minima

Why we included this project

Minima is a self-contained retrieval stack that keeps your documents on your own infrastructure. It runs as Docker containers that bundle the embedding model, reranker, and LLM, so you can go fully local with Ollama or point the indexer at an OpenAI-compatible endpoint like vLLM or TGI. The interesting part is how you query the index: the bundled chat UI works, but you can also connect ChatGPT custom GPTs or Anthropic Claude to answer from your private files, and MCP support opens the same knowledge base to other MCP-aware tools. For teams that want question answering over their own documents without sending them to a third party, this covers several deployment shapes in one project.