dibbed/tenant-rag ? reverse-engineered prompt

Build me a production ready multi tenant RAG backend for a SaaS app.

I want a FastAPI service where each customer has their own isolated data, their own API key, and their own vector store, so one tenant can never see another tenant’s documents or answers. It should let me create tenants, issue keys, upload text or files, and ask questions with grounded answers that include citations and a confidence score. Please make the whole thing work end to end with a simple local setup, a clear config file, and an API I can call from curl or another backend.

It should support document ingestion, semantic caching per tenant, and a choice of local or hosted LLM providers. Add a few basic admin or CLI commands for provisioning tenants and keys, and make sure there are tests and docs so I can run it and verify it easily. If you need to check current library docs online, go ahead.