileSystemWatcher as any)
+		// Use a function replacement instead of a direct stub
+		vscode.workspace.createFileSystemWatcher = function () {
+			return mockFileSystemWatcher
+		}
 
 		// Mock controller and context
-		mockController = {
-			context: { globalStorageUri: { fsPath: "/mock/storage" } } as vscode.ExtensionContext,
-		} as unknown as Controller
+		mockContext = {
+			globalStorageUri: { fsPath: "/mock/storage" },
+		} as unknown as vscode.ExtensionContext
 
 		// Mock disk module functions
 		mockTaskMetadata = { files_in_context: [], model_usage: [] }
 		getTaskMetadataStub = sandbox.stub(diskModule, "getTaskMetadata").resolves(mockTaskMetadata)
 		saveTaskMetadataStub = sandbox.stub(diskModule, "saveTaskMetadata").resolves()
 
-		setVscodeHostProviderMock()
+		// Reset HostProvider before initializing to avoid "already initialized" errors
+		HostProvider.reset()
+		HostProvider.initialize(
+			((_) => {}) as WebviewProviderCreator,
+			(() => {}) as DiffViewProviderCreator,
+			vscodeHostBridgeClient,
+			(_) => {},
+		)
 
 		// Create tracker instance
 		taskId = "test-task-id"
-		tracker = new FileContextTracker(mockController, taskId)
+		tracker = new FileContextTracker(mockContext, taskId)
 	})
 
 	afterEach(() => {
@@ -189,13 +194,17 @@ describe("FileContextTracker", () => {
 	it("should setup a file watcher for tracked files", async () => {
 		const filePath = "src/test-file.ts"
 
+		// Create a spy to track if createFileSystemWatcher was called
+		const createWatcherSpy = sinon.spy(vscode.workspace, "createFileSystemWatcher")
+
 		await tracker.trackFileContext(filePath, "read_tool")
 
-		// Verify chokidar.watch was called
-		expect(chokidarWatchStub.called).to.be.true
+		// Verify createFileSystemWatcher was called
+		expect(createWatcherSpy.called).to.be.true
+		createWatcherSpy.restore()
 
-		// Verify change listener was set up
-		expect(mockFileSystemWatcher.on.called).to.be.true
+		// Verify onDidChange was called to set up the change listener
+		expect(mockFileSystemWatcher.onDidChange.called).to.be.true
 	})
 
 	it("should track user edits when file watcher detects changes", async () => {
@@ -211,8 +220,8 @@ describe("FileContextTracker", () => {
 		// Create a spy on trackFileContext to verify it's called with the right parameters
 		const trackFileContextSpy = sandbox.spy(tracker, "trackFileContext")
 
-		// Get the callback that was registered with chokidar "change" event
-		const callback = mockFileSystemWatcher.on.firstCall.args[1]
+		// Get the callback that was registered with onDidChange
+		const callback = mockFileSystemWatcher.onDidChange.firstCall.args[0]
 
 		// Directly call the callback to simulate a file change event
 		callback(vscode.Uri.file(path.resolve("/mock/workspace", filePath)))
@@ -241,8 +250,8 @@ describe("FileContextTracker", () => {
 		// Create a spy on trackFileContext to verify it's not called
 		const trackFileContextSpy = sandbox.spy(tracker, "trackFileContext")
 
-		// Get the callback that was registered with chokidar "change" event
-		const callback = mockFileSystemWatcher.on.firstCall.args[1]
+		// Get the callback that was registered with onDidChange
+		const callback = mockFileSystemWatcher.onDidChange.firstCall.args[0]
 
 		// Directly call the callback to simulate a file change event
 		callback(vscode.Uri.file(path.resolve("/mock/workspace", filePath)))
@@ -262,9 +271,9 @@ describe("FileContextTracker", () => {
 		await tracker.trackFileContext(filePath, "read_tool")
 
 		// Call dispose
-		await tracker.dispose()
+		tracker.dispose()
 
-		// Verify the watcher was closed
-		expect(mockFileSystemWatcher.close.called).to.be.true
+		// Verify the watcher was disposed
+		expect(mockFileSystemWatcher.dispose.called).to.be.true
 	})
 })

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

@@ -1,12 +1,12 @@
 import * as path from "path"
 import * as vscode from "vscode"
-import chokidar, { FSWatcher } from "chokidar"
 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"
+import { HostProvider } from "@/hosts/host-provider"
 import { getCwd } from "@/utils/path"
-import { HistoryItem } from "@/shared/HistoryItem"
-import { Controller } from "@/core/controller"
 
 // 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.
@@ -23,16 +23,16 @@ If a file is modified outside of Cline, we detect and track this change to preve
 This is used when restoring a task (non-git "checkpoint" restore), and mid-task.
 */
 export class FileContextTracker {
-	private controller: Controller
+	private context: vscode.ExtensionContext
 	readonly taskId: string
 
 	// File tracking and watching
-	private fileWatchers = new Map<string, FSWatcher>()
+	private fileWatchers = new Map<string, vscode.FileSystemWatcher>()
 	private recentlyModifiedFiles = new Set<string>()
 	private recentlyEditedByCline = new Set<string>()
 
-	constructor(controller: Controller, taskId: string) {
-		this.controller = controller
+	constructor(context: vscode.ExtensionContext, taskId: string) {
+		this.context = context
 		this.taskId = taskId
 	}
 
@@ -51,21 +51,14 @@ export class FileContextTracker {
 			return
 		}
 
-		// Create a chokidar file watcher for this specific file
-		const resolvedFilePath = path.resolve(cwd, filePath)
-		const watcher = chokidar.watch(resolvedFilePath, {
-			persistent: true, // Keep process alive while watching
-			ignoreInitial: true, // Don't emit events for existing files on startup
-			atomic: true, // Handle atomic writes (editors that use temp files)
-			awaitWriteFinish: {
-				// Wait for writes to finish before emitting events
-				stabilityThreshold: 100, // Wait 100ms for file size to stabilize
-				pollInterval: 100, // Check every 100ms while waiting
-			},
-		})
+		// Create a file system watcher for this specific file
+		const fileUri = vscode.Uri.file(path.resolve(cwd, filePath))
+		const watcher = vscode.workspace.createFileSystemWatcher(
+			new vscode.RelativePattern(path.dirname(fileUri.fsPath), path.basename(fileUri.fsPath)),
+		)
 
 		// Track file changes
-		watcher.on("change", () => {
+		watcher.onDidChange(() => {
 			if (this.recentlyEditedByCline.has(filePath)) {
 				this.recentlyEditedByCline.delete(filePath) // This was an edit by Cline, no need to inform Cline
 			} else {
@@ -91,7 +84,7 @@ export class FileContextTracker {
 			}
 
 			// Add file to metadata
-			await this.addFileToFileContextTracker(this.controller.context, this.taskId, filePath, operation)
+			await this.addFileToFileContextTracker(this.context, this.taskId, filePath, operation)
 
 			// Set up file watcher for this file
 			await this.setupFileWatcher(filePath)
@@ -186,9 +179,10 @@ export class FileContextTracker {
 	/**
 	 * Disposes all file watchers
 	 */
-	async dispose(): Promise<void> {
-		const closePromises = Array.from(this.fileWatchers.values()).map((watcher) => watcher.close())
-		await Promise.all(closePromises)
+	dispose(): void {
+		for (const watcher of this.fileWatchers.values()) {
+			watcher.dispose()
+		}
 		this.fileWatchers.clear()
 	}
 
@@ -201,7 +195,7 @@ export class FileContextTracker {
 
 		try {
 			// Check task metadata for files that were edited by Cline or users after the message timestamp
-			const taskMetadata = await getTaskMetadata(this.controller.context, this.taskId)
+			const taskMetadata = await getTaskMetadata(this.context, this.taskId)
 
 			if (taskMetadata?.files_in_context) {
 				for (const fileEntry of taskMetadata.files_in_context) {
@@ -243,7 +237,7 @@ export class FileContextTracker {
 			const key = `pendingFileContextWarning_${this.taskId}`
 			// NOTE: Using 'as any' because dynamic keys like pendingFileContextWarning_${taskId}
 			// are legitimate workspace state keys but don't fit the strict LocalStateKey type system
-			this.controller.cacheService.setWorkspaceState(key as any, files)
+			await updateWorkspaceState(this.context, key as any, files)
 		} catch (error) {
 			console.error("Error storing pending file context warning:", error)
 		}
@@ -255,7 +249,7 @@ export class FileContextTracker {
 	async retrievePendingFileContextWarning(): Promise<string[] | undefined> {
 		try {
 			const key = `pendingFileContextWarning_${this.taskId}`
-			const files = this.controller.cacheService.getWorkspaceStateKey(key as any) as string[]
+			const files = (await getWorkspaceState(this.context, key as any)) as string[]
 			return files
 		} catch (error) {
 			console.error("Error retrieving pending file context warning:", error)
@@ -270,7 +264,7 @@ export class FileContextTracker {
 		try {
 			const files = await this.retrievePendingFileContextWarning()
 			if (files) {
-				this.controller.cacheService.setWorkspaceState(`pendingFileContextWarning_${this.taskId}` as any, undefined)
+				await updateWorkspaceState(this.context, `pendingFileContextWarning_${this.taskId}` as any, undefined)
 				return files
 			}
 		} catch (error) {
@@ -286,8 +280,7 @@ export class FileContextTracker {
 	static async cleanupOrphanedWarnings(context: vscode.ExtensionContext): Promise<void> {
 		const startTime = Date.now()
 		try {
-			// eslint-disable-next-line eslint-rules/no-direct-vscode-state-api
-			const taskHistory = (context.globalState.get("taskHistory") as HistoryItem[]) || []
+			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_"))
@@ -302,8 +295,7 @@ export class FileContextTracker {
 
 			if (orphanedPendingContextTasks.length > 0) {
 				for (const key of orphanedPendingContextTasks) {
-					// eslint-disable-next-line eslint-rules/no-direct-vscode-state-api
-					await context.workspaceState.update(key, undefined)
+					await updateWorkspaceState(context, key as any, undefined)
 				}
 			}
 

src/core/context/instructions/user-instructions/cline-rules.ts

@@ -4,8 +4,9 @@ import { fileExistsAtPath, isDirectory, readDirectory } from "@utils/fs"
 import { formatResponse } from "@core/prompts/responses"
 import fs from "fs/promises"
 import { ClineRulesToggles } from "@shared/cline-rules"
+import { getGlobalState, getWorkspaceState, updateGlobalState, updateWorkspaceState } from "@core/storage/state"
+import * as vscode from "vscode"
 import { synchronizeRuleToggles, getRuleFilesTotalContent } from "@core/context/instructions/user-instructions/rule-helpers"
-import { Controller } from "@/core/controller"
 
 export const getGlobalClineRules = async (globalClineRulesFilePath: string, toggles: ClineRulesToggles) => {
 	if (await fileExistsAtPath(globalClineRulesFilePath)) {
@@ -67,25 +68,25 @@ export const getLocalClineRules = async (cwd: string, toggles: ClineRulesToggles
 }
 
 export async function refreshClineRulesToggles(
-	controller: Controller,
+	context: vscode.ExtensionContext,
 	workingDirectory: string,
 ): Promise<{
 	globalToggles: ClineRulesToggles
 	localToggles: ClineRulesToggles
 }> {
 	// Global toggles
-	const globalClineRulesToggles = controller.cacheService.getGlobalStateKey("globalClineRulesToggles")
+	const globalClineRulesToggles = ((await getGlobalState(context, "globalClineRulesToggles")) as ClineRulesToggles) || {}
 	const globalClineRulesFilePath = await ensureRulesDirectoryExists()
 	const updatedGlobalToggles = await synchronizeRuleToggles(globalClineRulesFilePath, globalClineRulesToggles)
-	controller.cacheService.setGlobalState("globalClineRulesToggles", updatedGlobalToggles)
+	await updateGlobalState(context, "globalClineRulesToggles", updatedGlobalToggles)
 
 	// Local toggles
-	const localClineRulesToggles = controller.cacheService.getWorkspaceStateKey("localClineRulesToggles")
+	const localClineRulesToggles = ((await getWorkspaceState(context, "localClineRulesToggles")) as ClineRulesToggles) || {}
 	const localClineRulesFilePath = path.resolve(workingDirectory, GlobalFileNames.clineRules)
 	const updatedLocalToggles = await synchronizeRuleToggles(localClineRulesFilePath, localClineRulesToggles, "", [
 		[".clinerules", "workflows"],
 	])
-	controller.cacheService.setWorkspaceState("localClineRulesToggles", updatedLocalToggles)
+	await updateWorkspaceState(context, "localClineRulesToggles", updatedLocalToggles)
 
 	return {
 		globalToggles: updatedGlobalToggles,

src/core/context/instructions/user-instructions/external-rules.ts

@@ -3,6 +3,7 @@ import fs from "fs/promises"
 import { GlobalFileNames } from "@core/storage/disk"
 import { fileExistsAtPath, isDirectory } from "@utils/fs"
 import { formatResponse } from "@core/prompts/responses"
+import { getWorkspaceState, updateWorkspaceState } from "@core/storage/state"
 import {
 	synchronizeRuleToggles,
 	combineRuleToggles,
@@ -10,26 +11,26 @@ import {
 	readDirectoryRecursive,
 } from "@core/context/instructions/user-instructions/rule-helpers"
 import { ClineRulesToggles } from "@shared/cline-rules"
-import { Controller } from "@/core/controller"
+import * as vscode from "vscode"
 
 /**
  * Refreshes the toggles for windsurf and cursor rules
  */
 export async function refreshExternalRulesToggles(
-	controller: Controller,
+	context: vscode.ExtensionContext,
 	workingDirectory: string,
 ): Promise<{
 	windsurfLocalToggles: ClineRulesToggles
 	cursorLocalToggles: ClineRulesToggles
 }> {
 	// local windsurf toggles
-	const localWindsurfRulesToggles = controller.cacheService.getWorkspaceStateKey("localWindsurfRulesToggles")
+	const localWindsurfRulesToggles = ((await getWorkspaceState(context, "localWindsurfRulesToggles")) as ClineRulesToggles) || {}
 	const localWindsurfRulesFilePath = path.resolve(workingDirectory, GlobalFileNames.windsurfRules)
 	const updatedLocalWindsurfToggles = await synchronizeRuleToggles(localWindsurfRulesFilePath, localWindsurfRulesToggles)
-	controller.cacheService.setWorkspaceState("localWindsurfRulesToggles", updatedLocalWindsurfToggles)
+	await updateWorkspaceState(context, "localWindsurfRulesToggles", updatedLocalWindsurfToggles)
 
