Skip to main content

Build an agent with OpenAI function calling

Edit on Github

Function calling lets the model call functions you define and act on the results. You describe each tool with a JSON schema; when the model decides to use one it returns a tool call instead of a final answer. You run it, send the result back, and repeat until the model is done. This loop is the core of an agent. Set OPENAI_API_KEY before running.

import OpenAI from "npm:openai";

const client = new OpenAI();
Describe the tool with a JSON schema and a prescriptive description so the model knows when to call it.
const tools: OpenAI.Chat.Completions.ChatCompletionTool[] = [
  {
    type: "function",
    function: {
      name: "get_weather",
      description:
        "Get the current weather for a city. Call this whenever the user " +
        "asks about weather conditions.",
      parameters: {
        type: "object",
        properties: {
          city: { type: "string", description: "City name, e.g. Tokyo" },
        },
        required: ["city"],
      },
    },
  },
];
The actual implementation. A real app would call a weather API here.
function getWeather(city: string): string {
  const data: Record<string, string> = {
    Tokyo: "18°C, clear",
    Oslo: "4°C, snow",
  };
  return data[city] ?? `No data for ${city}`;
}

const messages: OpenAI.Chat.Completions.ChatCompletionMessageParam[] = [
  { role: "user", content: "Compare the weather in Tokyo and Oslo right now." },
];
Agent loop: keep calling the model until it stops requesting tools.
while (true) {
  const response = await client.chat.completions.create({
    model: "gpt-4o",
    messages,
    tools,
  });
  const message = response.choices[0].message;
Record the assistant turn; it holds any tool calls we must answer.
  messages.push(message);
No tool calls means the model produced its final answer.
  if (!message.tool_calls) {
    console.log(message.content);
    break;
  }
Run each requested tool and append its result as a tool message.
  for (const call of message.tool_calls) {
    if (call.type !== "function") continue;
    const { city } = JSON.parse(call.function.arguments);
    messages.push({
      role: "tool",
      tool_call_id: call.id,
      content: getWeather(city),
    });
  }
}

Run this example locally using the Deno CLI:

deno run -N -E https://docs.deno.com/examples/scripts/openai_tool_use.ts

Additional resources

Did you find what you needed?

Privacy policy