- Changed model identifier from `claude-opus-4-5-20251101` to `claude-opus-4-6` across various files, including documentation and code references.
- Updated the SDK to support adaptive thinking for Opus 4.6, allowing the model to determine its own reasoning depth.
- Enhanced the thinking level options to include 'adaptive' and adjusted related components to reflect this change.
- Updated tests to ensure compatibility with the new model and its features.
These changes improve the model's capabilities and user experience by leveraging adaptive reasoning.
* feat: add GitHub Copilot SDK provider integration
Adds comprehensive GitHub Copilot SDK provider support including:
- CopilotProvider class with CLI detection and OAuth authentication check
- Copilot models definition with GPT-4o, Claude, and o1/o3 series models
- Settings UI integration with provider tab, model configuration, and navigation
- Onboarding flow integration with Copilot setup step
- Model selector integration for all phase-specific model dropdowns
- Persistence of enabled models and default model settings via API sync
- Server route for Copilot CLI status endpoint
https://claude.ai/code/session_01D26w7ZyEzP4H6Dor3ttk9d
* chore: update package-lock.json
https://claude.ai/code/session_01D26w7ZyEzP4H6Dor3ttk9d
* refactor: rename Copilot SDK to Copilot CLI and use GitHub icon
- Update all references from "GitHub Copilot SDK" to "GitHub Copilot CLI"
- Change install command from @github/copilot-sdk to @github/copilot
- Update CopilotIcon to use official GitHub Octocat logo
- Update error codes and comments throughout codebase
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: update Copilot model definitions and add dynamic model discovery
- Update COPILOT_MODEL_MAP with correct models from CLI (claude-sonnet-4.5,
claude-haiku-4.5, claude-opus-4.5, claude-sonnet-4, gpt-5.x series, gpt-4.1,
gemini-3-pro-preview)
- Change default Copilot model to copilot-claude-sonnet-4.5
- Add model caching methods to CopilotProvider (hasCachedModels,
clearModelCache, refreshModels)
- Add API routes for dynamic model discovery:
- GET /api/setup/copilot/models
- POST /api/setup/copilot/models/refresh
- POST /api/setup/copilot/cache/clear
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* refactor: use @github/copilot-sdk instead of direct CLI calls
- Install @github/copilot-sdk package for proper SDK integration
- Rewrite CopilotProvider to use SDK's CopilotClient API
- Use client.createSession() for session management
- Handle SDK events (assistant.message, tool.execution_*, session.idle)
- Auto-approve permissions for autonomous agent operation
- Remove incorrect CLI flags (--mode, --output-format)
- Update default model to claude-sonnet-4.5
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: add Copilot and Gemini model support to model resolver
- Import isCopilotModel and isGeminiModel from types
- Add explicit checks for copilot- and gemini- prefixed models
- Pass through Copilot/Gemini models unchanged to their providers
- Update resolver documentation to list all supported providers
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: pass working directory to Copilot SDK and reduce event noise
- Create CopilotClient per execution with correct cwd from options.cwd
- This ensures the CLI operates in the correct project directory, not the
server's current directory
- Skip assistant.message_delta events (they create excessive noise)
- Only yield the final assistant.message event which has complete content
- Clean up client on completion and error paths
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: simplify Copilot SDK execution with sendAndWait
- Use sendAndWait() instead of manual event polling for more reliable
execution
- Disable streaming (streaming: false) to simplify response handling
- Increase timeout to 10 minutes for agentic operations
- Still capture tool execution events for UI display
- Add more debug logging for troubleshooting
- This should fix the "invalid_request_body" error on subsequent calls
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: allow Copilot model IDs with claude-, gemini-, gpt- prefixes
Copilot's bare model IDs legitimately contain prefixes like claude-,
gemini-, gpt- because those are the actual model names from the
Copilot CLI (e.g., claude-sonnet-4.5, gemini-3-pro-preview, gpt-5.1).