 	// local cursor toggles
-	const localCursorRulesToggles = controller.cacheService.getWorkspaceStateKey("localCursorRulesToggles")
+	const localCursorRulesToggles = ((await getWorkspaceState(context, "localCursorRulesToggles")) as ClineRulesToggles) || {}
 
 	// cursor has two valid locations for rules files, so we need to check both and combine
 	// synchronizeRuleToggles will drop whichever rules files are not in each given path, but combining the results will result in no data loss
@@ -40,7 +41,7 @@ export async function refreshExternalRulesToggles(
 	const updatedLocalCursorToggles2 = await synchronizeRuleToggles(localCursorRulesFilePath, localCursorRulesToggles)
 
 	const updatedLocalCursorToggles = combineRuleToggles(updatedLocalCursorToggles1, updatedLocalCursorToggles2)
-	controller.cacheService.setWorkspaceState("localCursorRulesToggles", updatedLocalCursorToggles)
+	await updateWorkspaceState(context, "localCursorRulesToggles", updatedLocalCursorToggles)
 
 	return {
 		windsurfLocalToggles: updatedLocalWindsurfToggles,

src/core/context/instructions/user-instructions/rule-helpers.ts

@@ -1,9 +1,10 @@
 import { fileExistsAtPath, isDirectory, readDirectory } from "@utils/fs"
 import { ensureRulesDirectoryExists, ensureWorkflowsDirectoryExists, GlobalFileNames } from "@core/storage/disk"
+import { getGlobalState, getWorkspaceState, updateGlobalState, updateWorkspaceState } from "@core/storage/state"
 import * as path from "path"
 import fs from "fs/promises"
 import { ClineRulesToggles } from "@shared/cline-rules"
-import { Controller } from "@/core/controller"
+import * as vscode from "vscode"
 
 /**
  * Recursively traverses directory and finds all files, including checking for optional whitelisted file extension
@@ -223,7 +224,7 @@ export const createRuleFile = async (isGlobal: boolean, filename: string, cwd: s
  * Delete a rule file or workflow file
  */
 export async function deleteRuleFile(
-	controller: Controller,
+	context: vscode.ExtensionContext,
 	rulePath: string,
 	isGlobal: boolean,
 	type: string,
@@ -247,31 +248,31 @@ export async function deleteRuleFile(
 		// Update the appropriate toggles
 		if (isGlobal) {
 			if (type === "workflow") {
-				const toggles = controller.cacheService.getGlobalStateKey("globalWorkflowToggles")
+				const toggles = ((await getGlobalState(context, "globalWorkflowToggles")) as ClineRulesToggles) || {}
 				delete toggles[rulePath]
-				controller.cacheService.setGlobalState("globalWorkflowToggles", toggles)
+				await updateGlobalState(context, "globalWorkflowToggles", toggles)
 			} else {
-				const toggles = controller.cacheService.getGlobalStateKey("globalClineRulesToggles")
+				const toggles = ((await getGlobalState(context, "globalClineRulesToggles")) as ClineRulesToggles) || {}
 				delete toggles[rulePath]
-				controller.cacheService.setGlobalState("globalClineRulesToggles", toggles)
+				await updateGlobalState(context, "globalClineRulesToggles", toggles)
 			}
 		} else {
 			if (type === "workflow") {
-				const toggles = controller.cacheService.getWorkspaceStateKey("workflowToggles")
+				const toggles = ((await getWorkspaceState(context, "workflowToggles")) as ClineRulesToggles) || {}
 				delete toggles[rulePath]
-				controller.cacheService.setWorkspaceState("workflowToggles", toggles)
+				await updateWorkspaceState(context, "workflowToggles", toggles)
 			} else if (type === "cursor") {
-				const toggles = controller.cacheService.getWorkspaceStateKey("localCursorRulesToggles")
+				const toggles = ((await getWorkspaceState(context, "localCursorRulesToggles")) as ClineRulesToggles) || {}
 				delete toggles[rulePath]
-				controller.cacheService.setWorkspaceState("localCursorRulesToggles", toggles)
+				await updateWorkspaceState(context, "localCursorRulesToggles", toggles)
 			} else if (type === "windsurf") {
-				const toggles = controller.cacheService.getWorkspaceStateKey("localWindsurfRulesToggles")
+				const toggles = ((await getWorkspaceState(context, "localWindsurfRulesToggles")) as ClineRulesToggles) || {}
 				delete toggles[rulePath]
-				controller.cacheService.setWorkspaceState("localWindsurfRulesToggles", toggles)
+				await updateWorkspaceState(context, "localWindsurfRulesToggles", toggles)
 			} else {
-				const toggles = controller.cacheService.getWorkspaceStateKey("localClineRulesToggles")
+				const toggles = ((await getWorkspaceState(context, "localClineRulesToggles")) as ClineRulesToggles) || {}
 				delete toggles[rulePath]
-				controller.cacheService.setWorkspaceState("localClineRulesToggles", toggles)
+				await updateWorkspaceState(context, "localClineRulesToggles", toggles)
 			}
 		}
 

src/core/context/instructions/user-instructions/workflows.ts

@@ -1,29 +1,30 @@
 import path from "path"
 import { GlobalFileNames, ensureWorkflowsDirectoryExists } from "@core/storage/disk"
 import { ClineRulesToggles } from "@shared/cline-rules"
+import { getWorkspaceState, updateWorkspaceState, getGlobalState, updateGlobalState } from "@core/storage/state"
+import * as vscode from "vscode"
 import { synchronizeRuleToggles } from "@core/context/instructions/user-instructions/rule-helpers"
-import { Controller } from "@/core/controller"
 
 /**
  * Refresh the workflow toggles
  */
 export async function refreshWorkflowToggles(
-	controller: Controller,
+	context: vscode.ExtensionContext,
 	workingDirectory: string,
 ): Promise<{
 	globalWorkflowToggles: ClineRulesToggles
 	localWorkflowToggles: ClineRulesToggles
 }> {
 	// Global workflows
-	const globalWorkflowToggles = controller.cacheService.getGlobalStateKey("globalWorkflowToggles")
+	const globalWorkflowToggles = ((await getGlobalState(context, "globalWorkflowToggles")) as ClineRulesToggles) || {}
 	const globalClineWorkflowsFilePath = await ensureWorkflowsDirectoryExists()
 	const updatedGlobalWorkflowToggles = await synchronizeRuleToggles(globalClineWorkflowsFilePath, globalWorkflowToggles)
-	controller.cacheService.setGlobalState("globalWorkflowToggles", updatedGlobalWorkflowToggles)
+	await updateGlobalState(context, "globalWorkflowToggles", updatedGlobalWorkflowToggles)
 
-	const workflowRulesToggles = controller.cacheService.getWorkspaceStateKey("workflowToggles")
+	const workflowRulesToggles = ((await getWorkspaceState(context, "workflowToggles")) as ClineRulesToggles) || {}
 	const workflowsDirPath = path.resolve(workingDirectory, GlobalFileNames.workflows)
 	const updatedWorkflowToggles = await synchronizeRuleToggles(workflowsDirPath, workflowRulesToggles)
-	controller.cacheService.setWorkspaceState("workflowToggles", updatedWorkflowToggles)
+	await updateWorkspaceState(context, "workflowToggles", updatedWorkflowToggles)
 
 	return {
 		globalWorkflowToggles: updatedGlobalWorkflowToggles,

src/core/controller/account/accountLoginClicked.ts

@@ -2,6 +2,8 @@ import { Controller } from "../index"
 import { AuthService } from "@/services/auth/AuthService"
 import { EmptyRequest, String } from "@shared/proto/cline/common"
 
+const authService = AuthService.getInstance()
+
 /**
  * Handles the user clicking the login link in the UI.
  * Generates a secure nonce for state validation, stores it in secrets,
@@ -11,5 +13,5 @@ import { EmptyRequest, String } from "@shared/proto/cline/common"
  * @returns The login URL as a string.
  */
 export async function accountLoginClicked(_controller: Controller, _: EmptyRequest): Promise<String> {
-	return await AuthService.getInstance().createAuthRequest()
+	return await authService.createAuthRequest()
 }

src/core/controller/account/accountLogoutClicked.ts

@@ -3,6 +3,7 @@ import { Empty } from "@shared/proto/cline/common"
 import type { EmptyRequest } from "@shared/proto/cline/common"
 import type { Controller } from "../index"
 
+const authService = AuthService.getInstance()
 /**
  * Handles the account logout action
  * @param controller The controller instance
@@ -11,6 +12,6 @@ import type { Controller } from "../index"
  */
 export async function accountLogoutClicked(controller: Controller, _request: EmptyRequest): Promise<Empty> {
 	await controller.handleSignOut()
-	await AuthService.getInstance().handleDeauth()
+	await authService.handleDeauth()
 	return Empty.create({})
 }

src/core/controller/account/authStateChanged.ts

@@ -1,5 +1,6 @@
 import { AuthStateChangedRequest, AuthState } from "@shared/proto/cline/account"
 import type { Controller } from "../index"
+import { updateGlobalState } from "../../storage/state"
 
 /**
  * Handles authentication state changes from the Firebase context.
@@ -11,7 +12,7 @@ import type { Controller } from "../index"
 export async function authStateChanged(controller: Controller, request: AuthStateChangedRequest): Promise<AuthState> {
 	try {
 		// Store the user info directly in global state
-		controller.cacheService.setGlobalState("userInfo", request.user)
+		await updateGlobalState(controller.context, "userInfo", request.user)
 
 		// Return the same user info
 		return AuthState.create({ user: request.user })

src/core/controller/account/openrouterAuthClicked.ts

@@ -1,17 +0,0 @@
-import { HostProvider } from "@/hosts/host-provider"
-import { openExternal } from "@/utils/env"
-
-import { Controller } from ".."
-import { EmptyRequest, Empty } from "@shared/proto/cline/common"
-
-/**
- * Initiates OpenRouter auth
- */
-export async function openrouterAuthClicked(_: Controller, __: EmptyRequest): Promise<Empty> {
-	const callbackUri = await HostProvider.get().getCallbackUri()
-	const authUri = `https://openrouter.ai/auth?callback_url=${callbackUri}/openrouter`
-
-	await openExternal(authUri)
-
-	return {}
-}

src/core/controller/account/subscribeToAuthStatusUpdate.ts

@@ -1,13 +1,5 @@
-import { AuthState, EmptyRequest } from "@/shared/proto/index.cline"
-import { AuthService } from "@services/auth/AuthService"
-import { Controller } from ".."
-import { StreamingResponseHandler } from "../grpc-handler"
+import { AuthService } from "../../../services/auth/AuthService"
 
-export async function subscribeToAuthStatusUpdate(
-	controller: Controller,
-	request: EmptyRequest,
-	responseStream: StreamingResponseHandler<AuthState>,
-	requestId?: string,
-): Promise<void> {
-	return AuthService.getInstance().subscribeToAuthStatusUpdate(controller, request, responseStream, requestId)
-}
+const authService = AuthService.getInstance()
+export const subscribeToAuthStatusUpdate = authService.subscribeToAuthStatusUpdate.bind(authService)
+export const sendAuthStatusUpdateEvent = authService.sendAuthStatusUpdate.bind(authService)

src/core/controller/browser/discoverBrowser.ts

@@ -1,6 +1,7 @@
 import { BrowserConnection } from "@shared/proto/cline/browser"
 import { EmptyRequest } from "@shared/proto/cline/common"
 import { Controller } from "../index"
+import { getAllExtensionState } from "@core/storage/state"
 import { BrowserSession } from "@services/browser/BrowserSession"
 import { discoverChromeInstances } from "@services/browser/BrowserDiscovery"
 
@@ -19,7 +20,7 @@ export async function discoverBrowser(controller: Controller, request: EmptyRequ
 			// This way we don't override the user's preference
 
 			// Test the connection to get the endpoint
-			const browserSettings = controller.cacheService.getGlobalStateKey("browserSettings")
+			const { browserSettings } = await getAllExtensionState(controller.context)
 			const browserSession = new BrowserSession(controller.context, browserSettings)
 			const result = await browserSession.testConnection(discoveredHost)
 

src/core/controller/browser/getBrowserConnectionInfo.ts

@@ -1,6 +1,7 @@
 import { BrowserConnectionInfo } from "@shared/proto/cline/browser"
 import { EmptyRequest } from "@shared/proto/cline/common"
 import { Controller } from "../index"
+import { getAllExtensionState } from "@core/storage/state"
 
 /**
  * Get information about the current browser connection
@@ -11,7 +12,7 @@ import { Controller } from "../index"
 export async function getBrowserConnectionInfo(controller: Controller, _: EmptyRequest): Promise<BrowserConnectionInfo> {
 	try {
 		// Get browser settings from extension state
-		const browserSettings = controller.cacheService.getGlobalStateKey("browserSettings")
+		const { browserSettings } = await getAllExtensionState(controller.context)
 
 		// Check if there's an active browser session by using the controller's handleWebviewMessage approach
 		// This is similar to what's done in controller/index.ts for the "getBrowserConnectionInfo" message

src/core/controller/browser/getDetectedChromePath.ts

@@ -1,6 +1,7 @@
 import { ChromePath } from "@shared/proto/cline/browser"
 import { EmptyRequest } from "@shared/proto/cline/common"
 import { Controller } from "../index"
+import { getAllExtensionState } from "../../storage/state"
 import { BrowserSession } from "../../../services/browser/BrowserSession"
 
 /**
@@ -11,7 +12,7 @@ import { BrowserSession } from "../../../services/browser/BrowserSession"
  */
 export async function getDetectedChromePath(controller: Controller, _: EmptyRequest): Promise<ChromePath> {
 	try {
-		const browserSettings = controller.cacheService.getGlobalStateKey("browserSettings")
+		const { browserSettings } = await getAllExtensionState(controller.context)
 		const browserSession = new BrowserSession(controller.context, browserSettings)
 		const result = await browserSession.getDetectedChromePath()
 

src/core/controller/browser/relaunchChromeDebugMode.ts

@@ -16,9 +16,11 @@ export async function relaunchChromeDebugMode(controller: Controller, _: EmptyRe
 		// Relaunch Chrome in debug mode
 		await browserSession.relaunchChromeDebugMode(controller)
 
