skills.rest
tottenjordan
tottenjordanCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

agent-engine

Deploy and manage ADK agents on Vertex AI Agent Engine with sessions, memory, and A2A support.

1|Updated Mar 19, 2026
Search Tags:#deployment#google cloud#adk#memory bank#a2a#vertex ai#agent engine
One-click install
npx skills add https://github.com/tottenjordan/me-skittles --skill agent-engine-tottenjordan
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: agent-engine
Source: https://github.com/tottenjordan/me-skittles/tree/main/gemini/agent-engine
Command: npx skills add https://github.com/tottenjordan/me-skittles --skill agent-engine-tottenjordan

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires google-cloud-aiplatform, google-adk, a2a-sdk, uvicorn, fastapi, and includes references (resource) components.

What problem does it solve? Moving an AI agent from local development to a managed production runtime on Google Cloud involves deployment packaging, session handling, scaling configuration, and memory management, all of which are error-prone without a clear reference. ## Core Features & Use Cases - Agent Deployment: Deploy ADK agents from agent objects or source files to Vertex AI Agent Engine with configurable requirements, environment variables, and resource limits. - Sessions and Memory Bank: Manage conversation sessions and generate, retrieve, and search long-term user memories with scope-based isolation. - Multi-Agent A2A Deployment: Deploy hierarchical multi-agent systems in phases (leaf, functional, orchestrator) using the A2A protocol and RemoteA2aAgent. - Use Case: You built an ADK agent locally and need to ship it to production with auto-scaling, per-user memory, and evaluation metrics. This Skill walks through client initialization, deployment config, querying, and troubleshooting. ## Quick Start Deploy my ADK agent to Vertex AI Agent Engine in project my-project with a staging bucket and show me how to query it.

Frequently Asked Questions about agent-engine

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I deploy an ADK agent to Vertex AI Agent Engine?▼

Wrap the agent in agent_engines.AdkApp and call client.agent_engines.create with a config containing a staging bucket and requirements. You can deploy from an agent object for development or from source files with an entrypoint module for CI/CD.

How do I query a deployed agent on Agent Engine?▼

Use remote_agent.query for synchronous calls or remote_agent.stream_query for streaming events, passing a user_id and message. Async streaming is available via async_stream_query, and a REST endpoint supports streamQuery over HTTP.

Does Agent Engine support long-term memory across sessions?▼

Yes, Memory Bank is auto-enabled with ADK 1.5.0 or later in the Agent Engine runtime. Generate memories from conversations or direct facts, retrieve them by user scope or similarity search, and use PreloadMemoryTool for read-only loading at conversation start.

Why does my Agent Engine deployment fail with module not found errors?▼

Cloudpickle fails when the agent imports local modules unavailable in the runtime. Build the agent inline in the deploy script with no external imports, or pass local packages via extra_packages and source_packages in the deployment config.

How do I reduce cold start latency on Agent Engine?▼

Set min_instances in the resource config to keep warm instances running, which can cut latency from about 4.7 seconds to 1.4 seconds. Increase container_concurrency in multiples of 9 for I/O-bound agents, watching for out-of-memory errors.

Can I deploy a multi-agent A2A system on Agent Engine?▼

Yes, deploy in phases: leaf agents first, then functional agents with leaf URLs as environment variables, then the orchestrator. Use RemoteA2aAgent to connect agents, and verify each agent URL ends with /.well-known/agent.json.

Related Skills

Explore More Software Engineering Skills→
SeanChenRSeanChenR

adk-dev-guide

Enforce spec-first ADK agent development with adk eval workflows.

Community
Advanced
wshobsonwshobson
39.3k

eval-harness-first

Build evaluation harnesses with goldens, graders, and baselines that gate fine-tuning runs.

Community
Advanced
ginkidaginkida
20

gcp

Design secure, scalable Google Cloud Platform infrastructure using gcloud CLI and Terraform.

Community
Advanced
rnavarychrnavarych
14

role-devops:gcp-expert

Guide Google Cloud Platform infrastructure design, security, and optimization.

Community
Advanced