langgraph/docs

LangGraph Documentation

For more information on contributing to our documentation, see the Contributing Guide.

Structure

The primary documentation is located in the docs/ directory. This directory contains both the source files for the main documentation as well as the API reference doc build process.

Main Documentation

Main documentation files are located in docs/docs/ and are written in Markdown format. The site uses MkDocs with the Material theme and includes:

• Concepts: Core LangGraph concepts and explanations
• Tutorials: Step-by-step learning guides
• How-tos: Task-focused guides for specific use cases
• Examples: Real-world applications and use cases
• Jupyter Notebooks: Interactive tutorials that are automatically converted to markdown

API Reference

API reference documentation is located in docs/docs/reference/ and is generated from docstrings in the codebase using the mkdocstrings plugin.

The API reference uses manual directives in markdown files that specify which classes/functions to document:

::: langgraph.graph.state.StateGraph
    options:
      members:
        - add_node
        - compile

Build Process

The build process follows these steps:

1. Content Processing Phase:

   • _scripts/notebook_hooks.py - Main processing pipeline that:
     • Converts Jupyter notebooks to markdown using notebook_convert.py
     • Adds automatic API reference links to code blocks using generate_api_reference_links.py
     • Handles conditional rendering for Python/JS versions
     • Processes highlight comments and custom syntax

2. API Reference Generation:

   • mkdocstrings plugin extracts docstrings from Python source code
   • Manual ::: module.Class directives in reference pages specify what to document
   • Cross-references are automatically generated between docs and API

3. Site Generation:

   • MkDocs processes all markdown files and generates static HTML
   • Custom hooks handle redirects and inject additional functionality

4. Deployment:

   • Site is deployed with Vercel
   • make build-docs generates production build (also usable for local testing)
   • Automatic redirects handle URL changes between versions

Local Development

For local development, use the Makefile targets:

# Serve docs locally with hot reloading
make serve-docs

# Clean build for production testing
make build-docs

# Serve with clean build
make serve-clean-docs

The serve-docs command:

• Watches source files for changes
• Includes dirty builds for faster iteration
• Serves on http://127.0.0.1:8000/langgraph/

Documentation Standards

Docstring Format: The API reference uses Google-style docstrings with Markdown markup. The mkdocstrings plugin processes these to generate documentation.

Required format:

def example_function(param1: str, param2: int = 5) -> bool:
    """Brief description of the function.

    Longer description can go here. Use Markdown syntax for
    rich formatting like **bold** and *italic*.

    Args:
        param1: Description of the first parameter.
        param2: Description of the second parameter with default value.

    Returns:
        Description of the return value.

    Raises:
        ValueError: When param1 is empty.
        TypeError: When param2 is not an integer.

    !!! warning
        This function is experimental and may change.

    !!! version-added "Added in version 0.2.0"
    """

Special Markers:

• MkDocs admonitions: !!! warning, !!! note, !!! version-added
• Code blocks: Standard markdown ``` syntax
• Cross-references: Automatic linking via generate_api_reference_links.py

Site Styling and Assets

Theme and Styling:

• Uses Material for MkDocs theme
• Custom CSS in docs/stylesheets/ for LangGraph-specific styling:
  • Brand colors and typography
  • Custom navigation components
  • Version admonitions and agent graph widgets

Static Assets:

• Logos and favicon in docs/static/
• Custom stylesheets in docs/stylesheets/

Content Processing:

• Automatic API reference link generation for code examples
• Jupyter notebook execution with VCR cassettes for reproducible builds
• Conditional rendering for multi-language support
• Extensive redirect mapping for URL stability

Analytics and Integration:

• Google Tag Manager integration via custom hooks
• GitHub integration (edit buttons, source links)
• Automatic cross-referencing between documentation sections

Execute notebooks

If you would like to automatically execute all of the notebooks, to mimic the "Run notebooks" GitHub action, you can run:

python _scripts/prepare_notebooks_for_ci.py
./_scripts/execute_notebooks.sh

Note: if you want to run the notebooks without %pip install cells, you can run:

python _scripts/prepare_notebooks_for_ci.py --comment-install-cells
./_scripts/execute_notebooks.sh

prepare_notebooks_for_ci.py script will add VCR cassette context manager for each cell in the notebook, so that:

• when the notebook is run for the first time, cells with network requests will be recorded to a VCR cassette file
• when the notebook is run subsequently, the cells with network requests will be replayed from the cassettes

Adding new notebooks

If you are adding a notebook with API requests, it's recommended to record network requests so that they can be subsequently replayed. If this is not done, the notebook runner will make API requests every time the notebook is run, which can be costly and slow.

To record network requests, please make sure to first run prepare_notebooks_for_ci.py script.

Then, run

jupyter execute <path_to_notebook>

Once the notebook is executed, you should see the new VCR cassettes recorded in cassettes directory and discard the updated notebook.

Updating existing notebooks

If you are updating an existing notebook, please make sure to remove any existing cassettes for the notebook in cassettes directory (each cassette is prefixed with the notebook name), and then run the steps from the "Adding new notebooks" section above.

To delete cassettes for a notebook, you can run:

rm cassettes/<notebook_name>*