Skip to content

google-adk.md

Latest commit

 

History

History
85 lines (68 loc) · 3.23 KB

google-adk.md

File metadata and controls

85 lines (68 loc) · 3.23 KB
Raw

Google ADK

Use Google's Agent Development Kit (ADK) agents directly with Conductor. The AdkBridge converts a native LlmAgent (or any BaseAgent) into an Agent, serialising its tools, instructions, and sub-agent graph into the format the server's GoogleADKNormalizer understands.

Dependency

implementation 'org.conductoross:conductor-client-ai:<VERSION>'
implementation 'com.google.adk:google-adk:1.3.0'

1.3.0 is the version exercised by this repository's agent examples. Replace <VERSION> with a published SDK version from Maven Central.

Usage

import com.google.adk.agents.LlmAgent;
import com.google.adk.tools.Annotations;
import com.google.adk.tools.FunctionTool;
import java.util.Map;
import org.conductoross.conductor.ai.Agent;
import org.conductoross.conductor.ai.AgentRuntime;
import org.conductoross.conductor.ai.frameworks.AdkBridge;
import org.conductoross.conductor.ai.model.AgentResult;

// A FunctionTool target — ADK reflects the method and its @Schema params
public static class WeatherService {
    public static Map<String, Object> getWeather(
            @Annotations.Schema(name = "city", description = "City to query") String city) {
        return Map.of("city", city, "condition", "Sunny", "tempC", 22);
    }
}

// Build a native ADK LlmAgent
LlmAgent adkAgent = LlmAgent.builder()
    .name("weather_agent")
    .model("gemini-2.0-flash")
    .instruction("Answer weather questions. Use the getWeather tool.")
    .tools(FunctionTool.create(WeatherService.class, "getWeather"))
    .build();

// Convert to an Agent
Agent agent = AdkBridge.toConductor(adkAgent);

// Run via AgentRuntime
try (AgentRuntime runtime = new AgentRuntime()) {
    AgentResult result = runtime.run(agent, "What's the weather in London?");
    System.out.println(result.getOutput());
}

agentBuilder — attach extra Conductor features

If you want to mix ADK agent structure with Conductor-specific features (guardrails, credentials, callbacks), use agentBuilder() which returns an Agent.Builder you can continue configuring:

import org.conductoross.conductor.ai.guardrail.RegexGuardrail;
import org.conductoross.conductor.ai.enums.Position;
import org.conductoross.conductor.ai.enums.OnFail;

Agent agent = AdkBridge.agentBuilder(adkAgent)
    .credentials("WEATHER_API_KEY")
    .guardrails(RegexGuardrail.builder()
        .name("no_pii").position(Position.OUTPUT)
        .patterns("\\b\\d{3}-\\d{2}-\\d{4}\\b")   // SSN-like
        .onFail(OnFail.RAISE).build())
    .maxTurns(10)
    .build();

What gets mapped

ADK concept Conductor mapping
LlmAgent.name() Agent.name
LlmAgent.model() Agent.model
LlmAgent.instruction() Agent.instructions
FunctionTool Conductor worker task (via WorkerManager)
AgentTool Sub-agent (nested Agent)
GoogleSearchTool HTTP tool
BuiltInCodeExecutionTool Code execution tool
Sub-agents (.subAgents()) Agent.agents
LoopAgent / SequentialAgent Strategy.SEQUENTIAL

Model requirement: Google ADK agents require a Gemini model (for example, gemini-2.0-flash). Configure the Google AI or Vertex AI credential on the Conductor server.