RAGLight

Why we included this project

RAGLight is a practical starting point if you want to stand up a retrieval-augmented generation pipeline without hand-wiring every component. The framework keeps the LLM, the embedding model, and the vector store as separate, swappable pieces, so you can move between OpenAI, Mistral, Ollama, LMStudio, or AWS Bedrock without rewriting your application code. A command-line wizard indexes a folder of documents and drops you into a chat session, and a REST server with a chat UI covers the jump from local prototyping to a Docker Compose deployment. The newer MCP integration lets the pipeline reach external tools and data sources, and hybrid BM25-plus-semantic search, query reformulation, and streaming output handle the rough edges you usually have to build yourself. Teams evaluating a lightweight, provider-agnostic RAG layer will find the configuration-driven approach quick to test against their own documents.