DeepDocs: Update your docs with AI on every pull request (Install)
DeepDocs is a LLM-powered app that keeps your GitHub documentation—like READMEs, API references, SDK guides, wikis, tutorials etc. up-to-date with your changing codebase, eliminating manual effort.
How it works (▶️ Watch)
- In your main repo, specify the path to your doc files.
- When a PR is opened, DeepDocs automatically updates the relevant docs based on code changes.
- Continuous Documentation: Update your docs with every PR with no manual effort
- Intelligent Updates: Preserves your existing doc format without rewriting from scratch
- Broad support: Works across mono and external repos
- Github native: Integrates smoothly into your dev workflow and with other doc hosting tools like Mintlify
Install it in your source code repo. If docs are in a separate repo, install it there too.
If DeepDocs isn't active on your repos yet, go to any repo's Settings > GitHub Apps, click Configure next to DeepDocsAI, select your source and docs repos (separately if needed), and hit Save.
In the main branch, place a deepdocs.yml file at the root of your source code repo. This file should specify the GitHub URLs of the doc files to be updated — all URLs must point to files that exist on the main branch. You've two options for it:
Option 1: A single repo or sub-folder Use this when you want DeepDocs to decide which docs to update within a folder or entire repo.
target_repo:
- url: "https://github.com/user/repo/tree/main/docs"In Option 1, committing deepdocs.yml triggers DeepDocs to sync your code and docs, and comment on the commit when done. You’re all set for step 3.
Option 2: Up-to 5 files in the source code or external repo Use this when you want to update specific doc files only.
target_files:
- url: "https://github.com/user/repo/blob/main/path/to/doc.md"Make a code change and open a pull request. DeepDocs will update the relevant docs in a new commit and add it to the same PR. If your docs are in a separate repo, it will open a PR there instead.
Each doc update consumes 1 credit, regardless of the number of code changes. If no update is needed, 0.5 credits are used for analysis.