Product information "Google AI Agent Pipelines"
Google AI Agent Pipelines, part of the Gemini Enterprise Agent Platform, is a fully managed orchestration service for building and running repeatable, production-grade ML and AI workflows on Google Cloud.
This integration brings Google AI Agent Pipelines into your UAC workflows. From a standard UAC task you can submit a pipeline run, pass runtime parameters, and either return immediately or block until the pipeline reaches a terminal state—all while UAC retains the run identifier and console URL as reusable output variables for downstream tasks.
Key Features
- Submit a new AI Agent Pipeline run from a compiled pipeline template (JSON or YAML) with full control over runtime parameters.
- Choose between two execution modes: Submit Only (non-blocking, returns as soon as GCP accepts the job) and Wait for Completion (polls GCP at a configurable interval until the pipeline succeeds, fails, or is cancelled).
- Cancel an in-flight pipeline run and optionally wait until the cancellation is confirmed.
- Pass pipeline runtime parameters either as an inline JSON object directly in the task form or by referencing a UAC Data Script entity, enabling dynamic parameterisation from upstream task outputs or UAC Variables.
- Control component-level caching to skip re-execution of pipeline steps whose inputs have not changed since a previous run. Reusing cached outputs avoids spinning up containers for unchanged components, reducing both pipeline execution time and GCP compute costs.
- Structured output fields available as UAC Variables for use by downstream tasks, notifications, or conditional branching logic within a workflow.
| Extension Name: | ue-google-ai-agent-pipelines |
|---|---|
| Universal Template Name: | Google AI Agent Pipelines |
| Vendor Product: | Gemini Enterprise Agent Platform |
| Version: | 1.0.0 |
| Vendor: | |
| Compatibility : | UC/UA 7.7 and above |
| Support: | Stonebranch Certified |
Use Case 1: Trigger AI Agent Pipeline, Wait for Completion, and Notify Stakeholders
Description
This workflow demonstrates orchestrating a Google Cloud AI Agent Pipeline run directly from Universal Controller. Using the ue-google-ai-agent-pipelines extension, it submits a compiled pipeline template to the Gemini Enterprise Agent Platform and polls the run until it reaches a terminal state. Depending on the observed outcome, the ue-email extension notifies stakeholders through a dedicated email task on either the Success or Failure exit path.
Business Value
- Give stakeholders immediate visibility into AI Agent Pipeline outcomes without requiring direct access to the Google Cloud Console
- Eliminate manual monitoring of long-running AI/ML pipeline executions by centralizing status polling and notification inside Universal Controller
- Standardize AI Agent Pipeline governance by combining Gemini Enterprise Agent Platform execution with UAC's existing scheduling, credential, and alerting capabilities
- Reduce time to awareness of pipeline failures, enabling faster remediation of failed training or inference runs
Technical Capabilities
- Submission of a compiled AI Agent Pipeline template to the Gemini Enterprise Agent Platform via the ue-google-ai-agent-pipelines extension, with configurable runtime parameters and execution caching
- Polling-based monitoring of the pipeline run until a terminal state is reached, with output fields tracking the run's resource name, state, and Console URL
- Conditional email notification via the ue-email extension, routed to a dedicated Success or Failure task depending on the pipeline's observed outcome
- Microsoft 365 (Azure AD tenant and sender account) based email delivery, keeping notification credentials centrally managed alongside the rest of the automation
This solution delivers governed, hands-off Google Cloud AI Agent Pipeline execution directly from Universal Controller, closing the loop between long-running AI/ML automation and the stakeholders who need to know how it turned out.