---
title: Friday Subconscious
emoji: 🧠
colorFrom: blue
colorTo: indigo
sdk: docker
pinned: false
---

# Project Jarvis

> Personal AI Assistant — 100% Local, Zero Cloud Dependencies

Jarvis is a voice-first personal AI assistant that runs entirely on your Mac. No data leaves your machine. No API keys required for core functionality.

## 🧠 What Powers Jarvis (All Local)

| Component | Technology | Purpose |
|---|---|---|
| **LLM Brain** | Ollama (llama3:8b) | Intent classification, reasoning, responses |
| **Speech-to-Text** | OpenAI Whisper (local) | Transcribes your voice commands |
| **Text-to-Speech** | Kokoro TTS + macOS native | Speaks responses back to you |
| **Wake Word** | openWakeWord | "Hey Jarvis" detection |
| **Memory** | ChromaDB + SQLite | Persistent memory and data storage |
| **Backend** | Python FastAPI | Service orchestration on localhost:7777 |

## 🚀 Quick Start

### 1. Run Setup (one-time)
```bash
bash scripts/setup.sh
```
This installs all dependencies, downloads models, and configures everything.

### 2. Start Jarvis
```bash
bash scripts/start.sh
```

### 3. Use Jarvis

- **Web Dashboard**: Open `http://localhost:7777` in your browser
- **CLI Mode**: `cd backend && source venv/bin/activate && python cli.py`
- **Voice Mode**: `cd backend && source venv/bin/activate && python cli.py --voice`
- **API Docs**: `http://localhost:7777/docs`

### 4. Stop Jarvis
```bash
bash scripts/stop.sh
```

## 📁 Project Structure

```
Jarvis/
├── backend/            # Python FastAPI server
│   ├── app/
│   │   ├── api/        # REST API endpoints
│   │   ├── core/       # Brain, intent routing, context management
│   │   ├── services/   # STT, TTS, LLM, wake word, audio
│   │   └── models/     # Pydantic schemas
│   ├── main.py         # Server entry point
│   ├── config.py       # Configuration
│   └── cli.py          # Terminal interface
├── static/             # Web dashboard
├── app/                # macOS Swift menu bar app (Phase 2)
├── models/             # Downloaded AI models
├── data/               # Runtime data (SQLite, ChromaDB)
└── scripts/            # Setup, start, stop scripts
```

## 🎙️ Voice Commands (Phase 0)

| Say This | Jarvis Does |
|---|---|
| "Hey Jarvis, what time is it?" | Tells you the current time |
| "Good morning" | Greets you with daily context |
| "What can you do?" | Lists available capabilities |
| "Draft an email to the team" | Drafts content for you |
| "Tell me a joke" | Engages in conversation |

## 🛣️ Roadmap

- **Phase 0** ✅ Core voice loop, LLM brain, web dashboard
- **Phase 1** 🔜 Expense tracking, SQLite database, ChromaDB memory
- **Phase 2** 🔜 Contact intelligence, Google Sheets integration
- **Phase 3** 🔜 OttoPilot CRM integration
- **Phase 4** 🔜 WhatsApp, Gmail, Calendar
- **Phase 5** 🔜 Call recording & memory
- **Phase 6** 🔜 Analytics & reporting
- **Phase 7** 🔜 macOS system intelligence
- **Phase 8** 🔜 Polish & hardening
- **Phase 9** 🔜 Mobile app (React Native)

## ⚙️ Configuration

Edit `backend/.env` to customize:
- `OLLAMA_MODEL` — Change the LLM model (default: llama3:8b)
- `WHISPER_MODEL` — STT model size (tiny/base/small/medium/large)
- `TTS_ENGINE` — kokoro (neural) or macos (system voice)
- `OPENAI_API_KEY` — Optional GPT-4 fallback

---

*Built with ❤️ by Antigravity — Confidential*