RAG system for querying clinical guidelines with citation-backed, hallucination-mitigated responses. Built with LangGraph, ChromaDB, and FastAPI.
⚠️ Engineering and research project — not medical advice.
PDF Guidelines → Ingestion Pipeline → ChromaDB (bge-small-en-v1.5)
↓
LangGraph Inference Engine
(retrieval → reranking → generation)
↓
FastAPI serving layer → Streamlit UI
| Metric | Before | After | Improvement |
|---|---|---|---|
| End-to-end latency | ~358s | ~7.1s | >98% |
| Retrieval latency | tracked | ||
| Indexing time | tracked | ||
| LLM inference | tracked |
Python · LangChain · LangGraph · ChromaDB · HuggingFace (bge-small-en-v1.5)
Google Gemini 2.5 Flash · FastAPI · Streamlit · Docker
git clone https://github.com/cleberfc23/clinical-evidence-navigator
cd clinical-evidence-navigator
docker compose up