README.md

metadata

title: RAG Onboarding Backend
emoji: 🚀
colorFrom: blue
colorTo: green
sdk: docker
pinned: false
license: mit

RAG Onboarding Backend - HuggingFace Space

Production-ready RAG (Retrieval-Augmented Generation) backend for corporate employee onboarding, deployed on HuggingFace Spaces.

🌟 Features

• FastAPI REST API - High-performance async API
• HuggingFace Models - Open-source LLMs and embeddings
• Vector Search - Qdrant for similarity search
• Caching - Redis for performance optimization
• Monitoring - Prometheus metrics
• Clean Architecture - Production-grade code structure

🚀 Quick Start

API Endpoints

• GET / - Service info
• GET /health - Health check
• POST /api/v1/query - RAG query processing
• GET /api/v1/metrics - Prometheus metrics
• GET /docs - Interactive API documentation

Example Query

curl -X POST "https://YOUR-SPACE-NAME.hf.space/api/v1/query" \
  -H "Content-Type: application/json" \
  -d '{
    "query_text": "What is the onboarding process?",
    "department": "HR",
    "top_k": 5
  }'

🔧 Configuration

Set the following secrets in your HuggingFace Space settings:

• GEMINI_API_KEY - Your Google Gemini API key
• DATABASE_URL - PostgreSQL connection string (use external DB like Supabase/Neon)
• REDIS_URL - Redis connection string (use Upstash Redis)
• QDRANT_URL - Qdrant vector DB URL (use Qdrant Cloud)
• QDRANT_API_KEY - Qdrant API key (if using cloud)

📊 Models Used

• LLM: Google Gemini 2.0 Flash (via API)
• Embeddings: sentence-transformers/all-MiniLM-L6-v2
• Reranking: cross-encoder/ms-marco-MiniLM-L-12-v2 (optional)

🏗️ Architecture

┌─────────────┐
│   Client    │
└──────┬──────┘
       │
       ▼
┌─────────────┐
│  FastAPI    │
└──────┬──────┘
       │
       ├─────► PostgreSQL (Documents)
       ├─────► Redis (Cache)
       ├─────► Qdrant (Vectors)
       └─────► Google Gemini (LLM)

📝 License

MIT License - See LICENSE file for details

🤝 Contributing

Contributions welcome! Please open an issue or submit a PR.

📧 Support

For questions or issues, please open a GitHub issue.