Skip to content

openai.md

Latest commit

 

History

History
94 lines (73 loc) · 3.24 KB

openai.md

File metadata and controls

94 lines (73 loc) · 3.24 KB
Raw

OpenAI Agents SDK

Use the Conductor Java Agent SDK with OpenAI Agents SDK-style tool definitions. The OpenAIAgent bridge accepts @Tool-annotated POJOs and registers them as Conductor worker tasks, routing the agent through the server's OpenAINormalizer.

Dependency

implementation 'org.conductoross:conductor-client-ai:<VERSION>'

The bridge uses the LangChain4j @Tool annotation as a practical equivalent of the Python OpenAI Agents SDK @function_tool decorator — add it if you need the annotation:

implementation 'dev.langchain4j:langchain4j:1.0.0'

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

Usage

import dev.langchain4j.agent.tool.Tool;
import dev.langchain4j.agent.tool.P;
import org.conductoross.conductor.ai.Agent;
import org.conductoross.conductor.ai.AgentRuntime;
import org.conductoross.conductor.ai.frameworks.OpenAIAgent;
import org.conductoross.conductor.ai.model.AgentResult;

public class ShoppingTools {

    @Tool(name = "search_products", value = "Search for products by keyword")
    public String searchProducts(@P("query") String query, @P("maxResults") int maxResults) {
        return callSearchApi(query, maxResults);
    }

    @Tool(name = "add_to_cart", value = "Add a product to the shopping cart")
    public String addToCart(@P("productId") String productId, @P("quantity") int quantity) {
        return cartService.add(productId, quantity);
    }
}

Agent agent = OpenAIAgent.builder()
    .name("shopping_assistant")
    .model("openai/gpt-4o-mini")
    .instructions("Help users find and purchase products.")
    .tools(new ShoppingTools())
    .build();

try (AgentRuntime runtime = new AgentRuntime()) {
    AgentResult result = runtime.run(agent, "Find me a blue jacket under $100");
    System.out.println(result.getOutput());
}

Handoffs

OpenAI Agents SDK-style handoffs let the LLM transfer control to a specialist agent:

Agent billingAgent = Agent.builder()
    .name("billing_agent")
    .model("openai/gpt-4o-mini")
    .instructions("Handle billing and payment questions.")
    .build();

Agent supportAgent = OpenAIAgent.builder()
    .name("support_agent")
    .model("openai/gpt-4o-mini")
    .instructions("Handle general support. Transfer billing issues to the billing agent.")
    .handoffs(billingAgent)                      // adds billing_agent as a handoff target
    .build();

Structured output

Agent agent = OpenAIAgent.builder()
    .name("classifier")
    .model("openai/gpt-4o-mini")
    .instructions("Classify the sentiment of the input.")
    .outputType("SentimentResult")               // server-side structured output type name
    .build();

Builder reference

Method Description
name(String) Required. Agent and workflow name.
model(String) Required. "provider/model" string.
instructions(String) System prompt.
tools(Object...) @Tool-annotated POJOs; each method becomes a worker task.
handoffs(Agent...) Sub-agents the LLM can hand off to.
outputType(String) Structured output type name for the server normalizer.