Show HN: Dirt-cheap custom ranking layer for your vector search

1 points by vtaya 10 hours ago

I built CoderSwap for coders who want experimental, cheap search that ships fast. It’s a tiny layer on top of your DB that does hybrid (vector + keyword) search with explainable scoring. You can express ranking rules in plain English (e.g., boost recent, prefer verified, down-rank competitors) and we compile them into a safe, tunable scoring policy. Why use it with your existing vector DB? It adds governance (versioning/rollback) and explainability without an ML team. Free tier: 100MB, 10k calls/month. Quickstart/docs on the site. Would love feedback: what’s confusing, where it breaks, and what knobs you’d want.