Conversation QA for voice agents

See exactly why a call passed, or failed. Five dimensions, scored separately, each with the evidence behind it.

$uvx hotato start --demo

conversation-test.v1 · support-bot call 8f2a

Deterministic checksbyte-stable, set the gate

Outcomerefund posted to the ledgerPASS

Policyrequired disclosure readPASS

Conversationtalked over the caller 1.32sFAIL

Speechlatency within budgetPASS

Reliability20 runs, all heldPASS

Advisorya model's read, logged

Model judgetone, sent to reviewADVISORY

Every dimension scored on its own. With the evidence behind each result.