The generic validateBareModelId function was incorrectly rejecting
these valid model IDs. Now we only check that the copilot- prefix
has been stripped by the ProviderFactory.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat: enable real-time streaming of tool events for Copilot
- Switch back to streaming mode (streaming: true) for real-time events
- Use async queue pattern to bridge SDK callbacks to async generator
- Events are now yielded as they happen, not batched at the end
- Tool calls (Read, Write, Edit, Bash, TodoWrite, etc.) show in real-time
- Better progress visibility during agentic operations
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat: expand Copilot tool name and input normalization
Tool name mapping additions:
- view → Read (Copilot's file viewing tool)
- create_file → Write
- replace, patch → Edit
- run_shell_command, terminal → Bash
- search_file_content → Grep
- list_directory → Ls
- google_web_search → WebSearch
- report_intent → ReportIntent (Copilot-specific planning)
- think, plan → Think, Plan
Input normalization improvements:
- Read/Write/Edit: Map file, filename, filePath → file_path
- Bash: Map cmd, script → command
- Grep: Map query, search, regex → pattern
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: convert git+ssh to git+https in package-lock.json
The @electron/node-gyp dependency was resolved with a git+ssh URL
which fails in CI environments without SSH keys. Convert to HTTPS.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: address code review feedback for Copilot SDK provider
- Add guard for non-text prompts (vision not yet supported)
- Clear runtime model cache on fetch failure
- Fix race condition in async queue error handling
- Import CopilotAuthStatus from shared types
- Fix comment mismatch for default model constant
- Add auth-copilot and deauth-copilot routes
- Extract shared tool normalization utilities
- Create base model configuration UI component
- Add comprehensive unit tests for CopilotProvider
- Replace magic strings with constants
- Add debug logging for cleanup errors
* fix: address CodeRabbit review nitpicks
- Fix test mocks to include --version check for CLI detection
- Add aria-label for accessibility on refresh button
- Ensure default model checkbox always appears checked/enabled
* fix: address CodeRabbit review feedback
- Fix test mocks by creating fresh provider instances after mock setup
- Extract COPILOT_DISCONNECTED_MARKER_FILE constant to common.ts
- Add AUTONOMOUS MODE comment explaining auto-approval of permissions
- Improve tool-normalization with union types and null guards
- Handle 'canceled' (American spelling) status in todo normalization
* refactor: extract copilot connection logic to service and fix test mocks
- Create copilot-connection-service.ts with connect/disconnect logic
- Update auth-copilot and deauth-copilot routes to use service
- Fix test mocks for CLI detection:
- Mock fs.existsSync for CLI path validation
- Mock which/where command for CLI path detection
---------
Co-authored-by: Claude <noreply@anthropic.com>
* feat: add Gemini CLI provider for AI model execution
- Add GeminiProvider class extending CliProvider for Gemini CLI integration
- Add Gemini models (Gemini 3 Pro/Flash Preview, 2.5 Pro/Flash/Flash-Lite)
- Add gemini-models.ts with model definitions and types
- Update ModelProvider type to include 'gemini'
- Add isGeminiModel() to provider-utils.ts for model detection
- Register Gemini provider in provider-factory with priority 4
- Add Gemini setup detection routes (status, auth, deauth)
- Add GeminiCliStatus to setup store for UI state management
- Add Gemini to PROVIDER_ICON_COMPONENTS for UI icon display
- Add GEMINI_MODELS to model-display for dropdown population
- Support thinking levels: off, low, medium, high
Based on https://github.com/google-gemini/gemini-cli
* chore: update package-lock.json
* feat(ui): add Gemini provider to settings and setup wizard
- Add GeminiCliStatus component for CLI detection display
- Add GeminiSettingsTab component for global settings
- Update provider-tabs.tsx to include Gemini as 5th tab
- Update providers-setup-step.tsx with Gemini provider detection
- Add useGeminiCliStatus hook for querying CLI status
- Add getGeminiStatus, authGemini, deauthGemini to HTTP API client
- Add gemini query key for React Query
- Fix GeminiModelId type to not double-prefix model IDs
* feat(ui): add Gemini to settings sidebar navigation
- Add 'gemini-provider' to SettingsViewId type
- Add GeminiIcon and gemini-provider to navigation config
- Add gemini-provider to NAV_ID_TO_PROVIDER mapping
- Add gemini-provider case in settings-view switch
- Export GeminiSettingsTab from providers index
This fixes the missing Gemini entry in the AI Providers sidebar menu.
* feat(ui): add Gemini model configuration in settings
- Create GeminiModelConfiguration component for model selection
- Add enabledGeminiModels and geminiDefaultModel state to app-store
- Add setEnabledGeminiModels, setGeminiDefaultModel, toggleGeminiModel actions
- Update GeminiSettingsTab to show model configuration when CLI is installed
- Import GeminiModelId and getAllGeminiModelIds from types
This adds the ability to configure which Gemini models are available
in the feature modal, similar to other providers like Codex and OpenCode.
* feat(ui): add Gemini models to all model dropdowns
- Add GEMINI_MODELS to model-constants.ts for UI dropdowns
- Add Gemini to ALL_MODELS array used throughout the app
- Add GeminiIcon to PROFILE_ICONS mapping
- Fix GEMINI_MODELS in model-display.ts to use correct model IDs
- Update getModelDisplayName to handle Gemini models correctly
Gemini models now appear in all model selection dropdowns including
Model Defaults, Feature Defaults, and feature card settings.
