t1-p3-05draft5–8h
Monitoring dashboard
requires: t1-p3-04
Spec
Log every query with latency, token cost, confidence score, and Phase 2 failure category to a time-series store, and build a dashboard showing daily query volume, average confidence, and cost per query.
Why this matters
Cost per query is the number that decides whether the feature survives contact with a finance review. Most teams discover they never measured it at exactly the wrong moment.
Verify
No result yet. Run `learn verify t1-p3-05` locally, or push to trigger CI.
contract
unrun
evidence
unrun
Rubric
# t1-p3-05 — Monitoring dashboard
**Task:** Log every query with latency, token cost, confidence score, and Phase 2 failure
category to a time-series store, and build a dashboard showing daily query volume,
average confidence, and cost per query.
**Strong:** All three named metrics (daily volume, average confidence, cost per query)
are visible on one dashboard, trending over time - not a single current-value snapshot -
and cost per query specifically is a real computed number, not eyeballed from raw logs.
**Passing but weak:** The dashboard exists and shows real data, but is missing one of the
three named metrics, or shows totals/snapshots instead of trends over time.
**Fails:** No working dashboard, or the underlying logging doesn't actually capture one of
the four required fields (latency, token cost, confidence score, failure category) - a
dashboard built on incomplete logs will silently mislead later.