Google has released AQuA (Ambient Quality Agent) as open‑source code. The tool runs beside an AI agent in Google Cloud, reads telemetry from Cloud Trace, Logging or BigQuery, and flags sessions that break defined quality rules. It groups similar failures, checks them against the exact source revision deployed, and suggests code locations for fixes. In a test on a travel‑booking agent, AQuA helped cut seat‑selection bypasses from 15 to two sessions after prompt tweaks.
Why it matters
Production AI teams can catch hidden bugs earlier, reducing customer‑facing errors without waiting for manual testing.
