Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/AsyncFuncAI/deepwiki-open/llms.txt

Use this file to discover all available pages before exploring further.

DeepWiki turns any GitHub, GitLab, Bitbucket, or local repository into a navigable wiki by combining repository analysis, vector embeddings, and AI generation. The result is a structured set of documentation pages — each with prose explanations and Mermaid diagrams — organized into sections that mirror the logical structure of the codebase.
DeepWiki supports public and private repositories on GitHub, GitLab, and Bitbucket, as well as local file system paths. For private repositories, see Access private repositories.

​
How the process works

DeepWiki performs seven steps for every generation request:
  1. Clone and analyze the repository’s file tree and code relationships
  2. Embed the codebase into a vector index for retrieval
  3. Select the AI provider and model you configured
  4. Generate prose documentation using context-aware retrieval
  5. Produce Mermaid diagrams that illustrate architecture and data flow
  6. Organize pages into a structured wiki with sections
  7. Cache the result so subsequent visits load instantly

​
Generating a wiki

1

Enter the repository URL

Open DeepWiki at http://localhost:3000 and paste a repository URL into the input field. Accepted formats include full HTTPS URLs (e.g. https://github.com/openai/codex) and local directory paths.
2

Select a provider and model

DeepWiki supports Google Gemini (default: gemini-2.5-flash), OpenAI (default: gpt-5-nano), OpenRouter, Azure OpenAI, and local Ollama models. Open the model selector to choose a different provider or enter a custom model identifier.
3

Configure filters (optional)

Expand the filter options to control which parts of the repository are indexed:
  • Excluded dirs — directories to skip entirely (e.g. node_modules, dist)
  • Excluded files — file patterns to omit (e.g. *.lock)
  • Included dirs — restrict indexing to specific directories only
4

Add an access token for private repos (optional)

Click + Add access tokens and enter a personal access token if the repository is private. The token is passed per-request and is not stored server-side.
5

Click Generate Wiki

DeepWiki clones the repository, builds embeddings, and streams generated pages into the wiki view. Large repositories may take a minute or two.

​
What a generated wiki contains

Each generated wiki includes:
  • Pages — AI-written documentation for individual modules, packages, or concepts found in the repository
  • Sections — logical groups that organize pages into a navigable hierarchy
  • Mermaid diagrams — automatically generated flowcharts and sequence diagrams embedded in relevant pages
The WikiStructureModel returned by the API has the following shape: 
{
  "id": "string",
  "title": "string",
  "description": "string",
  "pages": [...],
  "sections": [...],
  "rootSections": [...]
}
Each page (WikiPage) carries an id, title, content, a list of filePaths the content was derived from, an importance level (high, medium, or low), and a list of relatedPages IDs.

​
Wiki caching and refresh

Generated wikis are cached on disk under ~/.adalflow/wikicache/. When you navigate to a repository URL that has already been processed, DeepWiki loads the cached version immediately. To regenerate the wiki with fresh content — for example after the repository has been updated — use the refresh option in the UI. Cached data persists across server restarts; Docker setups mount ~/.adalflow to the host so caches survive container recreation.

​
Language selection

DeepWiki can generate wiki content in multiple languages. A language selector in the UI lets you choose the output language before generation. Supported language codes include en (English), zh (Simplified Chinese), zh-tw (Traditional Chinese), ja (Japanese), es (Spanish), kr (Korean), vi (Vietnamese), pt-br (Brazilian Portuguese), fr (French), and ru (Russian).

Build docs developers (and LLMs) love

Get started for freeTalk to us