← All features

Private Retrieval Learning

Turn a real miss into a regression test you own

Search quality tools usually improve by watching you: clicks become signals, signals become a model, and one day ranking changed and nobody can say why. GNO inverts that. Recording is off until you enable it, only judgments you make explicitly become qrels, exported fixtures contain no source text, and replay compares a candidate against a baseline read-only. It recommends. You promote.

Use cases
Reproducing a recurring retrieval miss
Evaluating a ranking or model change on your own corpus
Building private regression fixtures without exporting source text

What it gives you

  • Local-only trace IDs across query, Capsule, open, cite, and pin outcomes
  • Explicit relevant, irrelevant, and missing-expected judgments
  • Bounded inspection, retention, deletion, and physical purge
  • Deterministic content-free qrels export
  • Read-only baseline-versus-candidate replay
  • Stale and missing source disclosure
  • No upload, hidden click inference, or automatic ranking mutation

Try it yourself

Representative commands and entry points. Full reference lives in the documentation.

gno trace list --md
gno trace label <trace-id> --label relevant --target <ref>
gno trace export <trace-id> --format qrels --output qrels.json
gno trace replay <qrels-export-id> --candidate hybrid

Keep reading

Related features and docs.