Intelligent Vehicle Management Application | v1.3.1

A full-stack application for tracking maintenance, managing service reminders, and consulting your vehicle's documentation using AI-powered search with RAG (Retrieval-Augmented Generation).

• Dashboard - Vehicle overview with mileage tracking, maintenance forecast, CARFAX value estimate, and one-click reindexing
• Maintenance Log - Record oil changes, tire rotations, brake service, and more with cost tracking
• Receipt/Document Uploads - Attach PDFs, images, and receipts to maintenance records
• Service Records - Import CARFAX reports and track complete service history
• Smart Reminders - Date and mileage-based alerts with recurring support

• Floating Chat Widget - Ask questions about your vehicle from any page
• Context-Aware Chat - "Ask about this" button on maintenance records pre-fills contextual questions
• Document Search - Semantic search across owner's manual, QRG, and service records
• RAG Integration - Retrieval-Augmented Generation with source thumbnails
• MoE System - Mixture of Experts routing to specialized vehicle knowledge domains
• Local LLM Support - Run AI queries locally via Docker Model Runner with no cloud API costs

• Rich Maintenance Cards - Expandable cards with related manual page thumbnails
• Photo Management - Upload before/after/general photos for maintenance records
• Document Carousel - Browse related manual pages in timeline view
• Full-Size Image Modal - Click any thumbnail to view full-size document pages

• LLM Response Cache - Redis-backed cache for LLM responses to avoid redundant inference
• Permanent Search Cache - Vehicle queries and search results cached permanently in Redis
• Embedding Cache - Document embeddings cached to skip redundant computation
• Cache-Aware Reindexing - Stale LLM and search caches automatically flushed on document changes

• Automatic Document Processing - Upload PDFs and they're automatically ingested into both PostgreSQL and Qdrant
• Maintenance Record Embedding - Service history is embedded into the vector store for RAG search (e.g., "when was my last brake job?")
• Dual Vector Storage - pgvector for PostgreSQL + Qdrant for high-performance similarity search
• Background Processing - Non-blocking document ingestion and maintenance re-embedding on CRUD operations

git clone https://github.com/ry-ops/DriveIQ.git
cd DriveIQ

# Install PostgreSQL (macOS)
brew install postgresql@15
brew services start postgresql@15

# Create database
createdb driveiq
psql driveiq < database/init.sql

Create backend/.env:

# Database
DATABASE_URL=postgresql://localhost/driveiq

# Security
SECRET_KEY=your-secret-key-change-in-production
ALGORITHM=HS256
ACCESS_TOKEN_EXPIRE_MINUTES=30

# AI APIs
# ANTHROPIC_API_KEY=your-anthropic-api-key  # Not needed for local LLM

# Vector Database (optional)
QDRANT_HOST=localhost
QDRANT_PORT=6333
QDRANT_COLLECTION=driveiq_docs

# Redis
REDIS_URL=redis://localhost:6379

# Local LLM
USE_LOCAL_LLM=true
ANTHROPIC_BASE_URL=http://model-runner.docker.internal:12434
LOCAL_LLM_MODEL=ai/qwen3-coder

# Vehicle Info
VEHICLE_VIN=YOUR_VIN_HERE
VEHICLE_YEAR=2018
VEHICLE_MAKE=Toyota
VEHICLE_MODEL=4Runner
VEHICLE_TRIM=SR5 Premium

cd backend
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
uvicorn app.main:app --reload

Backend runs at: http://localhost:8000

Default credentials: admin / driveiq2024

cd frontend
npm install
npm run dev

Frontend runs at: http://localhost:3000

# Place PDFs in /docs directory
python scripts/ingest_documents.py

This processes the PDFs in /docs and creates vector embeddings for AI search using local sentence-transformers (no API key required for embeddings).

DriveIQ/
├── backend/                 # FastAPI application
│   ├── app/
│   │   ├── api/            # Route handlers
│   │   │   ├── auth.py
│   │   │   ├── vehicle.py
│   │   │   ├── maintenance.py  # + photo upload, related docs
│   │   │   ├── reminders.py
│   │   │   ├── search.py
│   │   │   ├── chat.py         # Chat widget endpoint
│   │   │   ├── uploads.py      # Auto-ingestion
│   │   │   ├── moe.py
│   │   │   └── import_data.py
│   │   ├── core/           # Config, database, security
│   │   │   ├── llm_client.py     # LLM abstraction (local inference)
│   │   │   ├── qdrant_client.py  # Qdrant integration
│   │   │   └── redis_client.py   # Caching, sessions, rate limiting
│   │   ├── models/         # SQLAlchemy models
│   │   ├── schemas/        # Pydantic schemas
│   │   └── services/       # Business logic
│   │       └── document_ingestion.py  # Dual vector ingestion
│   └── requirements.txt
├── frontend/               # React application
│   ├── src/
│   │   ├── components/     # Reusable components
│   │   │   ├── ChatWidget.tsx       # Floating chat
│   │   │   ├── MaintenanceCard.tsx  # Rich cards
│   │   │   └── ServiceHistoryTimeline.tsx
│   │   ├── context/
│   │   │   └── ChatContext.tsx      # Global chat state
│   │   ├── pages/          # Page components
│   │   ├── services/       # API client
│   │   └── types/          # TypeScript types
│   └── package.json
├── database/
│   └── init.sql           # Schema + seed data
├── docs/                   # Vehicle PDFs
│   ├── 4Runner Manual.pdf
│   ├── 4Runner QRG.pdf
│   └── CARFAX Report.pdf
└── scripts/
    └── ingest_documents.py # PDF embedding ingestion

