Minima is an open-source RAG platform offering on-premises containerized solutions, integrating with ChatGPT and MCP, or operating fully locally. It enables secure querying of local documents using various LLMs.
Minima is an open-source RAG system deployable on-premises via containers, integrating with ChatGPT and MCP, or functioning as a fully local RAG. It supports three modes: isolated on-premises operation, ChatGPT integration, and Anthropic Claude integration.
For containerized deployment, a .env file is required, specifying variables like LOCAL_FILES_PATH, EMBEDDING_MODEL_ID, EMBEDDING_SIZE, OLLAMA_MODEL, RERANKER_MODEL, USER_ID, and PASSWORD. Different docker-compose files are provided for each mode: Ollama, ChatGPT, and MCP. ChatGPT integration requires copying an OTP from the terminal and using the Minima GPT. Claude integration involves modifying the claude_desktop_config.json file. A fully local installation requires navigating to the electron directory, running npm install and npm start.
Minima can also be installed via Smithery for Claude Desktop using the command npx -y @smithery/cli install minima --client claude. Ensure Python >=3.10 and 'uv' are installed for MCP usage. Minima is licensed under the Mozilla Public License v2.0 (MPLv2).
dmayboroda/minima
November 2, 2024
March 28, 2025
Python