WrenchAI

AI-powered DIY car repair assistant with OBD-II diagnostics.

What It Does

1. Connect to your car via Bluetooth OBD-II dongle
2. Read diagnostic trouble codes (DTCs) and VIN
3. Interpret codes with AI
4. Guide you through repairs
5. Link to parts you need

Project Structure

wrenchai/
├── WrenchAI/          # React Native mobile app
│   ├── src/
│   │   ├── services/  # OBD, NHTSA, AI, Parts
│   │   ├── screens/   # UI screens
│   │   └── utils/
│   ├── App.js         # Main entry
│   └── app.json       # Expo config
│
└── backend/           # Node.js API
    └── server.js      # Proxies AI calls

Quick Start

Prerequisites

• Node.js 20+
• Expo CLI: npm install -g expo-cli
• Venice AI API key: https://venice.ai

1. Backend Setup

cd backend
cp .env.example .env
# Edit .env with your Venice API key
npm install
npm start

Backend runs on http://localhost:3000

2. Mobile App Setup

cd WrenchAI
npm install

Update src/config.js with your backend URL.

3. Run on Device

# Start Expo
npx expo start

# Scan QR code with Expo Go app on your phone

Hardware Required

• OBD-II Bluetooth dongle (ELM327 compatible)
  • Recommended: Veepeak, OBDLink LX, or similar
  • Avoid super cheap clones (connectivity issues)

Features

Feature	Status
Bluetooth OBD connection	✅ Basic
VIN auto-read	⚠️ Partial
DTC reading	✅ Basic
AI interpretation	✅ Via backend
Parts affiliate links	✅
Repair instructions	✅ Via backend

Testing

1. Plug OBD dongle into car port (usually under dash)
2. Turn ignition to ON (don't need to start engine)
3. Open app, tap "Scan for OBD Device"
4. Select device when found
5. View codes and AI analysis

What's Missing (TODO)

• Full VIN multi-frame parsing
• Complete DTC parsing (different protocols)
• Real-time sensor data (live PIDs)
• Offline mode with cached responses
• iOS/Android native app builds (currently Expo Go only)
• App Store submission

Architecture

Phone (React Native)
├── Bluetooth → OBD-II Dongle → Car ECU
└── HTTP → Backend API → Venice AI

Backend (Node.js)
├── Receives DTCs from phone
└── Calls Venice AI for interpretation

Cost

• Venice AI: $0-50/mo (free tier available)
• OBD Dongle: $20-50 (one-time)
• Hosting: $0 (local dev) or $5-20/mo (cloud)

License

MIT - Built for the DIY community.