-		// The actual result will be sent via the ProtoBus in the BrowserSession.relaunchChromeDebugMode method
+		// The actual result will be sent via postMessageToWebview in the BrowserSession.relaunchChromeDebugMode method
 		// Here we just return a message as a placeholder
-		return { value: "Chrome relaunch initiated" }
+		return StringMessage.create({
+			value: "Chrome relaunch initiated",
+		})
 	} catch (error) {
 		throw new Error(`Error relaunching Chrome: ${error instanceof Error ? error.message : globalThis.String(error)}`)
 	}

src/core/controller/browser/testBrowserConnection.ts

@@ -1,6 +1,7 @@
 import { BrowserConnection } from "@shared/proto/cline/browser"
 import { StringRequest } from "@shared/proto/cline/common"
 import { Controller } from "../index"
+import { getAllExtensionState } from "@core/storage/state"
 import { BrowserSession } from "@services/browser/BrowserSession"
 import { discoverChromeInstances } from "@services/browser/BrowserDiscovery"
 
@@ -12,7 +13,7 @@ import { discoverChromeInstances } from "@services/browser/BrowserDiscovery"
  */
 export async function testBrowserConnection(controller: Controller, request: StringRequest): Promise<BrowserConnection> {
 	try {
-		const browserSettings = controller.cacheService.getGlobalStateKey("browserSettings")
+		const { browserSettings } = await getAllExtensionState(controller.context)
 		const browserSession = new BrowserSession(controller.context, browserSettings)
 		const text = request.value || ""
 

src/core/controller/browser/updateBrowserSettings.ts

@@ -1,6 +1,7 @@
 import { UpdateBrowserSettingsRequest } from "@shared/proto/cline/browser"
 import { Boolean } from "@shared/proto/cline/common"
 import { Controller } from "../index"
+import { updateGlobalState, getGlobalState } from "../../storage/state"
 import { BrowserSettings as SharedBrowserSettings, DEFAULT_BROWSER_SETTINGS } from "../../../shared/BrowserSettings"
 
 /**
@@ -12,7 +13,7 @@ import { BrowserSettings as SharedBrowserSettings, DEFAULT_BROWSER_SETTINGS } fr
 export async function updateBrowserSettings(controller: Controller, request: UpdateBrowserSettingsRequest): Promise<Boolean> {
 	try {
 		// Get current browser settings to preserve fields not in the request
-		const currentSettings = controller.cacheService.getGlobalStateKey("browserSettings")
+		const currentSettings = (await getGlobalState(controller.context, "browserSettings")) as SharedBrowserSettings | undefined
 		const mergedWithDefaults = { ...DEFAULT_BROWSER_SETTINGS, ...currentSettings }
 
 		// Convert from protobuf format to shared format, merging with existing settings
@@ -35,11 +36,10 @@ export async function updateBrowserSettings(controller: Controller, request: Upd
 				// Otherwise, fall back to mergedWithDefaults.
 				"chromeExecutablePath" in request ? request.chromeExecutablePath : mergedWithDefaults.chromeExecutablePath,
 			disableToolUse: request.disableToolUse === undefined ? mergedWithDefaults.disableToolUse : request.disableToolUse,
-			customArgs: "customArgs" in request ? request.customArgs : mergedWithDefaults.customArgs,
 		}
 
 		// Update global state with new settings
-		controller.cacheService.setGlobalState("browserSettings", newBrowserSettings)
+		await updateGlobalState(controller.context, "browserSettings", newBrowserSettings)
 
 		// Update task browser settings if task exists
 		if (controller.task) {

src/core/controller/file/__tests__/ifFileExistsRelativePath.test.ts

@@ -1,101 +0,0 @@
-import { describe, it, beforeEach, afterEach } from "mocha"
-import { expect } from "chai"
-import * as sinon from "sinon"
-import { ifFileExistsRelativePath } from "../ifFileExistsRelativePath"
-import { Controller } from "@core/controller"
-import { StringRequest, BooleanResponse } from "@shared/proto/cline/common"
-import * as pathUtils from "@utils/path"
-
-describe("ifFileExistsRelativePath", () => {
-	let sandbox: sinon.SinonSandbox
-	let mockController: Controller
-	let getWorkspacePathStub: sinon.SinonStub
-	let consoleErrorStub: sinon.SinonStub
-
-	beforeEach(() => {
-		sandbox = sinon.createSandbox()
-
-		// Create a mock controller
-		mockController = {} as any
-
-		// Stub getWorkspacePath utility
-		getWorkspacePathStub = sandbox.stub(pathUtils, "getWorkspacePath")
-
-		// Stub console.error to prevent test output pollution
-		consoleErrorStub = sandbox.stub(console, "error")
-	})
-
-	afterEach(() => {
-		sandbox.restore()
-	})
-
-	it("should return BooleanResponse with boolean value", async () => {
-		getWorkspacePathStub.resolves("/workspace")
-
-		const request = StringRequest.create({
-			value: "src/test.ts",
-		})
-
-		const result = await ifFileExistsRelativePath(mockController, request)
-
-		// The result should be a BooleanResponse object
-		expect(result).to.have.property("value")
-		expect(typeof result.value).to.equal("boolean")
-	})
-
-	it("should return false and log error when no workspace path is available", async () => {
-		const noWorkspaceScenarios = [null, undefined]
-
-		for (const workspaceValue of noWorkspaceScenarios) {
-			getWorkspacePathStub.resolves(workspaceValue)
-			consoleErrorStub.resetHistory()
-
-			const request = StringRequest.create({
-				value: "src/test.ts",
-			})
-
-			const result = await ifFileExistsRelativePath(mockController, request)
-
-			expect(result).to.deep.equal(BooleanResponse.create({ value: false }))
-			expect(consoleErrorStub.called).to.be.true
-		}
-	})
-
-	it("should return false when path is invalid", async () => {
-		getWorkspacePathStub.resolves("/workspace")
-
-		const invalidPaths = ["", undefined]
-
-		for (const invalidPath of invalidPaths) {
-			const request = StringRequest.create({
-				value: invalidPath,
-			})
-
-			const result = await ifFileExistsRelativePath(mockController, request)
-
-			expect(result).to.deep.equal(BooleanResponse.create({ value: false }))
-		}
-	})
-
-	it("should handle valid relative paths correctly", async () => {
-		getWorkspacePathStub.resolves("/workspace")
-
-		// Test with valid workspace-relative paths only
-		const validPaths = ["src/file.ts", "./src/file.ts", "package.json", ".gitignore", "src/components/ui/Button/Button.tsx"]
-
-		for (const testPath of validPaths) {
-			const request = StringRequest.create({
-				value: testPath,
-			})
-
-			const result = await ifFileExistsRelativePath(mockController, request)
-
-			// Each should return a BooleanResponse
-			expect(result).to.have.property("value")
-			expect(typeof result.value).to.equal("boolean")
-		}
-
-		// Verify that getWorkspacePath was called for each path
-		expect(getWorkspacePathStub.callCount).to.equal(validPaths.length)
-	})
-})

src/core/controller/file/__tests__/openFileRelativePath.test.ts

@@ -1,117 +0,0 @@
-import { describe, it, beforeEach, afterEach } from "mocha"
-import { expect } from "chai"
-import * as sinon from "sinon"
-import { openFileRelativePath } from "../openFileRelativePath"
-import { Controller } from "@core/controller"
-import { StringRequest, Empty } from "@shared/proto/cline/common"
-import * as openFileIntegration from "@integrations/misc/open-file"
-import * as pathUtils from "@utils/path"
-import * as path from "path"
-
-describe("openFileRelativePath", () => {
-	let sandbox: sinon.SinonSandbox
-	let mockController: Controller
-	let openFileIntegrationStub: sinon.SinonStub
-	let getWorkspacePathStub: sinon.SinonStub
-	let consoleErrorStub: sinon.SinonStub
-
-	beforeEach(() => {
-		sandbox = sinon.createSandbox()
-
-		// Create a mock controller
-		mockController = {} as any
-
-		// Stub the openFileIntegration function
-		openFileIntegrationStub = sandbox.stub(openFileIntegration, "openFile")
-
-		// Stub getWorkspacePath utility
-		getWorkspacePathStub = sandbox.stub(pathUtils, "getWorkspacePath")
-
-		// Stub console.error to prevent test output pollution
-		consoleErrorStub = sandbox.stub(console, "error")
-	})
-
-	afterEach(() => {
-		sandbox.restore()
-	})
-
-	it("should return Empty response on successful execution", async () => {
-		getWorkspacePathStub.resolves("/workspace")
-
-		const request = StringRequest.create({
-			value: "src/test.ts",
-		})
-
-		const result = await openFileRelativePath(mockController, request)
-
-		expect(result).to.deep.equal(Empty.create())
-	})
-
-	it("should call openFileIntegration with absolute path when relative path is provided", async () => {
-		const workspacePath = "/workspace"
-		const relativePath = "src/components/Test.tsx"
-		const expectedAbsolutePath = path.resolve(workspacePath, relativePath)
-
-		getWorkspacePathStub.resolves(workspacePath)
-
-		const request = StringRequest.create({
-			value: relativePath,
-		})
-
-		await openFileRelativePath(mockController, request)
-
-		expect(openFileIntegrationStub.calledOnceWith(expectedAbsolutePath)).to.be.true
-	})
-
-	it("should not call openFileIntegration when path is invalid", async () => {
-		getWorkspacePathStub.resolves("/workspace")
-
-		const invalidPaths = ["", undefined]
-
-		for (const invalidPath of invalidPaths) {
-			const request = StringRequest.create({
-				value: invalidPath,
-			})
-
-			await openFileRelativePath(mockController, request)
-
-			expect(openFileIntegrationStub.called).to.be.false
-			openFileIntegrationStub.resetHistory()
-		}
-	})
-
-	it("should return Empty and log error when no workspace path is available", async () => {
-		const noWorkspaceScenarios = [null, undefined]
-
-		for (const workspaceValue of noWorkspaceScenarios) {
-			getWorkspacePathStub.resolves(workspaceValue)
-			consoleErrorStub.resetHistory()
-
-			const request = StringRequest.create({
-				value: "src/test.ts",
-			})
-
-			const result = await openFileRelativePath(mockController, request)
-
-			expect(result).to.deep.equal(Empty.create())
-			expect(consoleErrorStub.called).to.be.true
-			expect(openFileIntegrationStub.called).to.be.false
-		}
-	})
-
-	it("should handle nested directory paths", async () => {
-		const workspacePath = "/workspace"
-		const relativePath = "src/components/ui/Button/Button.tsx"
-		const expectedAbsolutePath = path.resolve(workspacePath, relativePath)
-
-		getWorkspacePathStub.resolves(workspacePath)
-
-		const request = StringRequest.create({
-			value: relativePath,
-		})
-
-		await openFileRelativePath(mockController, request)
-
-		expect(openFileIntegrationStub.calledOnceWith(expectedAbsolutePath)).to.be.true
-	})
-})

src/core/controller/file/createRuleFile.ts

@@ -51,9 +51,9 @@ export async function createRuleFile(controller: Controller, request: RuleFileRe
 		await openFile(controller, { value: filePath })
 	} else {
 		if (request.type === "workflow") {
-			await refreshWorkflowToggles(controller, cwd)
+			await refreshWorkflowToggles(controller.context, cwd)
 		} else {
-			await refreshClineRulesToggles(controller, cwd)
+			await refreshClineRulesToggles(controller.context, cwd)
 		}
 		await controller.postStateToWebview()
 

src/core/controller/file/deleteRuleFile.ts

@@ -28,7 +28,7 @@ export async function deleteRuleFile(controller: Controller, request: RuleFileRe
 		throw new Error("Missing or invalid parameters")
 	}
 
-	const result = await deleteRuleFileImpl(controller, request.rulePath, request.isGlobal, request.type)
+	const result = await deleteRuleFileImpl(controller.context, request.rulePath, request.isGlobal, request.type)
 
 	if (!result.success) {
 		throw new Error(result.message || "Failed to delete rule file")

src/core/controller/file/ifFileExistsRelativePath.ts

@@ -1,35 +0,0 @@
-import * as path from "path"
-import * as fs from "fs"
-import { Controller } from ".."
-import { StringRequest, BooleanResponse } from "@shared/proto/cline/common"
-import { getWorkspacePath } from "@utils/path"
-
-/**
- * Check if a file exists in the project using a relative path
- * @param controller The controller instance
- * @param request The request containing the relative file path to check
- * @returns BooleanResponse indicating whether the file exists
- */
-export async function ifFileExistsRelativePath(_controller: Controller, request: StringRequest): Promise<BooleanResponse> {
-	const workspacePath = await getWorkspacePath()
-
-	if (!workspacePath) {
-		// If no workspace is open, return false
-		console.error("Error in ifFileExistsRelativePath: No workspace path available") // TODO
-		return BooleanResponse.create({ value: false })
-	}
-
-	if (!request.value) {
-		// If no path provided, return false
-		return BooleanResponse.create({ value: false })
-	}
-
-	// Resolve the relative path to absolute path
-	const absolutePath = path.resolve(workspacePath, request.value)
-	// Check if the file exists
-	try {
-		return BooleanResponse.create({ value: fs.statSync(absolutePath).isFile() })
-	} catch {
-		return BooleanResponse.create({ value: false })
-	}
-}

src/core/controller/file/openFileRelativePath.ts

@@ -1,30 +0,0 @@
-import * as path from "path"
-import { Controller } from ".."
-import { Empty, StringRequest } from "@shared/proto/cline/common"
-import { openFile as openFileIntegration } from "@integrations/misc/open-file"
-import { getWorkspacePath } from "@utils/path"
-
-/**
- * Opens a file in the editor by a relative path
- * @param controller The controller instance
- * @param request The request message containing the relative file path in the 'value' field
- * @returns Empty response
- */
-export async function openFileRelativePath(_controller: Controller, request: StringRequest): Promise<Empty> {
-	const workspacePath = await getWorkspacePath()
-
-	if (!workspacePath) {
-		console.error("Error in openFileRelativePath: No workspace path available")
-		return Empty.create()
-	}
-
-	if (request.value) {
-		// Resolve the relative path to absolute path
-		const absolutePath = path.resolve(workspacePath, request.value)
-
-		// Open the file using the existing integration
-		openFileIntegration(absolutePath)
-	}
-
-	return Empty.create()
-}

