DEV Community

jg-noncelogic
jg-noncelogic

Posted on • Originally published at vecstore.app

Show HN: Search 10k paintings by describing them in plain English

Show HN: Search 10k paintings by describing them in plain English. Vecstore demo: image + semantic search, NSFW detection, sub‑200ms responses, 100+ languages, free tier. Try: https://vecstore.app (HN: https://news.ycombinator.com/item?id=47291954)

What this actually buys you: zero ML ops and instant image+text search you can call from your app. Demo claims <200ms and NSFW filtering. Real trade-off is control — vectors, retention, and query logic live with the provider.

If you build for agencies, financial advisers, or legal teams, require BYOK (bring your own key) or a self‑host option, mandatory human review before publish, exportable vectors, and a clear retention policy. Those constraints are product features, not bugs.

Quick integration checklist: index image embeddings + short captions (BLIP/CLIP), pre‑filter NSFW with a small model, batch index updates to cut cost, and map vector IDs to canonical metadata for audits. Which of these would save you the most dev time?

Top comments (0)