Back
Nichewit
Progressively use more resource intensive scraping methods based on patterns in HTTP response #nichewit
Experiment with Cloudflare AI workers for smart enrichment #nichewit
Restructure #nichewit datamodel to allow for more flexible campaigning and prompting.
Prefer category matching when building smart sheet #nichewit
#nichewit catch up and experiment with new classifications
Work on #nichewit, pull in production datasets to iterate bit faster locally
Get back on track with #nichewit "smart-sheet" generation
Dust off #nichewit
Quick #nichewit catchup with @pimbaljon
Rebuild Hetzner server with backup from last night because I made a 'lil mistake 😅 #nichewit
Look into some AI hallucinations #nichewit
Work on ideas for finishing #nichewit onboarding
Finally process the last few of the ~2.5M embeddings #nichewit
Enqueue embeddings creation for full companies database in #nichewit do start doing better similarity searches and recommendations
Generate pgvector embeddings for #nichewit to improve similarity searches.
Make #nichewit source parser smarter to prevent sending any irrelevant page content to the LLM
Upgrade Groq plan to increase rate limits #nichewit
Add Groq to #nichewit as well and tweak enrichment prompt further, getting somewhere!
Give up on hosting my own LLM for #nichewit and explore most affordable options