src/core/controller/file/openFocusChainFile.ts

@@ -1,40 +0,0 @@
-import { Controller } from ".."
-import { openFile as openFileIntegration } from "@integrations/misc/open-file"
-import { Empty, StringRequest } from "../../../shared/proto/cline/common"
-import { ensureFocusChainFile, extractFocusChainListFromText } from "../../task/focus-chain/file-utils"
-import { telemetryService } from "../../../services/posthog/PostHogClientProvider"
-
-/**
- * Opens or creates a focus chain checklist markdown file for editing
- * The file is stored at <globalStorage>/tasks/<taskId>/focus_chain_taskid_<taskId>.md
- */
-export async function openFocusChainFile(controller: Controller, request: StringRequest): Promise<Empty> {
-	if (!request.value) {
-		throw new Error("Task ID is required")
-	}
-
-	const taskId = request.value
-
-	// Get the current focus chain list from the task's most recent task_progress message
-	let initialFocusChainContent: string | undefined
-	const currentTask = controller.task
-	if (currentTask) {
-		// Get the task's message history and find the most recent task_progress message
-		// TODO - can we decouple this from ClineMessages?
-		const clineMessages = currentTask.messageStateHandler.getClineMessages()
-		const lastProgressMessage = clineMessages
-			.slice()
-			.reverse()
-			.find((m) => m.say === "task_progress")
-
-		if (lastProgressMessage && lastProgressMessage.text) {
-			initialFocusChainContent = extractFocusChainListFromText(lastProgressMessage.text) || undefined
-		}
-	}
-
-	const focusChainFilePath = await ensureFocusChainFile(controller.context, taskId, initialFocusChainContent)
-	telemetryService.captureFocusChainListOpened(taskId)
-	await openFileIntegration(focusChainFilePath)
-
-	return Empty.create()
-}

src/core/controller/file/refreshRules.ts

@@ -15,9 +15,9 @@ import { getCwd, getDesktopDir } from "@/utils/path"
 export async function refreshRules(controller: Controller, _request: EmptyRequest): Promise<RefreshedRules> {
 	try {
 		const cwd = await getCwd(getDesktopDir())
-		const { globalToggles, localToggles } = await refreshClineRulesToggles(controller, cwd)
-		const { cursorLocalToggles, windsurfLocalToggles } = await refreshExternalRulesToggles(controller, cwd)
-		const { localWorkflowToggles, globalWorkflowToggles } = await refreshWorkflowToggles(controller, cwd)
+		const { globalToggles, localToggles } = await refreshClineRulesToggles(controller.context, cwd)
+		const { cursorLocalToggles, windsurfLocalToggles } = await refreshExternalRulesToggles(controller.context, cwd)
+		const { localWorkflowToggles, globalWorkflowToggles } = await refreshWorkflowToggles(controller.context, cwd)
 
 		return RefreshedRules.create({
 			globalClineRulesToggles: { toggles: globalToggles },

src/core/controller/file/searchFiles.ts

@@ -1,5 +1,5 @@
 import { Controller } from ".."
-import { FileSearchRequest, FileSearchResults, FileSearchType } from "@shared/proto/cline/file"
+import { FileSearchRequest, FileSearchResults } from "@shared/proto/cline/file"
 import { searchWorkspaceFiles } from "@services/search/file-search"
 import { getWorkspacePath } from "@utils/path"
 import { convertSearchResultsToProtoFileInfos } from "@shared/proto-conversions/file/search-result-conversion"
@@ -23,20 +23,11 @@ export async function searchFiles(_controller: Controller, request: FileSearchRe
 	}
 
 	try {
-		// Map enum to string for the search service
-		let selectedTypeString: "file" | "folder" | undefined = undefined
-		if (request.selectedType === FileSearchType.FILE) {
-			selectedTypeString = "file"
-		} else if (request.selectedType === FileSearchType.FOLDER) {
-			selectedTypeString = "folder"
-		}
-
 		// Call file search service with query from request
 		const searchResults = await searchWorkspaceFiles(
 			request.query || "",
 			workspacePath,
 			request.limit || 20, // Use default limit of 20 if not specified
-			selectedTypeString,
 		)
 
 		// Convert search results to proto FileInfo objects using the conversion function

src/core/controller/file/subscribeToWorkspaceUpdates.ts

@@ -0,0 +1,58 @@
+import { Controller } from "../index"
+import { EmptyRequest, StringArray } from "@shared/proto/cline/common"
+import { StreamingResponseHandler, getRequestRegistry } from "../grpc-handler"
+
+// Keep track of active subscriptions
+const activeWorkspaceUpdateSubscriptions = new Set<StreamingResponseHandler<StringArray>>()
+
+/**
+ * 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<StringArray>,
+	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/toggleClineRule.ts

@@ -1,6 +1,7 @@
 import { ToggleClineRules } from "@shared/proto/cline/file"
 import type { ToggleClineRuleRequest } from "@shared/proto/cline/file"
 import type { Controller } from "../index"
+import { getGlobalState, getWorkspaceState, updateGlobalState, updateWorkspaceState } from "../../../core/storage/state"
 import { ClineRulesToggles as AppClineRulesToggles } from "@shared/cline-rules"
 
 /**
@@ -23,18 +24,18 @@ export async function toggleClineRule(controller: Controller, request: ToggleCli
 
 	// This is the same core logic as in the original handler
 	if (isGlobal) {
-		const toggles = controller.cacheService.getGlobalStateKey("globalClineRulesToggles")
+		const toggles = ((await getGlobalState(controller.context, "globalClineRulesToggles")) as AppClineRulesToggles) || {}
 		toggles[rulePath] = enabled
-		controller.cacheService.setGlobalState("globalClineRulesToggles", toggles)
+		await updateGlobalState(controller.context, "globalClineRulesToggles", toggles)
 	} else {
-		const toggles = controller.cacheService.getWorkspaceStateKey("localClineRulesToggles")
+		const toggles = ((await getWorkspaceState(controller.context, "localClineRulesToggles")) as AppClineRulesToggles) || {}
 		toggles[rulePath] = enabled
-		controller.cacheService.setWorkspaceState("localClineRulesToggles", toggles)
+		await updateWorkspaceState(controller.context, "localClineRulesToggles", toggles)
 	}
 
 	// Get the current state to return in the response
-	const globalToggles = controller.cacheService.getGlobalStateKey("globalClineRulesToggles")
-	const localToggles = controller.cacheService.getWorkspaceStateKey("localClineRulesToggles")
+	const globalToggles = ((await getGlobalState(controller.context, "globalClineRulesToggles")) as AppClineRulesToggles) || {}
+	const localToggles = ((await getWorkspaceState(controller.context, "localClineRulesToggles")) as AppClineRulesToggles) || {}
 
 	return ToggleClineRules.create({
 		globalClineRulesToggles: { toggles: globalToggles },

src/core/controller/file/toggleCursorRule.ts

@@ -1,6 +1,7 @@
 import type { ToggleCursorRuleRequest } from "@shared/proto/cline/file"
 import { ClineRulesToggles } from "@shared/proto/cline/file"
 import type { Controller } from "../index"
+import { getWorkspaceState, updateWorkspaceState } from "../../../core/storage/state"
 import { ClineRulesToggles as AppClineRulesToggles } from "@shared/cline-rules"
 
 /**
@@ -21,12 +22,12 @@ export async function toggleCursorRule(controller: Controller, request: ToggleCu
 	}
 
 	// Update the toggles in workspace state
-	const toggles = controller.cacheService.getWorkspaceStateKey("localCursorRulesToggles")
+	const toggles = ((await getWorkspaceState(controller.context, "localCursorRulesToggles")) as AppClineRulesToggles) || {}
 	toggles[rulePath] = enabled
-	controller.cacheService.setWorkspaceState("localCursorRulesToggles", toggles)
+	await updateWorkspaceState(controller.context, "localCursorRulesToggles", toggles)
 
 	// Get the current state to return in the response
-	const cursorToggles = controller.cacheService.getWorkspaceStateKey("localCursorRulesToggles")
+	const cursorToggles = ((await getWorkspaceState(controller.context, "localCursorRulesToggles")) as AppClineRulesToggles) || {}
 
 	return ClineRulesToggles.create({
 		toggles: cursorToggles,

src/core/controller/file/toggleWindsurfRule.ts

@@ -1,6 +1,7 @@
 import type { ToggleWindsurfRuleRequest } from "@shared/proto/cline/file"
 import { ClineRulesToggles } from "@shared/proto/cline/file"
 import type { Controller } from "../index"
+import { getWorkspaceState, updateWorkspaceState } from "../../../core/storage/state"
 import { ClineRulesToggles as AppClineRulesToggles } from "@shared/cline-rules"
 
 /**
@@ -21,9 +22,9 @@ export async function toggleWindsurfRule(controller: Controller, request: Toggle
 	}
 
 	// Update the toggles
-	const toggles = controller.cacheService.getWorkspaceStateKey("localWindsurfRulesToggles")
+	const toggles = ((await getWorkspaceState(controller.context, "localWindsurfRulesToggles")) as AppClineRulesToggles) || {}
 	toggles[rulePath] = enabled
-	controller.cacheService.setWorkspaceState("localWindsurfRulesToggles", toggles)
+	await updateWorkspaceState(controller.context, "localWindsurfRulesToggles", toggles)
 
 	// Return the toggles directly
 	return ClineRulesToggles.create({ toggles: toggles })

src/core/controller/file/toggleWorkflow.ts

@@ -1,5 +1,7 @@
 import { Controller } from ".."
+import { Metadata } from "@shared/proto/cline/common"
 import { ToggleWorkflowRequest, ClineRulesToggles } from "@shared/proto/cline/file"
+import { getWorkspaceState, updateWorkspaceState, getGlobalState, updateGlobalState } from "../../../core/storage/state"
 import { ClineRulesToggles as AppClineRulesToggles } from "../../../shared/cline-rules"
 
 /**
@@ -22,18 +24,18 @@ export async function toggleWorkflow(controller: Controller, request: ToggleWork
 	// Update the toggles based on isGlobal flag
 	if (isGlobal) {
 		// Global workflows
-		const toggles = controller.cacheService.getGlobalStateKey("globalWorkflowToggles")
+		const toggles = ((await getGlobalState(controller.context, "globalWorkflowToggles")) as AppClineRulesToggles) || {}
 		toggles[workflowPath] = enabled
-		controller.cacheService.setGlobalState("globalWorkflowToggles", toggles)
+		await updateGlobalState(controller.context, "globalWorkflowToggles", toggles)
 		await controller.postStateToWebview()
 
 		// Return the global toggles
 		return ClineRulesToggles.create({ toggles: toggles })
 	} else {
 		// Workspace workflows
-		const toggles = controller.cacheService.getWorkspaceStateKey("workflowToggles")
+		const toggles = ((await getWorkspaceState(controller.context, "workflowToggles")) as AppClineRulesToggles) || {}
 		toggles[workflowPath] = enabled
-		controller.cacheService.setWorkspaceState("workflowToggles", toggles)
+		await updateWorkspaceState(controller.context, "workflowToggles", toggles)
 		await controller.postStateToWebview()
 
