ADMIN · QUALITY CONTROL

Evaluation observatory

Durable golden-set runs live in PostgreSQL. Track quality, routing regressions, latency, cost, and coverage instead of relying on a one-off result.

Return to SpendIQ

Evaluate the snapshot

Starting a run captures today’s active golden questions and prompt versions before the worker begins. Every result is traceable to immutable inputs.

Quality over time

Pass rate for completed runs. January–April points labeled Sample are deterministic demo fixtures, not provider-produced results.

No baseline

Complete a run to establish the quality baseline.

No completed evaluations have been recorded yet.

No report selected

Run the expanded golden set, or select a completed run once history is available. Provider credentials never reach the browser.