llms-txt-actions

Make your documentation LLM friendly.

This Github action/CLI tool can automatically generate markdown (.md) files for each endpoint in your documentation as per the standard proposed by answer.ai, for more details read: https://llmstxt.org/.

It completely runs locally be default, by passing MODEL_API_KEY one can choose to engage hosted cloud AI services to generate page summaries.

Features

• 📄 Content Processing: Generate LLM-ready markdown (.md) files from popular document frameworks such as Readthedocs, MKDocs, Sphinx and more.
• 🌈 All Formats : Can process HTML, PDF, Images, DOCX, PPTX, XLSX (thanks to docling) and convert them to Markdown. (coming soon)
• 0️⃣ Zero Config: Works out of the box for most file based documentation framework.
• 💾 Generate Summaries: Using LLMs, we generate concise summary of each page.
• 📕 BYO-Model: Thanks to litellm, you can use upto 150 models like OpenAI, VertexAI, Cohere, Anthropic.

Quick Start

There are two ways to access this library.

1. Add this to your GitHub workflow:

    steps:
      - name: Make docs LLM ready
        uses: demodrive-ai/llms-txt-action@v1

  # OR You can choose to use an AI model to generate summaries, its completely optional.
      steps:
      - name: Make docs LLM ready
        env:
          MODEL_API_KEY: ${{ secrets.MODEL_API_KEY }}
        uses: demodrive-ai/llms-txt-action@v1
        # any other inputs you would like to set.

OR

You can use it outside of Github Action.

# python 3.9 or above
pip install llms-txt-action

llms-txt --docs-dir site/
# The first run takes a while as it downloads models files from the intrnet.

Input Parameters

Secret Parameters

Local Development

1. Clone and install:

   # clone the repo
   uv sync
   

2. Run the crawler:

   uv run python -m "llms_txt_action.entrypoint" --docs-dir site/
   

Examples

ReadtheDocs

To integrate llms-txt-action with ReadtheDocs, you'll need to configure two files in your project:

1. .readthedocs.yaml - The main ReadtheDocs configuration file that defines the build environment and process
2. docs/requirements.txt - Python package dependencies needed for building your documentation

Here's how to set up both files:

# .readthedocs.yaml
version: 2

build:
  os: ubuntu-22.04
  tools:
    python: "3.12" # ^3.9 is supported.
  jobs:
    post_build:
      - llms-txt --docs-dir $READTHEDOCS_OUTPUT/html

mkdocs:
  configuration: mkdocs.yml

python:
  install:
  - requirements: docs/requirements.txt

# docs/requirements.txt
llms-txt-action

MkDocs + Github Pages

MkDocs is a fast and simple static site generator that's geared towards building project documentation. Here's how to integrate llms-txt-action with MkDocs when deploying to GitHub Pages:

1. First, ensure you have a working MkDocs setup with your documentation source files.

2. Create or update your GitHub Actions workflow file (e.g., .github/workflows/docs.yml) with the following configuration:

# github action - .github/workflows/docs.yml

      - name: Generate static files
        run : mkdocs build

      - name: Make docs LLM ready
        uses: demodrive-ai/llms-txt-action@v1

      - name: Deploy to Github
        run : mkdocs gh-deploy --dirty
        # --dirty helps keep the generated .md and .txt files from getting deleted.

Sphinx + Github Pages

Sphinx is a popular documentation generator for Python projects. Here's how to integrate llms-txt-action with Sphinx and GitHub Pages:

1. First, ensure you have a working Sphinx documentation setup with a docs/ directory containing your source files and configuration.

2. Create or update your GitHub Actions workflow file (e.g., .github/workflows/docs.yml) with the following configuration:

#...
#...
      - name: Build HTML
        uses: ammaraskar/sphinx-action@master
      - name: Make docs LLM Ready
        uses: demodrive-ai/llms-txt-action@v1
        with:
          name: docs-dir
          path: docs/build/html/
      - name: Deploy
        uses: peaceiris/actions-gh-pages@v3
        with:
          github_token: ${{ secrets.GITHUB_TOKEN }}
          publish_dir: docs/build/html
#...
#...

License

This project is licensed under the MIT License - see the LICENSE file for details.