# JARVIS Windows — Project Roadmap

## Phase 1 — Core Stack ✅ (Week 1)
- [ ] FastAPI server with WebSocket endpoint
- [ ] Whisper STT (base model, local)
- [ ] Fish Audio TTS integration
- [ ] Claude API connection
- [ ] WebSocket manager (connect/disconnect/broadcast)
- [ ] Basic end-to-end voice loop (speak → transcribe → Claude → respond)

## Phase 2 — Google Workspace Integrations (Weeks 2–3)
### 2a — Auth + Core Services
- [ ] Google OAuth 2.0 setup (credentials.json → token.json)
- [ ] Google Calendar: read today's events, upcoming schedule
- [ ] Gmail: unread messages, search by sender/subject
- [ ] Google Tasks: list pending, create, complete

### 2b — Notes + Screen
- [ ] Google Keep via gkeepapi (or Markdown fallback)
- [ ] Screen vision: mss screenshot + win32gui active window
- [ ] Window list enumeration

### 2c — System Capabilities
- [ ] Terminal control via subprocess (PowerShell + CMD)
- [ ] File system: create projects, manage folders
- [ ] Git status/operations via subprocess
- [ ] Chrome automation via Selenium

## Phase 3 — Claude Tool Use (Week 4)
- [ ] Define tools manifest (all capabilities as function specs)
- [ ] Tool dispatch router (route Claude tool_use to capability modules)
- [ ] Multi-turn conversation with tool results
- [ ] Error handling + tool fallbacks

## Phase 4 — Dashboard + Packaging (Week 5)
- [ ] Wire WebSocket to Three.js dashboard frontend
- [ ] Live transcript display
- [ ] State: idle / listening / processing / speaking
- [ ] Windows startup registration (Task Scheduler)
- [ ] .env validation on startup

## Phase 5 — Optional Enhancements
- [ ] Docker containerization (for Unraid or headless server)
- [ ] Wake word detection ("Hey JARVIS")
- [ ] Conversation history / memory across sessions
- [ ] Google Drive file access
- [ ] React/Electron desktop wrapper