Course Welcome
Welcome to Agent Workshop.
Five modules on Google's Agent Development Kit, Vertex AI Vector Search, and Agent Engine. You'll go from a single-agent "what's the weather" demo to a multi-agent newsroom grounded in real-time search and a Vector Search archive — deployed on Cloud Run, with evaluations to keep it honest.
An agent is an LLM in a loop with tools, and this workshop takes you from that one idea to a multi-agent service running in production. It assumes Python comfort and a Google Cloud project; it does not assume any prior agent-framework experience. Start with Module 00 — what agents are, what ADK gives you, and the newsroom you'll build.
What you'll build
Module 00
Introduction
Welcome to the Agent Development Workshop! Get an overview of LLM agents, Google's Agent Development Kit (ADK), and the AI Newsroom application we will build.
4 lessons · 10 min
Module 01
Foundation
Build your first agent with Google's Agent Development Kit. Start from the Agent Starter Pack baseline and add a custom Python tool the model can call.
3 lessons · 20 min
Module 02
Agent Orchestration
Build a multi-agent AI Newsroom as a graph-based pipeline. A Planner, a Researcher, and an Editor-in-Chief, wired together with the ADK 2.0 Workflow API.
5 lessons · 30 min
Module 03
Advanced Orchestration
Evolve the linear pipeline into dynamic parallel research with asyncio.gather, then defeat URL hallucination by grounding citations in real Vertex AI Search metadata.
6 lessons · 30 min
Module 04
Vector Search
Add a Vertex AI Vector Search archive lookup alongside the live-news pipeline, with conditional routing in the workflow and structured output the UI can render.
4 lessons · 45 min
Before you begin
- Clone the workshop repo:
git clone github.com/LuisSala/agent-workshop. - Have a Google Cloud project with billing enabled and the Vertex AI API turned on.
- Install Python 3.11+ and
uv(recommended) orpip. - Open the Slides in another tab — we'll reference them in each module.