Skip to main content

Revamping LLM Context Files and Adding Per-Page Markdown Exports

· 2 min read

We've improved llms.txt and llms-full.txt for developer portals and added per-page Markdown exports. These updates make the documentation easier for AI tools to read and help them return answers that are more accurate, relevant, and grounded in the same content users see in the portal.

What's New?​

Easier navigation for AI tools​

llms-full.txt is now easier for AI tools to follow, so they can move through the documentation in a way that better matches how users explore the portal.

Clearer page-level documentation​

Each portal page can now be available as its own Markdown document. This gives AI tools a cleaner page-level source to read from, which helps them find the right topic faster and answer with better context.

llms.txt now follows the portal structure more closely and points to page-level Markdown documents instead of raw portal URLs.

Before:

### Models

- [Order](https://www.example.com/docs/#/http/x-redirect/JTI0bSUyRk9yZGVy) 
- [Category](https://www.example.com/docs/#/http/x-redirect/JTI0bSUyRkNhdGVnb3J5) 
- [User](https://www.example.com/docs/#/http/x-redirect/JTI0bSUyRlVzZXI) 
- [Tag](https://www.example.com/docs/#/http/x-redirect/JTI0bSUyRlRhZw)

After:

## Models
- Structures
  - [Order](https://www.example.com/docs/llms-pages/http/models/structures/order.md)
  - [Category](https://www.example.com/docs/llms-pages/http/models/structures/category.md)
  - [User](https://www.example.com/docs/llms-pages/http/models/structures/user.md)
  - [Tag](https://www.example.com/docs/llms-pages/http/models/structures/tag.md)

These Markdown documents are organized in a folder structure that matches the portal, and each one includes a Source link back to the corresponding portal page.

For example, a generated page-level Markdown file can include:

# Order

Source: https://www.example.com/docs/#/http/x-redirect/JTI0bSUyRk9yZGVy

More complete answers from portal content​

llms-full.txt now gives AI tools access to more of the information users expect to find in the portal.

This includes better coverage for:

  • Authentication details for endpoints that require auth.
  • Success response descriptions.
  • Request parameters, errors, usage examples, and server information.
  • Model descriptions.
  • Initialization examples for OneOf and AnyOf schemas when available.
  • Custom guide content and richer rendered structures such as tabs, callouts, code blocks, and linked resources.

Overall, these tools can now use a fuller and more helpful view of the documentation, which should lead to better answers for developers.

Why This Matters​

Previously, AI tools could miss useful page details or pick up less helpful summaries when working from generated documentation.

With this update:

  • They can navigate the docs more effectively.
  • Page summaries are clearer and less noisy.
  • Page-level Markdown gives AI workflows a cleaner input source.
  • More portal content is available to support better answers.

To learn more, see the LLM context generation documentation.