* fix(gemini): fix CLI integration and event handling
- Fix model ID prefix handling: strip gemini- prefix in agent-service,
add it back in buildCliArgs for CLI invocation
- Fix event normalization to match actual Gemini CLI output format:
- type: 'init' (not 'system')
- type: 'message' with role (not 'assistant')
- tool_name/tool_id/parameters/output field names
- Add --sandbox false and --approval-mode yolo for faster execution
- Remove thinking level selector from UI (Gemini CLI doesn't support it)
- Update auth status to show errors properly
* test: update provider-factory tests for Gemini provider
- Add GeminiProvider import and spy mock
- Update expected provider count from 4 to 5
- Add test for GeminiProvider inclusion
- Add gemini key to checkAllProviders test
* fix(gemini): address PR review feedback
- Fix npm package name from @anthropic-ai/gemini-cli to @google/gemini-cli
- Fix comments in gemini-provider.ts to match actual CLI output format
- Convert sync fs operations to async using fs/promises
* fix(settings): add Gemini and Codex settings to sync
Add enabledGeminiModels, geminiDefaultModel, enabledCodexModels, and
codexDefaultModel to SETTINGS_FIELDS_TO_SYNC for persistence across sessions.
* fix(gemini): address additional PR review feedback
- Use 'Speed' badge for non-thinking Gemini models (consistency)
- Fix installCommand mapping in gemini-settings-tab.tsx
- Add hasEnvApiKey to GeminiCliStatus interface for API parity
- Clarify GeminiThinkingLevel comment (CLI doesn't support --thinking-level)
* fix(settings): restore Codex and Gemini settings from server
Add sanitization and restoration logic for enabledCodexModels,
codexDefaultModel, enabledGeminiModels, and geminiDefaultModel
in refreshSettingsFromServer() to match the fields in SETTINGS_FIELDS_TO_SYNC.
* feat(gemini): normalize tool names and fix workspace restrictions
- Add tool name mapping to normalize Gemini CLI tool names to standard
names (e.g., write_todos -> TodoWrite, read_file -> Read)
- Add normalizeGeminiToolInput to convert write_todos format to TodoWrite
format (description -> content, handle cancelled status)
- Pass --include-directories with cwd to fix workspace restriction errors
when Gemini CLI has a different cached workspace from previous sessions
---------
Co-authored-by: Claude <noreply@anthropic.com>
- Combined CLI disconnection markers with OpenCode support
- Added OpenCode auth/deauth routes and API methods
- Resolved merge conflicts between feature branch and upstream v0.9.0rc
- Fix Claude, Codex, and Cursor auth handlers to check if CLI is already authenticated
- Use same detection logic as each provider's internal checkAuth/codexAuthIndicators()
- For Codex: Check for API keys and auth files before requiring manual login
- For Cursor: Check for env var and credentials files before requiring manual auth
- For Claude: Check for cached auth tokens, settings, and credentials files
- If CLI is already authenticated: Just reconnect by removing disconnected marker
- If CLI needs auth: Tell user to manually run login command
- This prevents timeout errors when login commands can't run in non-interactive mode
- Introduced a new method to check Codex authentication status, allowing for better handling of API keys and OAuth tokens.
- Updated API key management to include OpenAI, enabling users to manage their keys more effectively.
- Enhanced the CodexProvider to support session ID tracking and deduplication of text blocks in assistant messages.
- Improved error handling and logging in authentication routes, providing clearer feedback to users.
These changes improve the overall user experience and security of the Codex integration, ensuring smoother authentication processes and better management of API keys.
- Introduced a new method in ProviderFactory to check if a model supports vision/image input.
- Updated AgentService and AutoModeService to validate vision support before processing images, throwing an error if the model does not support it.
- Enhanced error messages to guide users on switching models or removing images if necessary.
These changes improve the robustness of image processing by ensuring compatibility with the selected models.
- Refactor model handling to support both Claude and Cursor models across various components.
- Introduce `stripProviderPrefix` utility for consistent model ID processing.
- Update `CursorProvider` to utilize `isCursorModel` for model validation.
- Implement model override functionality in GitHub issue validation and enhancement routes.
- Add `useCursorStatusInit` hook to initialize Cursor CLI status on app startup.
- Update UI components to reflect changes in model selection and validation processes.
This update improves the flexibility of AI model usage and enhances user experience by allowing quick model overrides.
