deepwiki-open

DeepWiki-Open is a self-hosted, open-source tool that automatically generates a structured wiki (documentation site) for any GitHub, GitLab, or Bitbucket repository. You point it at a repo, and it clones the code, analyzes the structure, generates written documentation, creates visual diagrams showing how components relate to each other, and organizes everything into an easy-to-navigate wiki — all powered by AI.

The process works by creating embeddings (a way of representing text and code as numbers that capture meaning) of the repository's contents, then using those embeddings to provide the AI model with relevant context when generating documentation. This technique is called RAG (Retrieval-Augmented Generation) — it lets the AI accurately describe the code by looking up the relevant parts rather than trying to hold everything in memory at once. Visual diagrams are generated using Mermaid, a text-based diagram format.

It supports multiple AI model providers so you can choose what powers the generation: Google Gemini, OpenAI, OpenRouter (a gateway to many models), local Ollama models (meaning you can run it completely offline without sending code to any external service), and Azure OpenAI. Private repositories are supported via personal access tokens.

The stack is Python (FastAPI) for the backend and a JavaScript frontend. You run it locally via Docker or by starting the backend and frontend separately. You would use this if you want to quickly understand an unfamiliar codebase or create documentation for your own project. The full README is longer than what was provided.