Teach Your LLM to Use Maverick

Teach Your LLM to Use Maverick#

Note

This page is about teaching LLMs to use Maverick. If you are interested in creating LLM-based players, take a look at this example: LLM as Player.

Maverick provides LLM-friendly documentation so that AI tools like Cursor, Windsurf, GitHub Copilot, ChatGPT, and Claude can understand and work with the library. Every page in this documentation has a Markdown equivalent generated by the sphinx-llm extension, following the llms.txt standard.

There are three ways to feed Maverick’s documentation to an LLM, each suited to a different situation.

Option 1: Full documentation bundle (llms-full.txt)#

llms-full.txt is the entire documentation concatenated into a single Markdown file. Use this when:

  • You want to ask broad questions that span multiple parts of the library (e.g. “how does the event system interact with player actions?”).

  • You are working in a chat interface where you paste context manually.

  • You want a self-contained snapshot that doesn’t require the LLM to make additional requests.

The drawback is size — it’s a large file and might consume an unreasonable portion of your models’ context window.

⬇ Download llms-full.txt

Option 2: The index file (llms.txt)#

llms.txt is a lightweight index listing every documentation page with its title and a short description, plus links to the per-page Markdown files. Use this when:

  • You are using an agentic AI tool (Cursor, Windsurf, Copilot Workspace) that can fetch URLs and decide which pages to read.

  • You want the LLM to always reference the latest published documentation rather than a stale downloaded snapshot.

  • Your query is focused enough that the LLM only needs to read a few pages rather than the entire library.

Point the tool at:

https://pymaverick.readthedocs.io/en/latest/llms.txt

The LLM can use the descriptions in the index to identify which pages are relevant, then fetch those individual .html.md files.

Example prompt:

Read https://pymaverick.readthedocs.io/en/latest/llms.txt to get an overview of the
Maverick poker library documentation. Then fetch whichever pages are relevant and help
me write a custom player that folds pre-flop unless it holds a pocket pair.

Note

As of today, these files cannot be natively leveraged by LLM frameworks or IDEs. Alternatively, a MCP server can be implemented to properly parse the llms.txt file.

Option 3: Per-page Markdown#

Every page in this documentation has a Markdown twin at the same URL with .md appended (e.g. api_reference.html.md). Use this when:

  • Your question is about a specific class, concept, or feature and a single page covers it.

  • You want to keep context usage minimal.

  • You are building a retrieval pipeline and want to index individual pages rather than the full bundle.

The easiest way to get the Markdown for the page you are currently reading is the download dropdown in the article header at the top of every page — click the download icon and select .md.