Perhaps somebody e.g. looking for a cool student project could consider how to transform and publish e.g. the openEHR archetypes, templates etc. from CKM and the openEHR specifications into chunks of suitable size and structure optimized for LLMs? And then check if the LLMs perform better.
(If we do want our future AI overlords to understand our EHRs properly… what could possibly go wrong…)
It would also be useful to be able to use this discourse forum conversations as context for an LLM. currently the scraping of content here is not allowed, but there is higher level knowledge that could be useful for a clinical modeling helper agent
I agree that the forum is a great source, but not sure I understand what you mean by “scraping of content here is not allowed”, could you explain more? The https://discourse.openehr.org/robots.txt file seems to allow most agents/robots to scrape posts etc. There is no llms.txt here yet, is that what you mean?
There is now! I found an LLMs.TXT plugin for Discourse which creates llms-full.txt and sitemaps. For some reason llms.txt is not working yet but the sitemaps.txt and llms-full.txt work fine. I will figure out and debug what is going wrong with the llms.txt path, but I thought worth sharing with you that it is now in place.
I found an LLMs.TXT plugin for Discourse which creates