llms.txt — AI discovery for Turtini and your modules

Turtini publishes its public help library at https://turtini.com/llms.txt and https://turtini.com/llms-full.txt — the emerging convention from llmstxt.org that AI clients (Claude, Gemini, Cursor, etc.) prefer over crawling a normal docs site.

What's there:
• /llms.txt — short manifest with the platform description and a categorized index linking every published help article
• /llms-full.txt — every article inlined as one big markdown document, optimized for single-shot ingestion into a model context
• /api/help/articles — JSON list (paginated, supports ?category and ?module filters)
• /api/help/articles/<id> — single article as JSON

Why it exists:
When you ask Claude "how do I set up Turtini's POS for my food truck?" you want Claude to give you the right answer based on Turtini's actual docs, not training-data hallucinations. Publishing llms.txt is the lowest-friction way to give every AI a path to the truth.

Filtering:
Only articles with status='published' and audience='public' appear in the public files. Admin-only and gov-only articles are never served from this surface.

Caching:
The function caches results in-memory for 5 minutes. Edge cache headers add another 5 minutes. So a help-article edit propagates to the public file within ~10 minutes.

Plugin manifests:
For ChatGPT custom GPTs and other tools that prefer a structured manifest, https://turtini.com/.well-known/ai-plugin.json points at the OpenAPI spec with the right discovery shape.

For module developers:
If you publish a marketplace module, your help articles are automatically added to the platform-wide llms.txt once they're approved and have audience='public'. No extra work — write good articles, get free AI distribution.