Relevan

About Relevan

Relevan builds search that learns from agents. Agents search constantly, and every search shows what worked, which results they used, which they skipped, and what they expected but didn’t get. Relevan turns that feedback into ranking changes for your index, and tests each change against real queries before it ships. It runs on OpenSearch, with our own tuning harness on top.

Product Description

Relevan is managed search on OpenSearch for AI agents and the apps they serve. Send your records through one API, and Relevan tunes each index on your own data:

  • Field weights
  • BM25
  • Hybrid retrieval
  • Entity matching
  • Learned re-ranking

Agents report which results helped through a feedback API, captured in the User Behavior Insights (UBI) format. When a search misses a document it should have found, the learning loop proposes changes and tests them against judged queries before any of them ships. Every response says which field matched and which words to try next, so an agent can correct its own search within one conversation. Set it up from a coding agent with our plugin or skills, or with the Python SDK. Start free at https://relevan.dev.

Office Location

Chesapeake, VA, USA