- Track tab creation and closure events in TabManager
- Add session tracking (created, resumed, completed) in ClaudeCodeSession
- Track model selection changes in ClaudeCodeSession
- Monitor agent execution events (success/failure) in AgentExecution
- Include execution duration metrics for agents
- Use useTrackEvent hook for consistent event tracking
- Add HooksEditor component for managing project hooks
- Add ProjectSettings component for project-specific configurations
- Create hooksManager utility for hook operations
- Add hooks type definitions
- Update backend commands to support hooks functionality
- Integrate hooks into main app, agent execution, and Claude sessions
- Update API and utilities to handle hooks data
- Cross-check database running sessions with actual process registry
- Filter out stale database entries for crashed processes without cleanup
- Add comprehensive comments explaining the reliability improvements
- Fix sticky header z-index layering issue in AgentExecution component
This addresses cases where agent processes crash without properly updating
the database status, ensuring the UI shows only truly active sessions.
Resolves#87 and #9 by implementing a robust three-tier process
termination approach:
1. ProcessRegistry kill - primary method using run_id tracking
2. ClaudeProcessState kill - fallback via stored process handle
3. System kill command - last resort using PID and OS commands
Key improvements:
- Enhanced logging throughout termination flow for better debugging
- Graceful fallback between termination methods
- Proper UI state management even when backend termination fails
- Track run_id in AgentExecution component for targeted process killing
- Comprehensive error handling with user-friendly feedback
- Consistent event emission for UI synchronization
This ensures agents can be properly stopped without requiring
application restart, addressing the core issue where STOP
requests were ignored and processes continued running.
This commit provides a comprehensive solution to Claude binary detection issues
by implementing a user-friendly version selector UI and improving the binary
discovery logic. It addresses all concerns raised in multiple PRs and comments.
Changes:
- Add ClaudeVersionSelector component for selecting from multiple installations
- Update ClaudeBinaryDialog to use version selector instead of manual path input
- Fix unused variable warning in production builds (claude.rs:442)
- Improve select_best_installation to handle production build restrictions
- Add listClaudeInstallations API endpoint to fetch all available installations
- Make Claude version indicator clickable to navigate to Settings
- Move Claude installation selector to General tab in Settings (per user request)
- Enhance dialog UX with loading states and clear installation instructions
- Add Radix UI radio-group dependency for version selector
Fixes:
- Production build warning about unused claude_path variable
- Version detection failures in production builds due to process restrictions
- Poor UX when Claude binary is not found (now shows helpful dialog)
- Inability to easily switch between multiple Claude installations
This implementation takes inspiration from:
- PR #3: Version selector dropdown approach (preferred by users)
- PR #4: Binary detection improvements and path validation
- PR #39: Additional detection methods and error handling
- Commit 5a29f9a: Shared claude binary detection module architecture
Addresses feedback from:
- getAsterisk/claudia#4 (comment): User preference for dropdown selector
- Production build restrictions that prevent version detection
- Need for better error handling when Claude is not installed
The solution provides a seamless experience whether Claude is installed via:
- npm/yarn/bun global installation
- nvm-managed Node.js versions
- Homebrew on macOS
- System-wide installation
- Local user installation (~/.local/bin, etc.)
Refs: #3, #4, #39, 5a29f9a
- Add run_id/session_id based event isolation for concurrent executions
- Enhance process registry with graceful shutdown and fallback kill methods
- Implement session-specific event listeners in React components
- Add proper process cleanup with timeout handling
- Support both isolated and backward-compatible event emissions
- Improve error handling and logging for process management
This change prevents event crosstalk between multiple concurrent agent/claude sessions
running simultaneously, ensuring proper isolation and user experience.
Restructure AgentExecution component layout to enhance user experience:
- Make header and configuration sections sticky to remain visible during scrolling
- Optimize scrollable output display area with proper overflow handling
- Improve container structure for better layout management and responsiveness
- Maintain existing functionality while providing better visual organization
- Refactor message filtering logic across AgentExecution, ClaudeCodeSession, SessionOutputViewer, and StreamMessage components to better handle visibility of user messages with tool results
- Replace inefficient every() loop with for-loop for better performance when checking message content visibility
- Add renderedSomething tracking in StreamMessage to prevent rendering empty components
- Optimize SessionOutputViewer with useMemo for displayableMessages filtering
- Remove unused GitHub API response fields (url, html_url, git_url) in Rust agents command
- Add Vite build configuration with manual code splitting for better bundle optimization
- Remove TypeScript expect-error comment for process global variable
These changes improve UI performance by reducing unnecessary re-renders and bundle size through better code splitting.
- Replace static rendering with @tanstack/react-virtual
- Optimize rendering for long conversation histories
- Maintain smooth auto-scroll behavior
- Integrate with enhanced message system from upstream
Significantly improves performance when handling extensive
agent execution outputs and long Claude sessions.
