This repository demonstrates a modular retrieval-augmented generation (RAG) pipeline built with LangChain and evaluated using the Ragas benchmark framework. It uses Mistral AI models for both text generation and embedding creation, with a simple FAISS-based vector store over a local knowledge base.

RAG system for NASA space mission data (Apollo 11, Apollo 13, Challenger). Built with ChromaDB + OpenAI embeddings for semantic search over mission transcripts, flight plans & audio data. Features a conversational LLM client, CLI pipeline, and evaluation suite.

RAG pipeline evaluation framework using RAGAS metrics (Context Precision, Recall, Faithfulness) with GPT-4, LangChain, and pytest

Production-grade RAG system with hybrid retrieval (semantic + BM25), Parent-Child chunking, and RAGAS evaluation. Optimized through 9 controlled experiments — context relevancy improved 315%.

A minimal Retrieval-Augmented Generation (RAG) chat app over the `civictechdc/cib-mango-tree` GitHub repo and a small set of related web pages. It ingests sources into PostgreSQL (with `pgvector`), serves answers via a FastAPI backend, and exposes a Streamlit chat UI.

A local full-stack medical research assistant combining semantic vector search and keyword queries with Ollama Qwen3 to answer clinical literature questions with cited references.