 		// Return the workspace toggles

src/core/controller/grpc-handler.test.ts

@@ -1,412 +0,0 @@
-import { describe, it, beforeEach, afterEach } from "mocha"
-import { expect } from "chai"
-import * as sinon from "sinon"
-import { handleGrpcRequest, handleGrpcRequestCancel, getRequestRegistry } from "./grpc-handler"
-import { Controller } from "@core/controller"
-import { GrpcRequest, GrpcCancel } from "@shared/WebviewMessage"
-import { serviceHandlers } from "@generated/hosts/vscode/protobus-services"
-
-describe("grpc-handler", () => {
-	let sandbox: sinon.SinonSandbox
-	let mockController: Controller
-	let mockPostMessageToWebview: sinon.SinonStub
-
-	let mockUnaryHandler: sinon.SinonStub
-	let mockUnaryFailingHandler: sinon.SinonStub
-	let mockStreamingHandler: sinon.SinonStub
-	let mockStreamingFailingHandler: sinon.SinonStub
-
-	const serviceName = "cline.TestService"
-	const mockResponse = { result: "result-1234" }
-
-	beforeEach(() => {
-		sandbox = sinon.createSandbox()
-
-		// Create a mock controller
-		mockController = {} as any
-		mockPostMessageToWebview = sandbox.stub().resolves()
-
-		// Create mock service handlers
-		mockUnaryHandler = sandbox.stub().resolves(mockResponse)
-		mockStreamingHandler = sandbox.stub().resolves()
-		mockUnaryFailingHandler = sandbox.stub().rejects(new Error("Test error unary"))
-		mockStreamingFailingHandler = sandbox.stub().rejects(new Error("Stream error"))
-		serviceHandlers[serviceName] = {
-			testUnary: mockUnaryHandler,
-			testUnaryFailing: mockUnaryFailingHandler,
-			testStreaming: mockStreamingHandler,
-			testStreamingFailing: mockStreamingFailingHandler,
-		}
-	})
-
-	afterEach(() => {
-		sandbox.restore()
-	})
-
-	describe("handleGrpcRequest", () => {
-		describe("Unary requests", () => {
-			it("should handle successful unary requests", async () => {
-				const request: GrpcRequest = {
-					service: serviceName,
-					method: "testUnary",
-					message: { input: "test" },
-					request_id: "test-123",
-					is_streaming: false,
-				}
-
-				await handleGrpcRequest(mockController, mockPostMessageToWebview, request)
-
-				// Verify the handler was called
-				expect(mockUnaryHandler.calledOnce).to.be.true
-				expect(mockUnaryHandler.firstCall.args[0]).to.equal(mockController)
-				expect(mockUnaryHandler.firstCall.args[1]).to.deep.equal({ input: "test" })
-
-				// Verify the response was sent
-				expect(mockPostMessageToWebview.calledOnce).to.be.true
-				const sentMessage = mockPostMessageToWebview.firstCall.args[0]
-				expect(sentMessage).to.deep.equal({
-					type: "grpc_response",
-					grpc_response: {
-						message: mockResponse,
-						request_id: "test-123",
-					},
-				})
-			})
-
-			it("should handle errors in unary requests", async () => {
-				const request: GrpcRequest = {
-					service: serviceName,
-					method: "testUnaryFailing",
-					message: { input: "test" },
-					request_id: "test-456",
-					is_streaming: false,
-				}
-
-				await handleGrpcRequest(mockController, mockPostMessageToWebview, request)
-
-				// Verify the error response was sent
-				expect(mockPostMessageToWebview.calledOnce).to.be.true
-				const sentMessage = mockPostMessageToWebview.firstCall.args[0]
-				expect(sentMessage).to.deep.equal({
-					type: "grpc_response",
-					grpc_response: {
-						error: "Test error unary",
-						request_id: "test-456",
-						is_streaming: false,
-					},
-				})
-			})
-
-			it("should handle unknown service errors", async () => {
-				const request: GrpcRequest = {
-					service: "UnknownService",
-					method: "someMethod",
-					message: {},
-					request_id: "test-789",
-					is_streaming: false,
-				}
-
-				await handleGrpcRequest(mockController, mockPostMessageToWebview, request)
-
-				// Verify the error response was sent
-				expect(mockPostMessageToWebview.calledOnce).to.be.true
-				const sentMessage = mockPostMessageToWebview.firstCall.args[0]
-				expect(sentMessage.type).to.equal("grpc_response")
-				expect(sentMessage.grpc_response?.error).to.include("Unknown service: UnknownService")
-				expect(sentMessage.grpc_response?.request_id).to.equal("test-789")
-			})
-
-			it("should handle unknown method errors", async () => {
-				const request: GrpcRequest = {
-					service: serviceName,
-					method: "unknownMethod",
-					message: {},
-					request_id: "test-999",
-					is_streaming: false,
-				}
-
-				await handleGrpcRequest(mockController, mockPostMessageToWebview, request)
-
-				// Verify the error response was sent
-				expect(mockPostMessageToWebview.calledOnce).to.be.true
-				const sentMessage = mockPostMessageToWebview.firstCall.args[0]
-				expect(sentMessage.type).to.equal("grpc_response")
-				expect(sentMessage.grpc_response?.error).to.include("Unknown rpc: cline.TestService.unknownMethod")
-				expect(sentMessage.grpc_response?.request_id).to.equal("test-999")
-			})
-		})
-		describe("Streaming requests", () => {
-			it("should handle successful streaming requests", async () => {
-				// Set up a streaming handler that sends multiple responses
-				const request: GrpcRequest = {
-					service: serviceName,
-					method: "testStreaming",
-					message: { input: "stream" },
-					request_id: "stream-123",
-					is_streaming: true,
-				}
-
-				// Reset the mock and set up the handler using callsFake
-				mockStreamingHandler.reset()
-				mockStreamingHandler.callsFake(async (controller: any, message: any, responseStream: any, requestId: string) => {
-					// Simulate streaming multiple messages
-					await responseStream({ value: 1 }, false, 0)
-					await responseStream({ value: 2 }, false, 1)
-					await responseStream({ value: 3 }, true, 2) // Last message
-				})
-
-				await handleGrpcRequest(mockController, mockPostMessageToWebview, request)
-
-				// Verify the handler was called
-				expect(mockStreamingHandler.calledOnce).to.be.true
-				expect(mockStreamingHandler.firstCall.args[0]).to.equal(mockController)
-				expect(mockStreamingHandler.firstCall.args[1]).to.deep.equal({ input: "stream" })
-				expect(mockStreamingHandler.firstCall.args[3]).to.equal("stream-123")
-
-				// Verify all streaming responses were sent
-				expect(mockPostMessageToWebview.callCount).to.equal(3)
-
-				// Check all responses
-				expect(mockPostMessageToWebview.firstCall.args[0]).to.deep.equal({
-					type: "grpc_response",
-					grpc_response: {
-						message: { value: 1 },
-						request_id: "stream-123",
-						is_streaming: true,
-						sequence_number: 0,
-					},
-				})
-				expect(mockPostMessageToWebview.secondCall.args[0]).to.deep.equal({
-					type: "grpc_response",
-					grpc_response: {
-						message: { value: 2 },
-						request_id: "stream-123",
-						is_streaming: true,
-						sequence_number: 1,
-					},
-				})
-				expect(mockPostMessageToWebview.thirdCall.args[0]).to.deep.equal({
-					type: "grpc_response",
-					grpc_response: {
-						message: { value: 3 },
-						request_id: "stream-123",
-						is_streaming: false, // Last message has is_streaming: false
-						sequence_number: 2,
-					},
-				})
-			})
-
-			it("should handle errors in streaming requests", async () => {
-				const request: GrpcRequest = {
-					service: serviceName,
-					method: "testStreamingFailing",
-					message: { input: "stream" },
-					request_id: "stream-456",
-					is_streaming: true,
-				}
-
-				await handleGrpcRequest(mockController, mockPostMessageToWebview, request)
-
-				// Verify the error response was sent
-				expect(mockPostMessageToWebview.calledOnce).to.be.true
-				const sentMessage = mockPostMessageToWebview.firstCall.args[0]
-				expect(sentMessage).to.deep.equal({
-					type: "grpc_response",
-					grpc_response: {
-						error: "Stream error",
-						request_id: "stream-456",
-						is_streaming: false,
-					},
-				})
-			})
-
-			it("should handle streaming with message, error, then another message", async () => {
-				// This test simulates a scenario where:
-				// 1. First message is sent successfully
-				// 2. An error occurs
-				// 3. Another message is attempted (which should not be sent after error)
-
-				const request: GrpcRequest = {
-					service: serviceName,
-					method: "testStreaming",
-					message: { input: "stream-with-error" },
-					request_id: "stream-error-mid",
-					is_streaming: true,
-				}
-
-				// Reset the mock and set up the handler to throw an error after being called
-				mockStreamingHandler.reset()
-				mockStreamingHandler.callsFake(async (controller: any, message: any, responseStream: any, requestId: string) => {
-					// Send first message successfully
-					await responseStream({ value: "first" }, false, 0)
-					// Throw an error
-					throw new Error("Mid-stream error")
-				})
-
-				await handleGrpcRequest(mockController, mockPostMessageToWebview, request)
-
-				// Verify the handler was called
-				expect(mockStreamingHandler.calledOnce).to.be.true
-
-				// Verify that we got the first message and then the error
-				expect(mockPostMessageToWebview.callCount).to.equal(2)
-
-				// Check first message was sent successfully
-				expect(mockPostMessageToWebview.firstCall.args[0]).to.deep.equal({
-					type: "grpc_response",
-					grpc_response: {
-						message: { value: "first" },
-						request_id: "stream-error-mid",
-						is_streaming: true,
-						sequence_number: 0,
-					},
-				})
-
-				// Check error response was sent
-				expect(mockPostMessageToWebview.secondCall.args[0]).to.deep.equal({
-					type: "grpc_response",
-					grpc_response: {
-						error: "Mid-stream error",
-						request_id: "stream-error-mid",
-						is_streaming: false,
-					},
-				})
-
-				// Try to send another message after the error (simulating what might happen
-				// if the handler tried to continue after an error)
-				const responseStream = mockStreamingHandler.firstCall.args[2]
-
-				// This should still work as the responseStream function is still valid
-				await responseStream({ value: "after-error" }, false, 1)
-
-				// Verify we now have 3 total calls (first message, error, after-error message)
-				expect(mockPostMessageToWebview.callCount).to.equal(3)
-
-				// Verify the message after error was still sent
-				// (In a real scenario, the handler would have stopped due to the error,
-				// but this tests that the responseStream function itself still works)
-				expect(mockPostMessageToWebview.thirdCall.args[0]).to.deep.equal({
-					type: "grpc_response",
-					grpc_response: {
-						message: { value: "after-error" },
-						request_id: "stream-error-mid",
-						is_streaming: true,
-						sequence_number: 1,
-					},
-				})
-			})
-		})
-
-		describe("handleGrpcRequestCancel", () => {
-			it("should cancel an active request", async () => {
-				// Register a request in the registry
-				const registry = getRequestRegistry()
-				const cleanupStub = sandbox.stub()
-				registry.registerRequest("cancel-123", cleanupStub)
-
-				const cancelRequest: GrpcCancel = {
-					request_id: "cancel-123",
-				}
-
-				await handleGrpcRequestCancel(mockPostMessageToWebview, cancelRequest)
-
-				// Verify the cleanup was called
-				expect(cleanupStub.calledOnce).to.be.true
-
-				// Verify the cancellation confirmation was sent
-				expect(mockPostMessageToWebview.calledOnce).to.be.true
-				const sentMessage = mockPostMessageToWebview.firstCall.args[0]
-				expect(sentMessage).to.deep.equal({
-					type: "grpc_response",
-					grpc_response: {
-						message: { cancelled: true },
-						request_id: "cancel-123",
-						is_streaming: false,
-					},
-				})
-
-				// Verify the request was removed from the registry
-				expect(registry.hasRequest("cancel-123")).to.be.false
-			})
-
-			it("should handle cancellation of non-existent request", async () => {
-				const cancelRequest: GrpcCancel = {
-					request_id: "non-existent",
-				}
-
-				await handleGrpcRequestCancel(mockPostMessageToWebview, cancelRequest)
-
-				// Verify no message was sent (request not found)
-				expect(mockPostMessageToWebview.called).to.be.false
-			})
-
-			it("should handle cleanup errors gracefully", async () => {
-				// Register a request with a failing cleanup
-				const registry = getRequestRegistry()
-				const cleanupStub = sandbox.stub().throws(new Error("Cleanup failed"))
-				registry.registerRequest("cancel-error", cleanupStub)
-
-				const cancelRequest: GrpcCancel = {
-					request_id: "cancel-error",
-				}
-
-				// Should not throw
-				await handleGrpcRequestCancel(mockPostMessageToWebview, cancelRequest)
-
-				// Verify the cleanup was attempted
-				expect(cleanupStub.calledOnce).to.be.true
-
-				// Verify the cancellation confirmation was still sent
-				expect(mockPostMessageToWebview.calledOnce).to.be.true
-
-				// Verify the request was removed despite the error
-				expect(registry.hasRequest("cancel-error")).to.be.false
-			})
-		})
-
-		describe("Concurrent requests", () => {
-			it("should handle concurrent requests", async () => {
-				// Set up handlers
-				mockUnaryHandler.resolves({ result: "unary" })
-				mockStreamingHandler.callsFake(async (_controller: any, _message: any, responseStream: any) => {
-					await responseStream({ value: "stream1" }, false, 0)
-					await responseStream({ value: "stream2" }, true, 1)
-				})
-
-				// Send multiple requests concurrently
-				const requests = [
-					handleGrpcRequest(mockController, mockPostMessageToWebview, {
-						service: serviceName,
-						method: "testUnary",
-						message: { id: 1 },
-						request_id: "concurrent-1",
-						is_streaming: false,
-					}),
-					handleGrpcRequest(mockController, mockPostMessageToWebview, {
-						service: serviceName,
-						method: "testStreaming",
-						message: { id: 2 },
-						request_id: "concurrent-2",
-						is_streaming: true,
-					}),
-					handleGrpcRequest(mockController, mockPostMessageToWebview, {
-						service: serviceName,
-						method: "testUnary",
-						message: { id: 3 },
-						request_id: "concurrent-3",
-						is_streaming: false,
-					}),
-				]
-
-				await Promise.all(requests)
-
-				// Verify all handlers were called
-				expect(mockUnaryHandler.callCount).to.equal(2)
-				expect(mockStreamingHandler.callCount).to.equal(1)
-
-				// Verify all responses were sent (2 unary + 2 streaming)
-				expect(mockPostMessageToWebview.callCount).to.equal(4)
-			})
-		})
-	})
-})

src/core/controller/grpc-handler.ts

@@ -1,8 +1,6 @@
 import { Controller } from "./index"
 import { serviceHandlers } from "@generated/hosts/vscode/protobus-services"
 import { GrpcRequestRegistry } from "./grpc-request-registry"
-import { GrpcCancel, GrpcRequest } from "@/shared/WebviewMessage"
-import { ExtensionMessage } from "@/shared/ExtensionMessage"
 
 /**
  * Type definition for a streaming response handler
@@ -13,122 +11,184 @@ export type StreamingResponseHandler<TResponse> = (
 	sequenceNumber?: number,
 ) => Promise<void>
 
-export type PostMessageToWebview = (message: ExtensionMessage) => Thenable<boolean | undefined>
+/**
+ * Handles gRPC requests from the webview
+ */
+export class GrpcHandler {
+	constructor(private controller: Controller) {}
+
+	/**
+	 * Handle a gRPC request from the webview
+	 * @param service The service name
+	 * @param method The method name
+	 * @param message The request message
+	 * @param requestId The request ID for response correlation
+	 * @param isStreaming Whether this is a streaming request
+	 * @returns The response message or error for unary requests, void for streaming requests
+	 */
+	async handleRequest(
+		service: string,
+		method: string,
+		message: any,
+		requestId: string,
+		isStreaming: boolean = false,
+	): Promise<{
+		message?: any
+		error?: string
+		request_id: string
+	} | void> {
+		try {
+			// If this is a streaming request, use the streaming handler
+			if (isStreaming) {
+				await this.handleStreamingRequest(service, method, message, requestId)
+				return
+			}
+
+			// Get the service handler from the config
+			const handler = getHandler(service, method)
+
+			// Handle unary request
+			return {
+				message: await handler(this.controller, message),
+				request_id: requestId,
+			}
+		} catch (error) {
+			console.log("Protobus error:", error)
+			return {
+				error: error instanceof Error ? error.message : String(error),
+				request_id: requestId,
+			}
+		}
+	}
+
+	/**
+	 * Handle a streaming gRPC request
+	 * @param service The service name
+	 * @param method The method name
+	 * @param message The request message
+	 * @param requestId The request ID for response correlation
+	 */
+	private async handleStreamingRequest(service: string, method: string, message: any, requestId: string): Promise<void> {
+		// Create a response stream function
+		const responseStream: StreamingResponseHandler<any> = async (
+			response: any,
+			isLast: boolean = false,
+			sequenceNumber?: number,
+		) => {
+			await this.controller.postMessageToWebview({
+				type: "grpc_response",
+				grpc_response: {
+					message: response,
+					request_id: requestId,
+					is_streaming: !isLast,
+					sequence_number: sequenceNumber,
+				},
+			})
+		}
+
+		try {
+			// Get the service handler from the config
+			const handler = getHandler(service, method)
+
+			// Handle streaming request and pass the requestId to all streaming handlers
+			await handler(this.controller, message, responseStream, requestId)
+
+			// Don't send a final message here - the stream should stay open for future updates
+			// The stream will be closed when the client disconnects or when the service explicitly ends it
+		} catch (error) {
+			// Send error response
+			console.log("Protobus error:", error)
+			await this.controller.postMessageToWebview({
+				type: "grpc_response",
+				grpc_response: {
+					error: error instanceof Error ? error.message : String(error),
+					request_id: requestId,
+					is_streaming: false,
+				},
+			})
+		}
+	}
+}
+
+// Registry to track active gRPC requests and their cleanup functions
+const requestRegistry = new GrpcRequestRegistry()
 
