DocsGPT — AI-Powered Documentation Assistant

Introduction

DocsGPT is an open-source documentation assistant that uses retrieval-augmented generation to answer questions about your project docs. It ingests Markdown, RST, PDF, and other formats, indexes them into a vector store, and provides a chat interface where users get accurate, source-cited answers.

What DocsGPT Does

• Ingests documentation from URLs, files, or Git repositories automatically
• Splits and embeds content into a vector database for semantic search
• Answers user questions with cited references to specific documentation pages
• Provides an embeddable widget for adding AI chat to any documentation site
• Supports multiple LLM providers including OpenAI, local models, and Anthropic

Architecture Overview

The backend is a Python Flask application that orchestrates document ingestion, embedding, and retrieval. Documents are chunked and stored in a vector database (FAISS or Elasticsearch). When a query arrives, relevant chunks are retrieved and passed to the configured LLM along with the question. The React frontend displays the conversation with source links.

Self-Hosting & Configuration

• Deploy with Docker Compose for a single-command setup including all services
• Configure the LLM provider and API keys in the .env file
• Choose between cloud LLMs or local models via llama.cpp or Ollama integration
• Set up document sources through the admin panel or API endpoints
• Use the JavaScript widget snippet to embed the chat on external sites

Key Features

• Source-grounded answers that cite the specific documentation page used
• Embeddable chat widget for adding AI help to any docs site
• Multi-format ingestion including Markdown, RST, PDF, DOCX, and web pages
• API endpoints for programmatic access and integration into CI/CD pipelines
• Admin dashboard for managing document sources and monitoring usage

Comparison with Similar Tools

• Mintlify — commercial docs platform with built-in AI; DocsGPT is self-hosted and open source
• GitBook AI — proprietary AI search for GitBook docs; DocsGPT works with any doc format
• Quivr — general-purpose RAG platform; DocsGPT specializes in documentation workflows
• Danswer — enterprise knowledge search; DocsGPT focuses specifically on developer documentation

FAQ

Q: Which LLMs does it support? A: OpenAI GPT models, Anthropic Claude, local models via Ollama, and any OpenAI-compatible API endpoint.

Q: Can I use it with private documentation? A: Yes. Self-hosted deployment keeps all data on your infrastructure with no external calls when using local models.

Q: How do I update the knowledge base when docs change? A: Re-run the ingestion process or configure automatic re-indexing through the admin panel.

Q: Is there a hosted version? A: The project offers a cloud-hosted option at docsgpt.co, but self-hosting is fully supported.

Sources

• https://github.com/arc53/DocsGPT
• https://docs.docsgpt.co