Home

Tutorial: google-adk-docker-model-runner

This project showcases how to build sophisticated AI agents using the Google Application Development Kit (ADK) and Docker model runners. Each agent is like a specialized worker that can use a Large Language Model (LLM) for understanding and generating text, and leverage Tools for external actions. The Runner acts as an execution engine, orchestrating the flow of agents, especially when combined using Agent Composition methods like sequential or parallel execution. A Session maintains the conversation state and shared memory between agents throughout an interaction.

Visual Overview

flowchart TD
    A0["Agent
"]
    A1["Model (LLM)
"]
    A2["Tool
"]
    A3["Runner
"]
    A4["Session
"]
    A5["Agent Composition (Sequential, Parallel, Loop)
"]
    A0 -- "Uses LLM" --> A1
    A0 -- "Uses Tool" --> A2
    A3 -- "Runs Agent" --> A0
    A3 -- "Uses Session" --> A4
    A5 -- "Contains Agents" --> A0
    A3 -- "Orchestrates Composition" --> A5

Loading

Chapters

1. Agent
2. Model (LLM)
3. Tool
4. Runner
5. Session
6. Agent Composition (Sequential, Parallel, Loop)

---