 /**
- * Handles a gRPC request from the webview.
+ * Handle a gRPC request from the webview
+ * @param controller The controller instance
+ * @param request The gRPC request
  */
 export async function handleGrpcRequest(
 	controller: Controller,
-	postMessageToWebview: PostMessageToWebview,
-	request: GrpcRequest,
-): Promise<void> {
-	if (request.is_streaming) {
-		await handleStreamingRequest(controller, postMessageToWebview, request)
-	} else {
-		await handleUnaryRequest(controller, postMessageToWebview, request)
-	}
-}
-
-/**
- * Handles a gRPC unary request from the webview.
- *
- * Calls the handler using the service and method name, and then posts the result back to the webview.
- */
-async function handleUnaryRequest(
-	controller: Controller,
-	postMessageToWebview: PostMessageToWebview,
-	request: GrpcRequest,
-): Promise<void> {
+	request: {
+		service: string
+		method: string
+		message: any
+		request_id: string
+		is_streaming?: boolean
+	},
+) {
 	try {
-		// Get the service handler from the config
-		const handler = getHandler(request.service, request.method)
-		// Handle unary request
-		const response = await handler(controller, request.message)
-		// Send response to the webview
-		await postMessageToWebview({
+		const grpcHandler = new GrpcHandler(controller)
+
+		// For streaming requests, handleRequest handles sending responses directly
+		if (request.is_streaming) {
+			try {
+				await grpcHandler.handleRequest(request.service, request.method, request.message, request.request_id, true)
+			} finally {
+				// Note: We don't automatically clean up here anymore
+				// The request will be cleaned up when it completes or is cancelled
+			}
+			return
+		}
+
+		// For unary requests, we get a response and send it back
+		const response = (await grpcHandler.handleRequest(
+			request.service,
+			request.method,
+			request.message,
+			request.request_id,
+			false,
+		)) as {
+			message?: any
+			error?: string
+			request_id: string
+		}
+
+		// Send the response back to the webview
+		await controller.postMessageToWebview({
 			type: "grpc_response",
-			grpc_response: {
-				message: response,
-				request_id: request.request_id,
-			},
+			grpc_response: response,
 		})
 	} catch (error) {
 		// Send error response
 		console.log("Protobus error:", error)
-		await postMessageToWebview({
+		await controller.postMessageToWebview({
 			type: "grpc_response",
 			grpc_response: {
 				error: error instanceof Error ? error.message : String(error),
 				request_id: request.request_id,
-				is_streaming: false,
 			},
 		})
 	}
 }
 
 /**
- * Handle a streaming gRPC request from the webview.
- *
- * Calls the handler using the service and method name, and creates a streaming response handler
- * which posts results back to the webview.
- */
-async function handleStreamingRequest(
-	controller: Controller,
-	postMessageToWebview: PostMessageToWebview,
-	request: GrpcRequest,
-): Promise<void> {
-	// Create a response stream function
-	const responseStream: StreamingResponseHandler<any> = async (
-		response: any,
-		isLast: boolean = false,
-		sequenceNumber?: number,
-	) => {
-		await postMessageToWebview({
-			type: "grpc_response",
-			grpc_response: {
-				message: response,
-				request_id: request.request_id,
-				is_streaming: !isLast,
-				sequence_number: sequenceNumber,
-			},
-		})
-	}
-
-	try {
-		// Get the service handler from the config
-		const handler = getHandler(request.service, request.method)
-
-		// Handle streaming request and pass the requestId to all streaming handlers
-		await handler(controller, request.message, responseStream, request.request_id)
-
-		// Don't send a final message here - the stream should stay open for future updates
-		// The stream will be closed when the client disconnects or when the service explicitly ends it
-	} catch (error) {
-		// Send error response
-		console.log("Protobus error:", error)
-		await postMessageToWebview({
-			type: "grpc_response",
-			grpc_response: {
-				error: error instanceof Error ? error.message : String(error),
-				request_id: request.request_id,
-				is_streaming: false,
-			},
-		})
-	}
-}
-
-/**
- * Handles a gRPC request cancellation from the webview.
+ * Handle a gRPC request cancellation from the webview
  * @param controller The controller instance
  * @param request The cancellation request
  */
-export async function handleGrpcRequestCancel(postMessageToWebview: PostMessageToWebview, request: GrpcCancel) {
+export async function handleGrpcRequestCancel(
+	controller: Controller,
+	request: {
+		request_id: string
+	},
+) {
 	const cancelled = requestRegistry.cancelRequest(request.request_id)
 
 	if (cancelled) {
 		// Send a cancellation confirmation
-		await postMessageToWebview({
+		await controller.postMessageToWebview({
 			type: "grpc_response",
 			grpc_response: {
 				message: { cancelled: true },
@@ -141,17 +201,6 @@ export async function handleGrpcRequestCancel(postMessageToWebview: PostMessageT
 	}
 }
 
-// Registry to track active gRPC requests and their cleanup functions
-const requestRegistry = new GrpcRequestRegistry()
-
-/**
- * Get the request registry instance
- * This allows other parts of the code to access the registry
- */
-export function getRequestRegistry(): GrpcRequestRegistry {
-	return requestRegistry
-}
-
 function getHandler(serviceName: string, methodName: string): any {
 	// Get the service handler from the config
 	const serviceConfig = serviceHandlers[serviceName]
@@ -164,3 +213,11 @@ function getHandler(serviceName: string, methodName: string): any {
 	}
 	return handler
 }
+
+/**
+ * Get the request registry instance
+ * This allows other parts of the code to access the registry
+ */
+export function getRequestRegistry(): GrpcRequestRegistry {
+	return requestRegistry
+}

src/core/controller/index.ts

@@ -1,24 +1,26 @@
 import { clineEnvConfig } from "@/config"
 import { HostProvider } from "@/hosts/host-provider"
 import { AuthService } from "@/services/auth/AuthService"
-import { PostHogClientProvider, telemetryService } from "@/services/posthog/PostHogClientProvider"
+import { telemetryService } from "@/services/posthog/telemetry/TelemetryService"
 import { ShowMessageType } from "@/shared/proto/host/window"
-import { getLatestAnnouncementId } from "@/utils/announcements"
 import { getCwd, getDesktopDir } from "@/utils/path"
 import { Anthropic } from "@anthropic-ai/sdk"
 import { buildApiHandler } from "@api/index"
 import { cleanupLegacyCheckpoints } from "@integrations/checkpoints/CheckpointMigration"
 import { downloadTask } from "@integrations/misc/export-markdown"
+import WorkspaceTracker from "@integrations/workspace/WorkspaceTracker"
 import { ClineAccountService } from "@services/account/ClineAccountService"
 import { McpHub } from "@services/mcp/McpHub"
 import { ApiProvider, ModelInfo } from "@shared/api"
 import { ChatContent } from "@shared/ChatContent"
-import { ExtensionState, Platform } from "@shared/ExtensionMessage"
+import { Mode } from "@shared/storage/types"
+import { ClineRulesToggles } from "@shared/cline-rules"
+import { ExtensionMessage, ExtensionState, Platform } from "@shared/ExtensionMessage"
 import { HistoryItem } from "@shared/HistoryItem"
 import { McpMarketplaceCatalog } from "@shared/mcp"
-import { Mode } from "@shared/storage/types"
 import { TelemetrySetting } from "@shared/TelemetrySetting"
 import { UserInfo } from "@shared/UserInfo"
+import { WebviewMessage } from "@shared/WebviewMessage"
 import { fileExistsAtPath } from "@utils/fs"
 import axios from "axios"
 import fs from "fs/promises"
@@ -26,13 +28,15 @@ 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 { CacheService, PersistenceErrorEvent } from "../storage/CacheService"
 import { ensureMcpServersDirectoryExists, ensureSettingsDirectoryExists, GlobalFileNames } from "../storage/disk"
+import { getAllExtensionState, getGlobalState, getWorkspaceState, storeSecret, updateGlobalState } from "../storage/state"
+import { CacheService, PersistenceErrorEvent } from "../storage/CacheService"
 import { Task } from "../task"
+import { handleGrpcRequest, handleGrpcRequestCancel } from "./grpc-handler"
 import { sendMcpMarketplaceCatalogEvent } from "./mcp/subscribeToMcpMarketplaceCatalog"
 import { sendStateUpdate } from "./state/subscribeToState"
-import { sendAddToInputEvent, sendAddToInputEventToClient } from "./ui/subscribeToAddToInput"
-import { WebviewProvider } from "../webview"
+import { sendAddToInputEvent } from "./ui/subscribeToAddToInput"
+import { getLatestAnnouncementId } from "@/utils/announcements"
 
 /*
 https://github.com/microsoft/vscode-webview-ui-toolkit-samples/blob/main/default/weather-webview/src/providers/WeatherViewProvider.ts
@@ -42,20 +46,25 @@ https://github.com/KumarVariable/vscode-extension-sidebar-html/blob/master/src/c
 
 export class Controller {
 	readonly id: string
+	private postMessage: (message: ExtensionMessage) => Thenable<boolean> | undefined
+
 	private disposables: vscode.Disposable[] = []
 	task?: Task
 
+	workspaceTracker: WorkspaceTracker
 	mcpHub: McpHub
 	accountService: ClineAccountService
 	readonly cacheService: CacheService
 
 	constructor(
 		readonly context: vscode.ExtensionContext,
+		postMessage: (message: ExtensionMessage) => Thenable<boolean> | undefined,
 		id: string,
 	) {
 		this.id = id
 
 		HostProvider.get().logToChannel("ClineProvider instantiated")
+		this.postMessage = postMessage
 		this.accountService = ClineAccountService.getInstance()
 		this.cacheService = new CacheService(context)
 		const authService = AuthService.getInstance(this)
@@ -89,9 +98,11 @@ export class Controller {
 			}
 		}
 
+		this.workspaceTracker = new WorkspaceTracker()
 		this.mcpHub = new McpHub(
 			() => ensureMcpServersDirectoryExists(),
 			() => ensureSettingsDirectoryExists(this.context),
+			(msg) => this.postMessageToWebview(msg),
 			this.context.extension?.packageJSON?.version ?? "1.0.0",
 		)
 
@@ -102,7 +113,7 @@ export class Controller {
 	}
 
 	async getCurrentMode(): Promise<Mode> {
-		return this.cacheService.getGlobalStateKey("mode")
+		return ((await getGlobalState(this.context, "mode")) as Mode | undefined) || "act"
 	}
 
 	/*
@@ -118,6 +129,7 @@ export class Controller {
 				x.dispose()
 			}
 		}
+		this.workspaceTracker.dispose()
 		this.mcpHub.dispose()
 
 		console.error("Controller disposed")
@@ -128,7 +140,7 @@ export class Controller {
 		try {
 			// TODO: update to clineAccountId and then move clineApiKey to a clear function.
 			this.cacheService.setSecret("clineAccountId", undefined)
-			this.cacheService.setGlobalState("userInfo", undefined)
+			await updateGlobalState(this.context, "userInfo", undefined)
 
 			// Update API providers through cache service
 			const apiConfiguration = this.cacheService.getApiConfiguration()
@@ -153,34 +165,35 @@ export class Controller {
 	}
 
 	async setUserInfo(info?: UserInfo) {
-		this.cacheService.setGlobalState("userInfo", info)
+		await updateGlobalState(this.context, "userInfo", info)
 	}
 
 	async initTask(task?: string, images?: string[], files?: string[], historyItem?: HistoryItem) {
 		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
 
+		// Get API configuration from cache for immediate access
 		const apiConfiguration = this.cacheService.getApiConfiguration()
-		const autoApprovalSettings = this.cacheService.getGlobalStateKey("autoApprovalSettings")
-		const browserSettings = this.cacheService.getGlobalStateKey("browserSettings")
-		const focusChainSettings = this.cacheService.getGlobalStateKey("focusChainSettings")
-		const focusChainFeatureFlagEnabled = this.cacheService.getGlobalStateKey("focusChainFeatureFlagEnabled")
-		const preferredLanguage = this.cacheService.getGlobalStateKey("preferredLanguage")
-		const openaiReasoningEffort = this.cacheService.getGlobalStateKey("openaiReasoningEffort")
-		const mode = this.cacheService.getGlobalStateKey("mode")
-		const shellIntegrationTimeout = this.cacheService.getGlobalStateKey("shellIntegrationTimeout")
-		const terminalReuseEnabled = this.cacheService.getGlobalStateKey("terminalReuseEnabled")
-		const terminalOutputLineLimit = this.cacheService.getGlobalStateKey("terminalOutputLineLimit")
-		const defaultTerminalProfile = this.cacheService.getGlobalStateKey("defaultTerminalProfile")
-		const enableCheckpointsSetting = this.cacheService.getGlobalStateKey("enableCheckpointsSetting")
-		const isNewUser = this.cacheService.getGlobalStateKey("isNewUser")
-		const taskHistory = this.cacheService.getGlobalStateKey("taskHistory")
-		const strictPlanModeEnabled = this.cacheService.getGlobalStateKey("strictPlanModeEnabled")
+
+		const {
+			autoApprovalSettings,
+			browserSettings,
+			preferredLanguage,
+			openaiReasoningEffort,
+			mode,
+			shellIntegrationTimeout,
+			terminalReuseEnabled,
+			terminalOutputLineLimit,
+			defaultTerminalProfile,
+			enableCheckpointsSetting,
+			isNewUser,
+			taskHistory,
+		} = await getAllExtensionState(this.context)
 
 		const NEW_USER_TASK_COUNT_THRESHOLD = 10
 
 		// Check if the user has completed enough tasks to no longer be considered a "new user"
 		if (isNewUser && !historyItem && taskHistory && taskHistory.length >= NEW_USER_TASK_COUNT_THRESHOLD) {
-			this.cacheService.setGlobalState("isNewUser", false)
+			await updateGlobalState(this.context, "isNewUser", false)
 			await this.postStateToWebview()
 		}
 
@@ -189,17 +202,12 @@ export class Controller {
 				...autoApprovalSettings,
 				version: (autoApprovalSettings.version ?? 1) + 1,
 			}
-			this.cacheService.setGlobalState("autoApprovalSettings", updatedAutoApprovalSettings)
+			await updateGlobalState(this.context, "autoApprovalSettings", updatedAutoApprovalSettings)
 		}
-		// Apply remote feature flag gate to focus chain settings
-		const effectiveFocusChainSettings = {
-			...(focusChainSettings || { enabled: false, remindClineInterval: 6 }),
-			enabled: Boolean(focusChainSettings?.enabled) && Boolean(focusChainFeatureFlagEnabled),
-		}
-
 		this.task = new Task(
-			this,
+			this.context,
 			this.mcpHub,
+			this.workspaceTracker,
 			(historyItem) => this.updateTaskHistory(historyItem),
 			() => this.postStateToWebview(),
 			(taskId) => this.reinitExistingTaskFromId(taskId),
@@ -207,11 +215,9 @@ export class Controller {
 			apiConfiguration,
 			autoApprovalSettings,
 			browserSettings,
-			effectiveFocusChainSettings,
 			preferredLanguage,
 			openaiReasoningEffort,
 			mode,
-			strictPlanModeEnabled ?? false,
 			shellIntegrationTimeout,
 			terminalReuseEnabled ?? true,
 			terminalOutputLineLimit ?? 500,
@@ -233,8 +239,43 @@ export class Controller {
 		}
 	}
 