Replace hardcoded switch statements with dynamic registry pattern.
Providers register with factory using registerProvider() function.
New features:
- registerProvider() function for dynamic registration
- canHandleModel() callback for model routing
- priority field for controlling match order
- aliases support (e.g., 'anthropic' -> 'claude')
- getRegisteredProviderNames() for introspection
Adding new providers now only requires calling registerProvider()
with a factory function and model matching logic.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Replace 'claude' | 'cursor' literal unions with ModelProvider type
from @automaker/types for better extensibility when adding new providers.
- Update ProviderFactory.getProviderNameForModel() return type
- Update RunningFeature.provider type in auto-mode-service
- Update getRunningAgents() return type
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Added CursorProvider to the ProviderFactory for handling cursor-* models.
- Updated getProviderNameForModel method to determine the appropriate provider based on model identifiers.
- Enhanced getAllProviders method to return both ClaudeProvider and CursorProvider.
- Updated documentation to reflect the completion of the Provider Factory integration phase.
- Updated model resolution logic to only check for gpt-* models, removing references to unsupported o1/o3 models in both model-resolver and provider-factory files.
- Enhanced comments for clarity regarding model support in Codex CLI.
Implements a flexible provider pattern that supports both Claude Agent SDK
and OpenAI Codex CLI, enabling future expansion to other AI providers
(Cursor, OpenCode, etc.) with minimal changes.
## Architecture Changes
### New Provider System
- Created provider abstraction layer with BaseProvider interface
- Model-based routing: model prefix determines provider
- `gpt-*`, `o*` → CodexProvider (subprocess CLI)
- `claude-*`, `opus/sonnet/haiku` → ClaudeProvider (SDK)
- Providers implement common ExecuteOptions interface
### New Files Created
- `providers/types.ts` - Shared interfaces (ExecuteOptions, ProviderMessage, etc.)
- `providers/base-provider.ts` - Abstract base class
- `providers/claude-provider.ts` - Claude Agent SDK wrapper
- `providers/codex-provider.ts` - Codex CLI subprocess executor
- `providers/codex-cli-detector.ts` - Installation & auth detection
- `providers/codex-config-manager.ts` - TOML config management
- `providers/provider-factory.ts` - Model-based provider routing
- `lib/subprocess-manager.ts` - Reusable subprocess utilities
## Features Implemented
### Codex CLI Integration
- Spawns Codex CLI as subprocess with JSONL output
- Converts Codex events to Claude SDK-compatible format
- Supports both `codex login` and OPENAI_API_KEY auth methods
- Handles: reasoning, messages, commands, todos, file changes
- Extracts text from content blocks for non-vision CLI
### Conversation History
- Added conversationHistory support to ExecuteOptions
- ClaudeProvider: yields previous messages to SDK
- CodexProvider: prepends history as text context
- Follow-up prompts maintain full conversation context
### Image Upload Support
- Images embedded as base64 for vision models
- Image paths appended to prompt text for Read tool access
- Auto-mode: copies images to feature folder
- Follow-up: combines original + new images
- Updates feature.json with image metadata
### Session Model Persistence
- Added `model` field to Session and SessionMetadata
- Sessions remember model preference across interactions
- API endpoints accept model parameter
- Auto-mode respects feature's model setting
## Modified Files
### Services
- `agent-service.ts`:
- Added conversation history building
- Uses ProviderFactory instead of direct SDK calls
- Appends image paths to prompts
- Added model parameter and persistence
- `auto-mode-service.ts`:
- Removed OpenAI model block restriction
- Uses ProviderFactory for all models
- Added image support in buildFeaturePrompt
- Follow-up: loads context, copies images, updates feature.json
- Returns to waiting_approval after follow-up
### Routes
- `agent.ts`: Added model parameter to /send endpoint
- `sessions.ts`: Added model field to create/update
- `models.ts`: Added Codex models (gpt-5.2, gpt-5.1-codex*)
### Configuration
- `.env.example`: Added OPENAI_API_KEY and CODEX_CLI_PATH
- `.gitignore`: Added provider-specific ignores
## Bug Fixes
- Fixed image path resolution (relative → absolute)
- Fixed Codex empty prompt when images attached
- Fixed follow-up status management (in_progress → waiting_approval)
- Fixed follow-up images not appearing in prompt text
- Removed OpenAI model restrictions in auto-mode
## Testing Notes
- Codex CLI authentication verified with both methods
- Image uploads work for both Claude (vision) and Codex (Read tool)
- Follow-up prompts maintain full context
- Conversation history persists across turns
- Model switching works per-session
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
