RedHub AI
How the roll-up gate works
How the Canary Call Library works
This tool grades every canary test call this release against your own baseline run on the voice behaviors a prompt-text diff never sees, then rolls the library up to its worst canary. It reads 4 canaries and rolls them into one verdict — the worst, never the average — then a gate that can only make it worse.
The 4 canaries
Some drills don't bend: barge-in below 5 of 5, or a scripted escalation that didn't fire, forces CANARY BROKE regardless of how correct the answers were — an agent that talks over callers fails the call, the words being right doesn't save it. Every canary rolls up worst-first, and a single broken canary escalates the whole library past SLIPS TO REVIEW to REGRESSIONS FOUND.
Why: Three canaries match the run you shipped. One — CN-002 — is below baseline: its p95 latency jumped from 950ms to 1250ms, past the slack, while the answer stayed perfectly correct. A prompt-text diff sees no change; the call just got slower. No canary broke a dispositive drill, so the library rolls up to SLIPS TO REVIEW rather than REGRESSIONS FOUND.
No single canary is the crisis — the clustering is. The roll-up takes the worst, the gate escalates when trouble compounds, and it names the one thing to fix first. Scope note. The Canary Call Library scores runs you record against your own baseline; it places no calls, connects to nothing, and grades no person. It is a release-quality aid, not a certification, and not legal advice.