+	// Send any JSON serializable data to the react app
+	async postMessageToWebview(message: ExtensionMessage) {
+		await this.postMessage(message)
+	}
+
+	/**
+	 * Sets up an event listener to listen for messages passed from the webview context and
+	 * executes code based on the message that is received.
+	 *
+	 * @param webview A reference to the extension webview
+	 */
+	async handleWebviewMessage(message: WebviewMessage) {
+		switch (message.type) {
+			case "fetchMcpMarketplace": {
+				await this.fetchMcpMarketplace(message.bool)
+				break
+			}
+			case "grpc_request": {
+				if (message.grpc_request) {
+					await handleGrpcRequest(this, message.grpc_request)
+				}
+				break
+			}
+			case "grpc_request_cancel": {
+				if (message.grpc_request_cancel) {
+					await handleGrpcRequestCancel(this, message.grpc_request_cancel)
+				}
+				break
+			}
+
+			// Add more switch case statements here as more webview message commands
+			// are created within the webview context (i.e. inside media/main.js)
+		}
+	}
+
 	async updateTelemetrySetting(telemetrySetting: TelemetrySetting) {
-		this.cacheService.setGlobalState("telemetrySetting", telemetrySetting)
+		await updateGlobalState(this.context, "telemetrySetting", telemetrySetting)
 		const isOptedIn = telemetrySetting !== "disabled"
 		telemetryService.updateTelemetryState(isOptedIn)
 		await this.postStateToWebview()
@@ -244,21 +285,21 @@ export class Controller {
 		const didSwitchToActMode = modeToSwitchTo === "act"
 
 		// Store mode to global state
-		this.cacheService.setGlobalState("mode", modeToSwitchTo)
+		await updateGlobalState(this.context, "mode", modeToSwitchTo)
 
 		// Capture mode switch telemetry | Capture regardless of if we know the taskId
-		telemetryService.captureModeSwitch(this.task?.ulid ?? "0", modeToSwitchTo)
+		telemetryService.captureModeSwitch(this.task?.taskId ?? "0", modeToSwitchTo)
 
 		// Update API handler with new mode (buildApiHandler now selects provider based on mode)
 		if (this.task) {
 			const apiConfiguration = this.cacheService.getApiConfiguration()
-			this.task.api = buildApiHandler({ ...apiConfiguration, ulid: this.task.ulid }, modeToSwitchTo)
+			this.task.api = buildApiHandler({ ...apiConfiguration, taskId: this.task.taskId }, modeToSwitchTo)
 		}
 
 		await this.postStateToWebview()
 
 		if (this.task) {
-			this.task.updateMode(modeToSwitchTo)
+			this.task.mode = modeToSwitchTo
 			if (this.task.taskState.isAwaitingPlanResponse && didSwitchToActMode) {
 				this.task.taskState.didRespondToPlanAskBySwitchingMode = true
 				// Use chatContent if provided, otherwise use default message
@@ -304,8 +345,7 @@ export class Controller {
 				this.task.taskState.abandoned = true
 			}
 			await this.initTask(undefined, undefined, undefined, historyItem) // clears task again, so we need to abortTask manually above
-			// Dont send the state to the webview, the new Cline instance will send state when it's ready.
-			// Sending the state here sent an empty messages array to webview leading to virtuoso having to reload the entire list
+			// 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
 		}
 	}
 
@@ -316,14 +356,13 @@ export class Controller {
 			const clineProvider: ApiProvider = "cline"
 
 			// Get current settings to determine how to update providers
-			const planActSeparateModelsSetting = this.cacheService.getGlobalStateKey("planActSeparateModelsSetting")
-
+			const { planActSeparateModelsSetting } = await getAllExtensionState(this.context)
 			const currentMode = await this.getCurrentMode()
 
 			// Get current API configuration from cache
 			const currentApiConfiguration = this.cacheService.getApiConfiguration()
 
-			const updatedConfig = { ...currentApiConfiguration }
+			let updatedConfig = { ...currentApiConfiguration }
 
 			if (planActSeparateModelsSetting) {
 				// Only update the current mode's provider
@@ -342,10 +381,10 @@ export class Controller {
 			this.cacheService.setApiConfiguration(updatedConfig)
 
 			// Mark welcome view as completed since user has successfully logged in
-			this.cacheService.setGlobalState("welcomeViewCompleted", true)
+			await updateGlobalState(this.context, "welcomeViewCompleted", true)
 
 			if (this.task) {
-				this.task.api = buildApiHandler({ ...updatedConfig, ulid: this.task.ulid }, currentMode)
+				this.task.api = buildApiHandler({ ...updatedConfig, taskId: this.task.taskId }, currentMode)
 			}
 
 			await this.postStateToWebview()
@@ -383,7 +422,7 @@ export class Controller {
 			}
 
 			// Store in global state
-			this.cacheService.setGlobalState("mcpMarketplaceCatalog", catalog)
+			await updateGlobalState(this.context, "mcpMarketplaceCatalog", catalog)
 			return catalog
 		} catch (error) {
 			console.error("Failed to fetch MCP marketplace:", error)
@@ -421,7 +460,7 @@ export class Controller {
 			}
 
 			// Store in global state
-			this.cacheService.setGlobalState("mcpMarketplaceCatalog", catalog)
+			await updateGlobalState(this.context, "mcpMarketplaceCatalog", catalog)
 			return catalog
 		} catch (error) {
 			console.error("Failed to fetch MCP marketplace:", error)
@@ -446,7 +485,7 @@ export class Controller {
 
 	/**
 	 * RPC variant that silently refreshes the MCP marketplace catalog and returns the result
-	 * Unlike silentlyRefreshMcpMarketplace, this doesn't send a message to the webview
+	 * Unlike silentlyRefreshMcpMarketplace, this doesn't post a message to the webview
 	 * @returns MCP marketplace catalog or undefined if refresh failed
 	 */
 	async silentlyRefreshMcpMarketplaceRPC() {
@@ -458,6 +497,31 @@ export class Controller {
 		}
 	}
 
+	private async fetchMcpMarketplace(forceRefresh: boolean = false) {
+		try {
+			// Check if we have cached data
+			const cachedCatalog = (await getGlobalState(this.context, "mcpMarketplaceCatalog")) as
+				| McpMarketplaceCatalog
+				| undefined
+			if (!forceRefresh && cachedCatalog?.items) {
+				await sendMcpMarketplaceCatalogEvent(cachedCatalog)
+				return
+			}
+
+			const catalog = await this.fetchMcpMarketplaceFromApi(false)
+			if (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"
+			HostProvider.window.showMessage({
+				type: ShowMessageType.ERROR,
+				message: errorMessage,
+			})
+		}
+	}
+
 	// OpenRouter
 
 	async handleOpenRouterCallback(code: string) {
@@ -489,9 +553,9 @@ export class Controller {
 
 		await this.postStateToWebview()
 		if (this.task) {
-			this.task.api = buildApiHandler({ ...updatedConfig, ulid: this.task.ulid }, currentMode)
+			this.task.api = buildApiHandler({ ...updatedConfig, taskId: this.task.taskId }, currentMode)
 		}
-		// Dont send settingsButtonClicked because its bad ux if user is on welcome
+		// await this.postMessageToWebview({ type: "action", action: "settingsButtonClicked" }) // bad ux if user is on welcome
 	}
 
 	private async ensureCacheDirectoryExists(): Promise<string> {
@@ -524,6 +588,10 @@ export class Controller {
 
 	// 'Add to Cline' context menu in editor and code action
 	async addSelectedCodeToChat(code: string, filePath: string, languageId: string, diagnostics?: vscode.Diagnostic[]) {
+		// Ensure the sidebar view is visible
+		await vscode.commands.executeCommand("claude-dev.SidebarProvider.focus")
+		await setTimeoutPromise(100)
+
 		// Post message to webview with the selected code
 		const fileMention = await this.getFileMentionFromPath(filePath)
 
@@ -533,10 +601,7 @@ export class Controller {
 			input += `\nProblems:\n${problemsString}`
 		}
 
-		const lastActiveWebview = WebviewProvider.getLastActiveInstance()
-		if (lastActiveWebview) {
-			await sendAddToInputEventToClient(lastActiveWebview.getClientId(), input)
-		}
+		await sendAddToInputEvent(input)
 
 		console.log("addSelectedCodeToChat", code, filePath, languageId)
 	}
@@ -546,6 +611,14 @@ export class Controller {
 		// Ensure the sidebar view is visible
 		await vscode.commands.executeCommand("claude-dev.SidebarProvider.focus")
 		await setTimeoutPromise(100)
+
+		// Post message to webview with the selected terminal output
+		// await this.postMessageToWebview({
+		//     type: "addSelectedTerminalOutput",
+		//     output,
+		//     terminalName
+		// })
+
 		await sendAddToInputEvent(`Terminal output:\n\`\`\`\n${output}\n\`\`\``)
 
 		console.log("addSelectedTerminalOutputToChat", output, terminalName)
@@ -603,7 +676,7 @@ export class Controller {
 		taskMetadataFilePath: string
 		apiConversationHistory: Anthropic.MessageParam[]
 	}> {
-		const history = this.cacheService.getGlobalStateKey("taskHistory")
+		const history = ((await getGlobalState(this.context, "taskHistory")) as HistoryItem[] | undefined) || []
 		const historyItem = history.find((item) => item.id === id)
 		if (historyItem) {
 			const taskDirPath = path.join(this.context.globalStorageUri.fsPath, "tasks", id)
@@ -638,9 +711,9 @@ export class Controller {
 
 	async deleteTaskFromState(id: string) {
 		// Remove the task from history
-		const taskHistory = this.cacheService.getGlobalStateKey("taskHistory")
+		const taskHistory = ((await getGlobalState(this.context, "taskHistory")) as HistoryItem[] | undefined) || []
 		const updatedTaskHistory = taskHistory.filter((task) => task.id !== id)
-		this.cacheService.setGlobalState("taskHistory", updatedTaskHistory)
+		await updateGlobalState(this.context, "taskHistory", updatedTaskHistory)
 
 		// Notify the webview that the task has been deleted
 		await this.postStateToWebview()
@@ -656,35 +729,35 @@ export class Controller {
 	async getStateToPostToWebview(): Promise<ExtensionState> {
 		// Get API configuration from cache for immediate access
 		const apiConfiguration = this.cacheService.getApiConfiguration()
-		const lastShownAnnouncementId = this.cacheService.getGlobalStateKey("lastShownAnnouncementId")
-		const taskHistory = this.cacheService.getGlobalStateKey("taskHistory")
-		const autoApprovalSettings = this.cacheService.getGlobalStateKey("autoApprovalSettings")
-		const browserSettings = this.cacheService.getGlobalStateKey("browserSettings")
-		const focusChainSettings = this.cacheService.getGlobalStateKey("focusChainSettings")
-		const focusChainFeatureFlagEnabled = this.cacheService.getGlobalStateKey("focusChainFeatureFlagEnabled")
-		const preferredLanguage = this.cacheService.getGlobalStateKey("preferredLanguage")
-		const openaiReasoningEffort = this.cacheService.getGlobalStateKey("openaiReasoningEffort")
-		const mode = this.cacheService.getGlobalStateKey("mode")
-		const strictPlanModeEnabled = this.cacheService.getGlobalStateKey("strictPlanModeEnabled")
-		const userInfo = this.cacheService.getGlobalStateKey("userInfo")
-		const mcpMarketplaceEnabled = this.cacheService.getGlobalStateKey("mcpMarketplaceEnabled")
-		const mcpDisplayMode = this.cacheService.getGlobalStateKey("mcpDisplayMode")
-		const telemetrySetting = this.cacheService.getGlobalStateKey("telemetrySetting")
-		const planActSeparateModelsSetting = this.cacheService.getGlobalStateKey("planActSeparateModelsSetting")
-		const enableCheckpointsSetting = this.cacheService.getGlobalStateKey("enableCheckpointsSetting")
-		const globalClineRulesToggles = this.cacheService.getGlobalStateKey("globalClineRulesToggles")
-		const globalWorkflowToggles = this.cacheService.getGlobalStateKey("globalWorkflowToggles")
-		const shellIntegrationTimeout = this.cacheService.getGlobalStateKey("shellIntegrationTimeout")
-		const terminalReuseEnabled = this.cacheService.getGlobalStateKey("terminalReuseEnabled")
-		const defaultTerminalProfile = this.cacheService.getGlobalStateKey("defaultTerminalProfile")
-		const isNewUser = this.cacheService.getGlobalStateKey("isNewUser")
-		const welcomeViewCompleted = this.cacheService.getGlobalStateKey("welcomeViewCompleted")
-		const mcpResponsesCollapsed = this.cacheService.getGlobalStateKey("mcpResponsesCollapsed")
-		const terminalOutputLineLimit = this.cacheService.getGlobalStateKey("terminalOutputLineLimit")
-		const localClineRulesToggles = this.cacheService.getWorkspaceStateKey("localClineRulesToggles")
-		const localWindsurfRulesToggles = this.cacheService.getWorkspaceStateKey("localWindsurfRulesToggles")
-		const localCursorRulesToggles = this.cacheService.getWorkspaceStateKey("localCursorRulesToggles")
-		const workflowToggles = this.cacheService.getWorkspaceStateKey("workflowToggles")
+
+		const {
+			lastShownAnnouncementId,
+			taskHistory,
+			autoApprovalSettings,
+			browserSettings,
+			preferredLanguage,
+			openaiReasoningEffort,
+			mode,
+			userInfo,
+			mcpMarketplaceEnabled,
+			mcpDisplayMode,
+			telemetrySetting,
+			planActSeparateModelsSetting,
+			enableCheckpointsSetting,
+			globalClineRulesToggles,
+			globalWorkflowToggles,
+			shellIntegrationTimeout,
+			terminalReuseEnabled,
+			defaultTerminalProfile,
+			isNewUser,
+			welcomeViewCompleted,
+			mcpResponsesCollapsed,
+			terminalOutputLineLimit,
+			localClineRulesToggles,
+			localWindsurfRulesToggles,
+			localCursorRulesToggles,
+			localWorkflowToggles,
+		} = await getAllExtensionState(this.context)
 
 		const currentTaskItem = this.task?.taskId ? (taskHistory || []).find((item) => item.id === this.task?.taskId) : undefined
 		const checkpointTrackerErrorMessage = this.task?.taskState.checkpointTrackerErrorMessage
@@ -698,7 +771,7 @@ export class Controller {
 		const latestAnnouncementId = getLatestAnnouncementId(this.context)
 		const shouldShowAnnouncement = lastShownAnnouncementId !== latestAnnouncementId
 		const platform = process.platform as Platform
-		const distinctId = PostHogClientProvider.getInstance().distinctId
+		const distinctId = telemetryService.distinctId
 		const version = this.context.extension?.packageJSON?.version ?? ""
 		const uriScheme = vscode.env.uriScheme
 
