Indexed:261 news/91 AI tools/372 open-source projects/64 learning resources
HomeNewsOpen SourceModelsAI ToolsLearn
Home / Open Source / kotaemon

kotaemon

An open-source document intelligent Q&A tool based on RAG technology, supporting conversational interaction with your documents.

PythonApache-2.0Search & RAG
⭐ GitHubhttps://github.com/Cinnamon/kotaemon
25,655
Stars
+183
Star growth
Jul 15, 2026
Last updated
2,212
Clicks

Kotaemon - Open Source RAG Document Intelligent Question Answering Tool

Project Overview

Kotaemon is an open-source RAG (Retrieval-Augmented Generation) tool specifically designed for conversational interaction with documents. The project aims to provide both end-users and developers with a fully functional RAG user interface, meeting users' needs for document question answering and assisting developers in building their own RAG pipelines.

Core Features

End-User Features

  • Clean and Beautiful UI: Provides a user-friendly RAG question answering interface.
  • Multiple LLM Support: Compatible with various LLM API providers (OpenAI, AzureOpenAI, Cohere, etc.) and local LLMs (via ollama and llama-cpp-python).
  • Easy Installation: Offers simple scripts for quick startup.
  • Multi-User Support: Supports multi-user login, allowing the creation of private/public document collections, supporting collaboration and sharing.

Developer Features

  • RAG Pipeline Framework: Provides tools for building RAG document question answering pipelines.
  • Customizable UI: Built on Gradio, allowing visualization of the actual effects of the RAG pipeline.
  • Hybrid RAG Pipeline: Employs hybrid (full-text and vector) retrievers and re-ranking to ensure optimal retrieval quality.
  • Extensibility: Supports custom UI elements and various document indexing and retrieval strategies.

Advanced Features

  • Multi-Modal Question Answering Support: Supports multi-document question answering involving charts and tables, and supports multi-modal document parsing.
  • Advanced Citation Functionality: Provides detailed citation information to ensure the correctness of LLM answers, with direct viewing of highlighted citations in the browser's PDF viewer.
  • Complex Reasoning Methods: Supports question decomposition to answer complex/multi-hop questions, and supports Agent-based reasoning such as ReAct, ReWOO, etc.
  • Configurable Settings Interface: Allows adjustment of key aspects of the retrieval and generation process (including prompts) on the UI.

Technical Architecture

Supported Document Formats

  • PDF, HTML, MHTML, XLSX
  • Supports more formats (.doc, .docx, etc.) through the Unstructured library.

Data Storage Options

  • Document Storage: Elasticsearch, LanceDB, SimpleFileDocumentStore
  • Vector Storage: ChromaDB, LanceDB, InMemory, Milvus, Qdrant

LLM Model Support

  • API Providers: OpenAI, Azure OpenAI, Cohere, Groq, etc.
  • Local Models: Supports GGUF format models via Ollama and llama-cpp-python.

GraphRAG Integration

The project supports two GraphRAG implementations:

NanoGraphRAG (Recommended)

  • More direct Kotaemon integration.
  • Automatically identifies default LLM and embedding models.

MS GraphRAG

  • Official Microsoft GraphRAG implementation.
  • Only supports OpenAI or Ollama API.
  • Supports local models and custom settings.

Installation and Deployment

Docker Deployment (Recommended)

Provides slim and full Docker images:

  • Slim Version: Supports basic file types, smaller image size.
  • Full Version: Includes the unstructured package, supports more file types.

Local Installation

  1. Download the latest version of the release file.
  2. Configure the .env file.
  3. Run python app.py to start the service.
  4. The default username and password are admin.

Customization and Extension

Custom Reasoning Pipeline

  • New .py implementations can be added in libs/ktem/ktem/reasoning/.
  • Enable new reasoning pipelines through flowsettings.

Custom Indexing Pipeline

  • Refer to the example implementation in libs/ktem/ktem/index/file/graph.
  • Supports GraphRAG indexing pipelines.

Configuration Files

  • flowsettings.py: Application configuration.
  • .env: Model and credential configuration.

Application Scenarios

  1. Enterprise Knowledge Management: Building internal document question answering systems.
  2. Academic Research: Intelligent question answering on research literature.
  3. Legal Document Analysis: Quickly retrieving and analyzing legal documents.
  4. Technical Documentation Support: Providing intelligent query for technical documentation.
  5. Personal Knowledge Base: Managing and querying personal document collections.

Project Advantages

  • Open Source and Free: Completely open source, free to use and modify.
  • Fully Functional: Comprehensive functionality, from simple question answering to complex reasoning.
  • Easy to Use: Intuitive web interface, usable without technical background.
  • Highly Customizable: Supports multiple configuration and extension options.
  • Active Community: Continuous updates and community support.

Technical Features

  • Developed based on Python.
  • Uses Gradio to build the web interface.
  • Supports Docker containerized deployment.
  • Modular architecture design.
  • Supports GPU acceleration (local models).

Summary

Kotaemon is a powerful RAG UI tool that allows users to interact with documents conversationally. It not only provides end-users with a simple and easy-to-use document question answering interface but also provides developers with a complete framework for building custom RAG systems. With features such as hybrid retrieval, multi-modal support, and advanced citation, Kotaemon provides a comprehensive and professional solution in the field of document intelligent question answering.