ragmill
PyPIv0.4.0MITPython >= 3.9
7
A lightweight, zero-config local pipeline engine for AI data ingestion, semantic chunking, embeddings and vector search — with optional cloud backends and a REST API.
terminal
pip install ragmill- ▸Offline-first, no API keysLocal ONNX embeddings plus a local Qwen2.5-1.5B LLM via llama-cpp-python — the whole pipeline runs with zero network calls.
- ▸Full RAG loopingest → chunk → embed → store → search → grounded, cited chat. Reads txt, md, log, rst, pdf and docx.
- ▸Pluggable vector storesSQLite by default; Pinecone or Qdrant by setting env vars, no code change. Backends migrate via JSONL export/import.
- ▸Swappable chat backendsLocal, Gemini or OpenAI, selected at runtime with RAGMILL_CHAT_BACKEND.
- ▸Incremental folder syncRe-ingests only what changed instead of rebuilding the whole index.
- ▸Zero-dependency coreThe base install pulls nothing; every heavy capability is an optional extra. Docker Compose profiles ship for the server modes.
interfaces
Python APIragmill CLIFastAPI REST server (with browser chatbox)Standalone setup UI that writes .env
releases · latest 2026-07-23
0.2.00.3.00.3.10.3.20.4.0
ragllm-ingestionsemantic-chunkingvector-embeddingsvector-searchdata-pipelineonnx