@@ -709,18 +782,14 @@ export class Controller {
 			currentTaskItem,
 			checkpointTrackerErrorMessage,
 			clineMessages,
-			currentFocusChainChecklist: this.task?.taskState.currentFocusChainChecklist || null,
 			taskHistory: processedTaskHistory,
 			shouldShowAnnouncement,
 			platform,
 			autoApprovalSettings,
 			browserSettings,
-			focusChainSettings,
-			focusChainFeatureFlagEnabled,
 			preferredLanguage,
 			openaiReasoningEffort,
 			mode,
-			strictPlanModeEnabled,
 			userInfo,
 			mcpMarketplaceEnabled,
 			mcpDisplayMode,
@@ -732,7 +801,7 @@ export class Controller {
 			localClineRulesToggles: localClineRulesToggles || {},
 			localWindsurfRulesToggles: localWindsurfRulesToggles || {},
 			localCursorRulesToggles: localCursorRulesToggles || {},
-			localWorkflowToggles: workflowToggles || {},
+			localWorkflowToggles: localWorkflowToggles || {},
 			globalWorkflowToggles: globalWorkflowToggles || {},
 			shellIntegrationTimeout,
 			terminalReuseEnabled,
@@ -769,15 +838,37 @@ export class Controller {
 	For now we'll store the conversation history in memory, and if we need to store in state directly we'd need to do a manual conversion to ensure proper json stringification.
 	*/
 
+	// getApiConversationHistory(): Anthropic.MessageParam[] {
+	// 	// const history = (await this.getGlobalState(
+	// 	// 	this.getApiConversationHistoryStateKey()
+	// 	// )) as Anthropic.MessageParam[]
+	// 	// return history || []
+	// 	return this.apiConversationHistory
+	// }
+
+	// setApiConversationHistory(history: Anthropic.MessageParam[] | undefined) {
+	// 	// await this.updateGlobalState(this.getApiConversationHistoryStateKey(), history)
+	// 	this.apiConversationHistory = history || []
+	// }
+
+	// addMessageToApiConversationHistory(message: Anthropic.MessageParam): Anthropic.MessageParam[] {
+	// 	// const history = await this.getApiConversationHistory()
+	// 	// history.push(message)
+	// 	// await this.setApiConversationHistory(history)
+	// 	// return history
+	// 	this.apiConversationHistory.push(message)
+	// 	return this.apiConversationHistory
+	// }
+
 	async updateTaskHistory(item: HistoryItem): Promise<HistoryItem[]> {
-		const history = this.cacheService.getGlobalStateKey("taskHistory")
+		const history = ((await getGlobalState(this.context, "taskHistory")) as HistoryItem[]) || []
 		const existingItemIndex = history.findIndex((h) => h.id === item.id)
 		if (existingItemIndex !== -1) {
 			history[existingItemIndex] = item
 		} else {
 			history.push(item)
 		}
-		this.cacheService.setGlobalState("taskHistory", history)
+		await updateGlobalState(this.context, "taskHistory", history)
 		return history
 	}
 }

src/core/controller/models/refreshBasetenModels.ts

@@ -1,232 +0,0 @@
-import { Controller } from ".."
-import { EmptyRequest } from "@shared/proto/cline/common"
-import { OpenRouterCompatibleModelInfo, OpenRouterModelInfo } from "@shared/proto/cline/models"
-import { basetenModels } from "../../../shared/api"
-import axios from "axios"
-import path from "path"
-import fs from "fs/promises"
-import { fileExistsAtPath } from "@utils/fs"
-import { GlobalFileNames } from "@core/storage/disk"
-
-/**
- * Refreshes the Baseten models and returns the updated model list
- * @param controller The controller instance
- * @param request Empty request object
- * @returns Response containing the Baseten models
- */
-export async function refreshBasetenModels(
-	controller: Controller,
-	request: EmptyRequest,
-): Promise<OpenRouterCompatibleModelInfo> {
-	console.log("=== refreshBasetenModels called ===")
-	const basetenModelsFilePath = path.join(await ensureCacheDirectoryExists(controller), GlobalFileNames.basetenModels)
-
-	// Get the Baseten API key from the controller's state
-	const basetenApiKey = controller.cacheService.getSecretKey("basetenApiKey")
-
-	let models: Record<string, Partial<OpenRouterModelInfo>> = {}
-	try {
-		if (!basetenApiKey) {
-			console.log("No Baseten API key found, using static models as fallback")
-			// Don't throw an error, just use static models
-			for (const [modelId, modelInfo] of Object.entries(basetenModels)) {
-				models[modelId] = {
-					maxTokens: modelInfo.maxTokens,
-					contextWindow: modelInfo.contextWindow,
-					supportsImages: modelInfo.supportsImages,
-					supportsPromptCache: modelInfo.supportsPromptCache,
-					inputPrice: modelInfo.inputPrice,
-					outputPrice: modelInfo.outputPrice,
-					cacheWritesPrice: (modelInfo as any).cacheWritesPrice || 0,
-					cacheReadsPrice: (modelInfo as any).cacheReadsPrice || 0,
-					description: (modelInfo as any).description || `${modelId} model`,
-				}
-			}
-		} else {
-			// Ensure the API key is properly formatted
-			const cleanApiKey = basetenApiKey.trim()
-			if (!cleanApiKey) {
-				throw new Error("Invalid Baseten API key format")
-			}
-
-			console.log("Fetching Baseten models with API key:", cleanApiKey.substring(0, 10) + "...")
-
-			const response = await axios.get("https://inference.baseten.co/v1/models", {
-				headers: {
-					Authorization: `Bearer ${cleanApiKey}`,
-					"Content-Type": "application/json",
-					"User-Agent": "Cline-VSCode-Extension",
-				},
-				timeout: 10000, // 10 second timeout
-			})
-
-			if (response.data?.data) {
-				const rawModels = response.data.data
-
-				for (const rawModel of rawModels) {
-					// Filter out non-chat models and validate model capabilities
-					if (!isValidChatModel(rawModel)) {
-						continue
-					}
-
-					// Only include models that are listed in the static basetenModels
-					if (!(rawModel.id in basetenModels)) {
-						console.log(`Skipping model ${rawModel.id} - not in static basetenModels list`)
-						continue
-					}
-
-					// Check if we have static pricing information for this model
-					const staticModelInfo = basetenModels[rawModel.id as keyof typeof basetenModels]
-
-					const modelInfo: Partial<OpenRouterModelInfo> = {
-						maxTokens: staticModelInfo?.maxTokens || 8192,
-						contextWindow: staticModelInfo?.contextWindow || 8192,
-						supportsImages: staticModelInfo?.supportsImages || false,
-						supportsPromptCache: staticModelInfo?.supportsPromptCache || false,
-						inputPrice: staticModelInfo?.inputPrice || 0,
-						outputPrice: staticModelInfo?.outputPrice || 0,
-						cacheWritesPrice: staticModelInfo?.cacheWritesPrice || 0,
-						cacheReadsPrice: staticModelInfo?.cacheReadsPrice || 0,
-						description: generateModelDescription(rawModel, staticModelInfo),
-					}
-
-					models[rawModel.id] = modelInfo
-				}
-			} else {
-				console.error("Invalid response from Baseten API")
-			}
-			await fs.writeFile(basetenModelsFilePath, JSON.stringify(models))
-			console.log("Baseten models fetched and saved:", Object.keys(models))
-		}
-	} catch (error) {
-		console.error("Error fetching Baseten models:", error)
-
-		// Provide more specific error messages
-		let errorMessage = "Unknown error occurred"
-		if (axios.isAxiosError(error)) {
-			if (error.response?.status === 401) {
-				errorMessage = "Invalid Baseten API key. Please check your API key in settings."
-			} else if (error.response?.status === 403) {
-				errorMessage = "Access forbidden. Please verify your Baseten API key has the correct permissions."
-			} else if (error.response?.status === 429) {
-				errorMessage = "Rate limit exceeded. Please try again later."
-			} else if (error.code === "ECONNABORTED") {
-				errorMessage = "Request timeout. Please check your internet connection."
-			} else {
-				errorMessage = `API request failed: ${error.response?.status || error.code || "Unknown error"}`
-			}
-		} else if (error instanceof Error) {
-			errorMessage = error.message
-		}
-
-		console.error("Baseten API Error:", errorMessage)
-
-		// If we failed to fetch models, try to read cached models first
-		const cachedModels = await readBasetenModels(controller)
-		if (cachedModels && Object.keys(cachedModels).length > 0) {
-			console.log("Using cached Baseten models")
-			// Filter cached models to only include those in static basetenModels
-			for (const [modelId, modelInfo] of Object.entries(cachedModels)) {
-				if (modelId in basetenModels) {
-					models[modelId] = modelInfo
-				}
-			}
-		} else {
-			// Fall back to static models from shared/api.ts
-			console.log("Using static Baseten models as fallback")
-			for (const [modelId, modelInfo] of Object.entries(basetenModels)) {
-				models[modelId] = {
-					maxTokens: modelInfo.maxTokens,
-					contextWindow: modelInfo.contextWindow,
-					supportsImages: modelInfo.supportsImages,
-					supportsPromptCache: modelInfo.supportsPromptCache,
-					inputPrice: modelInfo.inputPrice,
-					outputPrice: modelInfo.outputPrice,
-					cacheWritesPrice: (modelInfo as any).cacheWritesPrice || 0,
-					cacheReadsPrice: (modelInfo as any).cacheReadsPrice || 0,
-					description: (modelInfo as any).description || `${modelId} model`,
-				}
-			}
-		}
-	}
-
-	// 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 ?? 8192,
-			contextWindow: model.contextWindow ?? 8192,
-			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 ?? "",
-			tiers: model.tiers ?? [],
-		}
-	}
-
-	return OpenRouterCompatibleModelInfo.create({ models: typedModels })
-}
-
-/**
- * Ensures the cache directory exists and returns its path
- */
-async function ensureCacheDirectoryExists(controller: Controller): Promise<string> {
-	const cacheDir = path.join(controller.context.globalStorageUri.fsPath, "cache")
-	await fs.mkdir(cacheDir, { recursive: true })
-	return cacheDir
-}
-
-/**
- * Reads cached Baseten models from disk
- */
-async function readBasetenModels(controller: Controller): Promise<Record<string, Partial<OpenRouterModelInfo>> | undefined> {
-	const basetenModelsFilePath = path.join(await ensureCacheDirectoryExists(controller), GlobalFileNames.basetenModels)
-	const fileExists = await fileExistsAtPath(basetenModelsFilePath)
-	if (fileExists) {
-		try {
-			const fileContents = await fs.readFile(basetenModelsFilePath, "utf8")
-			return JSON.parse(fileContents)
-		} catch (error) {
-			console.error("Error reading cached Baseten models:", error)
-			return undefined
-		}
-	}
-	return undefined
-}
-
-/**
- * Validates if a model is suitable for chat completions
- */
-function isValidChatModel(rawModel: any): boolean {
-	// Filter out non-chat models (whisper, TTS, guard models, etc.)
-	if (rawModel.id.includes("whisper") || rawModel.id.includes("tts") || rawModel.id.includes("embedding")) {
-		return false
-	}
-
-	// Check if model supports chat completions
-	if (rawModel.object === "model" && rawModel.id) {
-		return true
-	}
-
-	return false
-}
-
-/**
- * Generates a descriptive name for the model
- */
-function generateModelDescription(rawModel: any, staticModelInfo?: any): string {
-	// Use static description if available
-	if (staticModelInfo?.description) {
-		return staticModelInfo.description
-	}
-
-	// Generate description based on model characteristics
-	const modelId = rawModel.id
-	const ownedBy = rawModel.owned_by || "Unknown"
-
-	return `${ownedBy} model: ${modelId}`
-}

src/core/controller/models/refreshGroqModels.ts

@@ -1,13 +1,13 @@
 import { Controller } from ".."
 import { EmptyRequest } from "@shared/proto/cline/common"
 import { OpenRouterCompatibleModelInfo, OpenRouterModelInfo } from "@shared/proto/cline/models"
+import { getAllExtensionState } from "../../storage/state"
 import { groqModels } from "../../../shared/api"
 import axios from "axios"
 import path from "path"
 import fs from "fs/promises"
 import { fileExistsAtPath } from "@utils/fs"
 import { GlobalFileNames } from "@core/storage/disk"
-import { telemetryService } from "@/services/posthog/PostHogClientProvider"
 
 /**
  * Refreshes the Groq models and returns the updated model list
@@ -18,7 +18,9 @@ import { telemetryService } from "@/services/posthog/PostHogClientProvider"
 export async function refreshGroqModels(controller: Controller, request: EmptyRequest): Promise<OpenRouterCompatibleModelInfo> {
 	const groqModelsFilePath = path.join(await ensureCacheDirectoryExists(controller), GlobalFileNames.groqModels)
 
-	const groqApiKey = controller.cacheService.getSecretKey("groqApiKey")
+	// Get the Groq API key from the controller's state
+	const { apiConfiguration } = await getAllExtensionState(controller.context)
+	const groqApiKey = apiConfiguration?.groqApiKey
 
 	let models: Record<string, Partial<OpenRouterModelInfo>> = {}
 	try {
@@ -109,12 +111,7 @@ export async function refreshGroqModels(controller: Controller, request: EmptyRe
 			errorMessage = error.message
 		}
 
-		telemetryService.captureProviderApiError({
-			ulid: controller.task?.ulid || "",
-			errorMessage,
-			errorStatus: error.status,
-			model: "groq",
-		})
+		console.error("Groq API Error:", errorMessage)
 
 		// If we failed to fetch models, try to read cached models first
 		const cachedModels = await readGroqModels(controller)
@@ -184,7 +181,7 @@ async function readGroqModels(controller: Controller): Promise<Record<string, Pa
  */
 function isValidChatModel(rawModel: any): boolean {
 	// Check if model is active (if the property exists)
-	if (Object.hasOwn(rawModel, "active") && !rawModel.active) {
+	if (rawModel.hasOwnProperty("active") && !rawModel.active) {
 		return false
 	}
 	// Filter out non-chat models (whisper, TTS, guard models, etc.)