Skip to main content

​
Overview

The rootAgent is the main LLM agent in the ADK Utils Example project. It demonstrates integration with the Google ADK framework and showcases the use of OllamaModel from the @yagolopez/adk-utils package.

Key Features

  • Multi-model support (Gemini, Ollama)
  • Three powerful function tools
  • Flexible model configuration
  • Real-time interaction capabilities

​
Agent Configuration

The rootAgent is configured with the following properties:
​
name
string
Agent identifier: "agent1"
​
model
OllamaModel | string
The LLM model to use. Currently configured to use gpt-oss:120b-cloud via Ollama’s cloud endpoint.
​
description
string
Agent description: “Agent with three function tools: get_current_time, create_mermaid_diagram and view_source_code. It retrieves the current time, creates mermaid diagrams and visualizes source code.”
​
instruction
string
System instructions that guide the agent’s behavior and tool usage.
​
tools
FunctionTool[]
Array of three function tools: getCurrentTime, createMermaidDiagram, and viewSourceCode.

​
Complete Implementation

Here’s the full agent definition from agent1.ts: 
import { FunctionTool, LlmAgent } from "@google/adk";
import { z } from "zod";
import { OllamaModel } from "@yagolopez/adk-utils";

export const rootAgent = new LlmAgent({
  name: "agent1",
  model: new OllamaModel("gpt-oss:120b-cloud", "https://ollama.com"),
  description:
    "Agent with three function tools: get_current_time, create_mermaid_diagram and view_source_code. It retrieves the current time, creates mermaid diagrams and visualizes source code.",
  instruction: `You are a helpful assistant.
                If the user ask for the time in a city, Use the 'get_current_time' tool for this purpose.
                If the user asks for a diagram or visual representation, use the 'create_mermaid_diagram' tool.
                If the user asks to view source code, use the 'view_source_code' tool.`,
  tools: [getCurrentTime, createMermaidDiagram, viewSourceCode],
});

​
OllamaModel Integration

The agent uses OllamaModel from @yagolopez/adk-utils to connect to various Ollama-compatible endpoints: 
// Current configuration (cloud-hosted model)
model: new OllamaModel("gpt-oss:120b-cloud", "https://ollama.com")

​
Available Model Options

The source code includes several commented model options: 
model: 'gemini-2.5-flash'

​
Modifying the Agent

​
Changing the Model

To switch between different models, simply uncomment the desired model configuration: 
export const rootAgent = new LlmAgent({
  name: "agent1",
  // Option 1: Use Gemini
  model: 'gemini-2.5-flash',
  
  // Option 2: Use local Ollama
  // model: new OllamaModel("qwen3:0.6b", "http://localhost:11434"),
  
  // ... rest of configuration
});

​
Adding New Tools

To add a new tool to the agent:
  1. Create the tool using FunctionTool:
const myNewTool = new FunctionTool({
  name: "my_new_tool",
  description: "Description of what the tool does",
  parameters: z.object({
    param1: z.string().describe("Parameter description"),
  }),
  execute: ({ param1 }) => {
    // Tool implementation
    return {
      status: "success",
      report: `Result: ${param1}`,
    };
  },
});
  1. Add to the tools array:
tools: [getCurrentTime, createMermaidDiagram, viewSourceCode, myNewTool],
  1. Update the instruction to guide the agent on when to use the new tool:
instruction: `You are a helpful assistant.
              If the user ask for the time in a city, Use the 'get_current_time' tool for this purpose.
              If the user asks for a diagram or visual representation, use the 'create_mermaid_diagram' tool.
              If the user asks to view source code, use the 'view_source_code' tool.
              If the user asks for [specific condition], use the 'my_new_tool' tool.`,

​
Customizing Instructions

The instruction field is crucial for guiding agent behavior. It should:
  • Clearly define the agent’s role
  • Specify when to use each tool
  • Provide context for decision-making
  • Use clear, directive language

​
Agent Location

Source: ~/workspace/source/app/agents/agent1.ts

​
Next Steps

Tools Reference

Learn about the three function tools available to the agent

OllamaModel

Explore the OllamaModel utility for connecting to Ollama endpoints

Build docs developers (and LLMs) love

Get started for freeTalk to us