Go Proposal Digest Jev Search
A Cloudflare site search reranks Go proposal articles with Jev relatedness judgments.
Added to Jevfast
How it uses Jev
The author describes a GoProposalWeeklyDigest search flow: Cloudflare Workers AI bge-m3 embeds article content through GitHub Actions, Vectorize retrieves 20 candidates, and TypeSafe Jev Noul judges relatedness using article titles and summaries. Results are filtered at 0.2 and ranked in descending order. The available post and article summary do not show the exact Jev question key or full wire request.
Why it is interesting
It uses a typed Jev judgment after vector retrieval to rerank a small candidate set for a domain-specific article search. A live example query for “generic” makes the user-facing search experience inspectable, while the public material leaves the exact request shape unspecified.