4 ms·
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 af
by dleeftink 3y ago
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?
- zachblank 3y agoFounder here. Docs where just updated to answer your question broadly (https://docs.isomeric.ai/api-reference/prompting 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 https://docs.isomeric.ai/api-reference/prompting#pattern). As for canvas. No, only text can currently be ingested.