Kotaemon

Cinnamon's open source app gives you a web interface for asking questions about PDFs and other files, plus a Python framework for building the retrieval pipeline behind it. By default it indexes documents into a vector store and a full-text index, merges both at query time with reranking, and answers with citations that open in a built-in PDF viewer with matched passages highlighted. It offers multi-user login with private and shared collections, question decomposition and ReAct or ReWOO agents for multi-hop questions, and optional GraphRAG, NanoGraphRAG or LightRAG indexing. Models can be hosted APIs such as OpenAI, Azure OpenAI, Cohere or Groq, or local ones served by Ollama or llama-cpp-python, so fully offline use is possible. Parsing can be upgraded with Docling locally or Azure and Adobe cloud services for tables, figures and OCR, and storage options include Chroma, LanceDB, Milvus, Qdrant and Elasticsearch. The quickest start is a lite, full or Ollama-bundled Docker image on port 7860; a manual install needs Python 3.10+. Apache-2.0 licensed. Ideal for teams wanting a private, extensible document assistant.