AI Inspiration WorldWhat it is
AQuA, short for Ambient Quality Agent, is a recipe Google publishes in its adk-recipes repository. The README says it gives production ADK agents continuous observability and autonomous diagnosis from the telemetry they already emit. It can run on a schedule, after each deployment or on demand; it uses a large language model to review agent sessions, lets you add custom metrics in Python, groups repeated failures together, and shows results on an IAP-protected dashboard. Your agents don't need to change: it works from the telemetry they already produce.
Who it's for:developers already running production agents on Google ADK who can configure their organisation's Google Cloud project. Readers who don't write code won't get much from it.
Before you start: it needs Python 3.11 to 3.13, uv and npm, agents-cli 1.6 or later, Terraform, and a Google organisation project with Application Default Credentials (ADC). It reads production conversation and call logs plus source snapshots and sends them to models such as Gemini for review, so it needs read access to your telemetry dataset and bucket. It can launch investigations automatically on a schedule and after deployments, and can keep agent memory. The dashboard is protected by Cloud Run IAP by default; the configuration has an option to turn IAP off, which weakens that protection, and we don't recommend it. The code is under the Apache-2.0 licence, and running it uses paid services such as Google Cloud, Gemini, BigQuery and Cloud Run.
Based on the project's public documentation. Not tested by AI Inspiration World.