Skip to main content

Building Your First Agent Programmatically

This tutorial will guide you through creating and running a basic AI agent by instantiating the Agent class directly using Agent Forge.

Prerequisites​

  • Agent Forge installed (see Installation).
  • An LLM API Key (e.g., OpenAI) set up. You can set this as an environment variable (e.g., OPENAI_API_KEY). See the Installation guide for more on environment variables.

Step 1: Create Agent Forge Instance & LLM Provider​

First, you need to initialize an LLM provider and then an AgentForge instance. The LLM class acts as a wrapper around different Large Language Model providers, offering a unified interface.

// myFirstAgent.ts
import { AgentForge, LLM, Agent } from "agent-forge";
import dotenv from 'dotenv';

dotenv.config(); // Load environment variables from .env file

async function main() {
  // Create an LLM provider
  const apiKey = process.env.OPENAI_API_KEY;
  if (!apiKey) {
    // It's good practice to handle the absence of an API key.
    // For a production app, you might throw an error or use a default behavior.
    console.error("Error: OPENAI_API_KEY is not set. Please set it in your .env file or directly.");
    return;
  }

  const llmProvider = new LLM("openai", {
    apiKey,
  });

  // Create the AgentForge instance (optional for direct agent instantiation if not using other Forge features like YAML loading or global tool registration)
  const forge = new AgentForge(llmProvider);
  // console.log("AgentForge instance and LLM provider created.");

  // Define and Instantiate Your Agent
  const researchAgent = new Agent({
    name: "ResearchAgentProgrammatic",
    role: "Research Assistant",
    description: "An agent that helps with online research, created programmatically.",
    objective: "Find accurate and relevant information based on user queries.",
    model: "gpt-4", // Or your preferred model like gpt-3.5-turbo. Ensure compatibility with your API key.
    temperature: 0.7,
    llm: llmProvider, // Assign the LLM provider directly
    tools: [
      // Example: If you have a WebSearchTool class instance, you could add it here:
      // new WebSearchTool() 
      // For this first example, we are not adding specific tools to keep it simple.
    ],
  });
  // console.log(`Agent '${researchAgent.name}' created programmatically.`);
  
  // Run Your Agent
  // console.log("Running agent...");
  try {
    const result = await researchAgent.run("What are the latest developments in AI?");
    console.log("\nAgent Execution Result:", result);
  } catch (error) {
    console.error("Error running agent:", error);
  }
}

main().catch(console.error);

This example shows how to create a functional agent by directly instantiating and configuring the Agent class. The next section, "YAML-Defined Agents," will show how to achieve similar configurations using external YAML files for better organization.

Next Steps​

Now that you've built a basic agent programmatically, you can explore: