LLM-friendly Docs (`llms.txt`)¶

pyinfra publishes two plain-text artifacts alongside the rendered HTML, following the llmstxt.org convention. They let AI assistants, agents, and static crawlers consume the docs without scraping the site.

What's published¶

llms.txt: a small curated index (~15 KB). Lists every hand-written guide, every operation/fact module, and every connector with a one-line summary and an absolute link into the rendered HTML docs.
llms-full.txt: a single-file monolith (~290 KB). Contains the raw text of the hand-written guides plus the source docstrings and signatures of every built-in operation, fact, and connector.

Both files are regenerated on every docs build and live at the site root of each versioned doc tree:

https://docs.pyinfra.com/en/latest/llms.txt
https://docs.pyinfra.com/en/latest/llms-full.txt

Swap latest for a specific version (for example 3.x or a tag like v3.3.2) to pin against a known release.

How to use them¶

For an AI assistant or code editor¶

Point the tool at llms.txt first. It's small, cheap to keep in context, and links out to whatever section the tool needs. Drop llms-full.txt in only when the task needs deep reference content (writing a custom operation, looking up every argument of files.put, etc.).

As context for a chat session¶

curl -sS https://docs.pyinfra.com/en/latest/llms-full.txt | pbcopy

Paste the result into a new chat. One shot, entire docs.

From code¶

import urllib.request

index = urllib.request.urlopen(
    "https://docs.pyinfra.com/en/latest/llms.txt",
).read().decode()

What's inside¶

llms.txt sections:

Using pyinfra (getting started, CLI, arguments, FAQ)
Deploy reference (operations index, facts index, connectors)
One entry per built-in operation module
One entry per built-in fact module
One entry per built-in connector
How pyinfra works (deploy process, API reference)
Optional pages (install, changelog, compatibility, performance, etc.)

llms-full.txt inlines the hand-written guides verbatim and appends per-module sections pulled directly from source docstrings. Signatures, idempotency notes, and usage examples are all preserved.

Version pinning and freshness¶

The files are rebuilt on every docs deploy, so they always match the HTML on the same version tree. If an LLM caches them, consider re-fetching when pyinfra releases a new version.

LLM-friendly Docs (llms.txt)¶