See:
Another example: AI seems to also struggle parsing the https://telefunc.com/event-based page (when I point it to that page, it doesn't understand that a .telefunc.js file should be named after the file name of a component). I suspect a similar issue where information isn't clearly available when only reading HTML.
This seems to be a quick-win that is relatively easy to implement? (And in general and regardless of whether we go down the llm.txt path, it's always good to have HTML that clearly expose all the information to AIs.)
So far, I think we should do this only to improve AI's crawling of Vike docs via HTML. To improve user's AI experience, we should probably focus more on the following. Once we have this, maybe a simpler solution would be be to tell AI crawlers to crawl .md instead of crawling the HTML (if it's possible?).
See also the Discussion with Claude link in #154 which explains how AI parses HTML.
See:
Another example: AI seems to also struggle parsing the https://telefunc.com/event-based page (when I point it to that page, it doesn't understand that a
.telefunc.jsfile should be named after the file name of a component). I suspect a similar issue where information isn't clearly available when only reading HTML.This seems to be a quick-win that is relatively easy to implement? (And in general and regardless of whether we go down the
llm.txtpath, it's always good to have HTML that clearly expose all the information to AIs.)So far, I think we should do this only to improve AI's crawling of Vike docs via HTML. To improve user's AI experience, we should probably focus more on the following. Once we have this, maybe a simpler solution would be be to tell AI crawlers to crawl
.mdinstead of crawling the HTML (if it's possible?).llms.txtper-page #154See also the
Discussion with Claudelink in #154 which explains how AI parses HTML.