• POST /api/auth/login - Login with username/password
• POST /api/auth/register - Register new user
• GET /api/auth/me - Get current user info

• GET /api/vehicle - Get vehicle information
• PATCH /api/vehicle - Update vehicle details
• PATCH /api/vehicle/mileage/{mileage} - Quick mileage update

• GET /api/maintenance - List all maintenance records
• POST /api/maintenance - Create new record (auto-syncs with reminders, re-embeds for RAG)
• PATCH /api/maintenance/{id} - Update record (re-embeds for RAG)
• DELETE /api/maintenance/{id} - Delete record (re-embeds for RAG)
• POST /api/maintenance/reindex - Manually trigger maintenance record re-embedding
• GET /api/maintenance/types/summary - Get summary by type
• GET /api/maintenance/related-docs/{type} - Get related manual pages (RAG)
• POST /api/maintenance/{id}/documents - Upload receipt/document
• GET /api/maintenance/{id}/documents - List documents for record
• POST /api/maintenance/{id}/photos - Upload photo (before/after/general)
• GET /api/maintenance/{id}/photos - List photos for record
• DELETE /api/maintenance/{id}/photos/{filename} - Delete photo

• POST /api/chat - Send message to AI assistant (with session support)
• DELETE /api/chat/{session_id} - Clear chat session

• POST /api/import/carfax - Import CARFAX PDF
• GET /api/import/service-records - List all service records
• POST /api/import/service-record - Add manual service record
• GET /api/import/kpis - Get maintenance KPIs

• GET /api/reminders - List reminders
• GET /api/reminders/upcoming - Get due/upcoming reminders
• POST /api/reminders - Create reminder
• POST /api/reminders/{id}/complete - Mark complete (handles recurrence)
• DELETE /api/reminders/{id} - Delete reminder

• POST /api/search - Semantic search in documents
• POST /api/search/ask - AI-powered Q&A with RAG

• GET /api/uploads - List uploaded documents
• POST /api/uploads - Upload document (auto-ingests)
• DELETE /api/uploads/{filename} - Delete document

• POST /api/moe/ask - Ask with automatic expert routing
• POST /api/moe/feedback - Submit response feedback
• GET /api/moe/stats - Get performance statistics
• GET /api/moe/experts - List available experts

API Documentation: http://localhost:8000/docs

• vehicles - Vehicle info, VIN, mileage tracking
• maintenance_records - Service history with costs, notes, and photos
• maintenance_logs - CARFAX imports and manual service records
• reminders - Date/mileage-based alerts with recurrence
• document_chunks - Vectorized PDF content (384-dim embeddings)

Uses cosine similarity for semantic search:

SELECT content, 1 - (embedding <=> CAST(:embedding AS vector)) as score
FROM document_chunks
ORDER BY embedding <=> CAST(:embedding AS vector)
LIMIT 5;

• Model: all-MiniLM-L6-v2 (sentence-transformers)
• Dimensions: 384
• No API key required for embeddings

• Docker Model Runner with OpenAI-compatible API (e.g., ai/qwen3-coder, ai/glm-4.7-flash, ai/devstral-small-2)
• Unified llm_client.py abstraction with automatic Redis response caching
• All inference runs locally — no cloud API keys or costs

1. User asks question or clicks "Ask about this"
2. Query embedded using sentence-transformers
3. Similar chunks retrieved from pgvector/Qdrant (includes embedded maintenance records)
4. LLM cache checked for identical prior queries
5. Context + question sent to local LLM via Docker Model Runner
6. Response cached and returned with source thumbnails

• Maintenance Expert - Service intervals, fluid specs, routine maintenance
• Technical Expert - Engine specs, towing capacity, electrical systems
• Safety Expert - Safety features, warnings, recalls, emergencies
• General Assistant - General vehicle questions

• macOS (for Homebrew setup) or manual PostgreSQL installation
• Python 3.11+
• Node.js 18+
• Docker Desktop with Model Runner enabled

# Backend
cd backend
pytest

# Frontend
cd frontend
npm test

# Start all services
docker-compose --profile local-llm up -d

Services:

• Added Docker Model Runner support for local LLM inference (no API key needed)
• Added LLM client abstraction layer (llm_client.py) for local inference
• Added Redis LLM response cache to avoid redundant inference calls
• Added Redis Insight GUI dashboard (port 5540)
• Added permanent Redis caching for vehicle queries and search results
• Added maintenance record embedding into vector store for RAG search
• Added dashboard "Reindex Docs" button with background re-embedding
• Added cache-aware reindexing with automatic stale cache flush
• Added POST /api/maintenance/reindex endpoint
• Maintenance CRUD operations now trigger background re-embedding
• Fixed chat close button visibility at all viewport sizes

• Added floating chat widget with context-aware "Ask about this" integration
• Added rich MaintenanceCard component with related manual page thumbnails
• Added photo upload (before/after/general) for maintenance records
• Added document carousel to timeline view
• Added automatic document ingestion on upload
• Added dual vector storage (PostgreSQL + Qdrant)
• Added Redis session management for chat history
• Updated architecture diagram

• Initial release with maintenance tracking, reminders, and AI search

This is a personal vehicle management project. Feel free to fork and adapt for your own vehicle!

MIT License - see LICENSE for details

Built with Commit-Relay and Git-Steer by Ry-Ops