Good feedback. We're moving fast and onboarding for sure needs to be improved. We added a playground when you first login and expanded docs here: https://docs.isomeric.ai/api-reference/prompting
Cool tech, very interested to try it out. How much 'influence' do users have on the labelling/nesting structure of the resulting JSON? Are pages parsed whole after a prompt has been fine-tuned or only scraped for the specific parts that users are interested in (i.e. selectors)? Is screen scraping possible of canvas based content?
In a nutshell you prompt the Isomeric LLM with the format of the data you want to see. This way you can completely control the output. We have support for even finer grained control through things like regular expressions too (https://docs.isomeric.ai/api-reference/prompting#pattern).
As for canvas. No, only text can currently be ingested.
4 comments
[ 3.2 ms ] story [ 23.4 ms ] threadUpdate: I found them in the bottom right corner of the page. This process could use better onboarding.
Good feedback. We're moving fast and onboarding for sure needs to be improved. We added a playground when you first login and expanded docs here: https://docs.isomeric.ai/api-reference/prompting
In a nutshell you prompt the Isomeric LLM with the format of the data you want to see. This way you can completely control the output. We have support for even finer grained control through things like regular expressions too (https://docs.isomeric.ai/api-reference/prompting#pattern).
As for canvas. No, only text can currently be ingested.