> As an anecdote, I had an easier time using Cursor + Claude to build the app in FastAPI and Next.js, and a harder time with FastHTML and SvelteKit. Since FastHTML is barely a couple weeks old (at the time of writing), its code and docs likely hasn’t made its way into the training data of most LLMs yet, explaining their limited proficiency with FastHTML.
In cursor settings, click features and then add the documentation URL for each framework or library you are using so they can be indexed.
It would be best if you did this regardless of how well trained a model is on certain code - it helps immensely.
FastHTML has markdown formatted docs which can be used by Claude, just add .md to the end of the URL:
You can find markdown docs for most libraries on GiHub, where you can have Cursor index.
I suspect that with the increased use of LLM-aware code editors, single-page markdown-formatted documentation will become more common (even better would be if Cursor hosted an external vector db with up-to-date docs and tutorials for all the most popular libraries and frameworks).
I'm the maintainer of an open Source CLI. The documentation site [1] is a static HTML site generated by Jekyll, from a bunch of Markdown files. Is there advices for LLM ready single page doc?
- should I just aggregate the Markdown files and that's all?
In cursor settings, click features and then add the documentation URL for each framework or library you are using so they can be indexed.
It would be best if you did this regardless of how well trained a model is on certain code - it helps immensely.
FastHTML has markdown formatted docs which can be used by Claude, just add .md to the end of the URL:
https://docs.fastht.ml/ref/handlers.html.md
You can find markdown docs for most libraries on GiHub, where you can have Cursor index.
I suspect that with the increased use of LLM-aware code editors, single-page markdown-formatted documentation will become more common (even better would be if Cursor hosted an external vector db with up-to-date docs and tutorials for all the most popular libraries and frameworks).