Google Cloud has updated its Vertex AI Agent Builder with new observability dashboards, faster build-and-deploy tools, and stronger governance controls, aiming to make it easier for developers to move AI agents from prototype to production at scale. The update adds an observability dashboard within the Agent Engine runtime to track token usage, latency, and error rates, along with a new evaluation layer that can simulate user interactions to test agent reliability.
Yasmeen Ahmad, managing director, Google Data Cloud, said that the greatest barrier to productivity in data science was switching between database/data warehouse environments to get data with SQL code, only to export and load it into a Python notebook to do machine learning, while configuring a separate Spark cluster. They might then switch to a BI tool just to visualize results, she said.
To ground its Gemini models, Google uses a proprietary technology called FastSearch. FastSearch is based on RankEmbed signals - a set of search ranking signals-and generates abbreviated, ranked web results that a model can use to produce a grounded response. FastSearch delivers results more quickly than Search because it retrieves fewer documents, but the resulting quality is lower than Search's fully ranked web results.