YC Indexor

How it uses Jev

A search embeds the user's query against company descriptions and logos, uses OCR and tags to narrow 6,241 companies to at most 320, then sends those finalists to Jev for calibrated relevance scores in one bounded parallel request. A separate preparation step uses Jev to reassign YC's tags.

Why it's interesting

It combines text, visual and OCR retrieval before spending Jev calls on only the finalists. The probabilities become an interactive search result, and the code exposes the fallback when Jev is unavailable.

How often does Jev run?

This project calls Jev on demand, in response to user input or events.