📚 Production RAG System

Ask questions over your own documents using hybrid AI search.

---

📸 Screenshots

RAG Chat Interface

Monitoring Dashboard

---

✨ Features

• Hybrid retrieval — BM25 keyword search + vector semantic search
• Cross-encoder reranking for higher precision
• Citation enforcement — declines if context doesn't support the answer
• LangSmith tracing — every query tracked end-to-end
• Monitoring dashboard — latency, quality scores, declined queries
• Chat UI built with Streamlit

---

🛠️ Tech Stack

• LLM: OpenAI GPT-4o-mini
• Embeddings: text-embedding-3-small
• Vector DB: ChromaDB
• Keyword Search: BM25
• Reranker: ms-marco-MiniLM-L-6-v2
• Tracing: LangSmith
• UI: Streamlit

---

🚀 How to Run

1. Add your documents to ./docs/
2. Create .env file with your API keys
3. Run python ingest.py
4. Run streamlit run app.py
5. Run streamlit run dashboard.py --server.port 8502 for monitoring

---