Brummer MCP Server

# AI Coder PTY Integration Plan ## Overview This plan outlines the integration of the PTY (pseudo-terminal) system into Brummer's main TUI, transforming the AI Coder tab from a simple process view to a full-featured tmux-style terminal interface. ## Current State Analysis ### What's Already Built 1. **PTY Foundation** (`internal/aicoder/`) - `pty_session.go` - Core PTY session with terminal emulation - `pty_manager.go` - Multi-session management - `data_injector.go` - Contextual data injection system - `brummer_data_provider.go` - Integration with Brummer's data 2. **TUI Components** (`internal/tui/`) - `ai_coder_pty_view.go` - Standalone PTY view component - `ai_coder_view.go` - Current AI coder tab (needs integration) - `model.go` - Main TUI model with tab switching 3. **Manager Integration** - `AICoderManager` has PTY support methods - CLI command configuration for various tools - Interactive vs non-interactive mode handling ### Integration Gaps 1. AI Coder tab (View 8) still shows simple process list 2. PTY view not connected to main TUI navigation 3. `/ai` command doesn't create PTY sessions 4. No automatic PTY output streaming to TUI 5. Missing session persistence across tab switches ## Integration Architecture ### Phase 1: Connect PTY View to AI Coder Tab **Goal**: Replace the current AI coder view with the PTY-enabled view 1. **Update `model.go`** - Add `aiCoderPTYView *AICoderPTYView` field - Initialize PTY view with BrummerDataProvider - Route View 8 rendering to PTY view 2. **Create BrummerDataProvider Implementation** - Bridge between TUI Model and PTY data requirements - Access to logStore, errorDetector, proxy data - Real-time data injection capabilities 3. **Update Tab Switching Logic** - Preserve PTY session state when switching tabs - Handle terminal focus state transitions - Maintain full-screen mode across tab switches ### Phase 2: Integrate /ai Command with PTY **Goal**: Make `/ai` command create interactive PTY sessions 1. **Update Command Handler** - Modify `handleAICommand` in `model.go` - Create PTY session instead of regular process - Auto-attach to new session in AI Coder tab 2. **Session Creation Flow** ``` /ai claude → CreateInteractiveCLISession("claude") /ai claude "fix tests" → CreateTaskCLISession("claude", "fix tests") ``` 3. **Automatic Tab Switching** - Switch to AI Coder tab when creating session - Focus terminal for immediate interaction - Show session info in status bar ### Phase 3: Event Integration **Goal**: Connect PTY events to TUI update cycle 1. **PTY Event Routing** - Create `PTYEventListener` in TUI model - Convert PTY events to tea.Msg types - Handle output, close, resize events 2. **Automatic Event Forwarding** - Monitor Brummer events (errors, test failures) - Forward to active PTY session in debug mode - Visual indicators for auto-injected data 3. **Session Lifecycle Management** - Handle session creation/deletion - Update session list in real-time - Clean up resources on session close ### Phase 4: Enhanced Features **Goal**: Polish the integration for natural workflow 1. **Session Persistence** - Maintain sessions across Brummer restarts - Save/restore terminal buffer state - Reconnect to running CLI processes 2. **Multi-Session Workflow** - Quick session switching (Ctrl+1-9) - Session overview panel - Visual session activity indicators 3. **Smart Defaults** - Auto-create session on first `/ai` use - Remember last used AI provider - Context-aware initial prompts ## Implementation Tasks ### Task 1: Create TUI Data Provider **File**: `internal/tui/brummer_data_provider_impl.go` - Implement `BrummerDataProvider` interface - Access TUI model's data stores - Thread-safe data access methods ### Task 2: Update Model Structure **File**: `internal/tui/model.go` - Add PTY-related fields - Initialize PTY manager with data provider - Update view initialization ### Task 3: Route AI Coder View **File**: `internal/tui/model.go` (View() method) - Replace case 8 rendering - Pass through window size messages - Handle PTY-specific commands ### Task 4: Create PTY Event Bridge **File**: `internal/tui/pty_events.go` - PTY event to tea.Msg conversion - Event subscription management - Async event handling ### Task 5: Update AI Command Handler **File**: `internal/tui/model.go` (handleAICommand) - Parse provider and task from command - Create appropriate PTY session - Auto-switch to AI Coder tab ### Task 6: Add Output Streaming **File**: `internal/tui/model.go` (Init/Update) - Subscribe to PTY output events - Convert to tea.Cmd for updates - Handle backpressure ### Task 7: Implement Session Management **File**: `internal/tui/ai_session_manager.go` - Session state persistence - Graceful reconnection logic - Resource cleanup ### Task 8: Add Visual Polish **Files**: Various TUI components - Session activity indicators - Status bar integration - Help text updates ## Testing Strategy ### Unit Tests 1. Data provider implementation 2. Event conversion logic 3. Session management operations ### Integration Tests 1. `/ai` command → PTY session creation 2. Tab switching with active sessions 3. Data injection key bindings 4. Full-screen mode transitions ### Manual Testing Scenarios 1. Create multiple AI sessions 2. Switch between sessions rapidly 3. Inject data during active conversation 4. Test terminal resize handling 5. Verify cleanup on exit ## Migration Path ### Step 1: Parallel Implementation - Keep existing AI coder view functional - Build PTY integration alongside - Feature flag for PTY mode ### Step 2: Gradual Rollout - Enable PTY for new sessions only - Migrate existing sessions on demand - Collect feedback and iterate ### Step 3: Full Cutover - Remove old AI coder view code - Clean up legacy process handling - Update documentation ## Success Criteria 1. **Seamless Integration** - `/ai` command "just works" - Natural tab navigation - Intuitive key bindings 2. **Performance** - Smooth terminal rendering - No lag in typing - Efficient event handling 3. **Reliability** - Sessions persist correctly - Clean shutdown/cleanup - No resource leaks 4. **User Experience** - Clear visual feedback - Helpful error messages - Discoverable features ## Next Steps 1. Review and approve this plan 2. Create detailed task breakdown 3. Begin implementation with Task 1 4. Iterate based on testing feedback This integration will transform Brummer's AI Coder feature from a simple process viewer to a powerful, tmux-style development environment where AI assistants feel like true pair programming partners.