RAG Knowledge Base System
Hybrid RAG pipeline combining dense vector search (HuggingFace + Pinecone) with BM25 sparse retrieval, Cohere Reranking, and Gemini generation. Plug-and-play retriever/LLM architecture deployed via FastAPI, Docker, and Streamlit.