danny-avila/rag_api ? reverse-engineered prompt

Build me a FastAPI service for ID based RAG with PostgreSQL and pgvector.

I want an API where people can upload documents, store them as embeddings, and then query them later by file id. It should support adding, retrieving, searching, and deleting documents, and it should work asynchronously so it feels fast under load. Please make the retrieval safe so users can only see or delete what belongs to them, and if a file id is outside their scope it should just look like it was never there. Use JWT based caller identity, and let an entity id be used for agent owned knowledge bases when needed.

Use LangChain for the vector and document handling, and make it easy to run with Docker as well as locally. If you need to check current docs or best practices for FastAPI or pgvector, look them up online. Keep the setup simple, with clear environment variables and a clean README style starting point.