Using Structured Outputs throws body used already for: https://api.openai.com/v1/chat/completions in 4.59.0

[jreidgreer]

Confirm this is a Node library issue and not an underlying OpenAI API issue

• This is an issue with the Node library

Describe the bug

Since 4.59.0, using structured outputs with Zod causes the following error:

TypeError: body used already for: https://api.openai.com/v1/chat/completions
    at Response.consumeBody (/dev/app/node_modules/node-fetch/lib/index.js:344:30)
    at Response.json (/dev/app/node_modules/node-fetch/lib/index.js:269:22)
    at Proxy.defaultParseResponse (/dev/app/node_modules/openai/src/core.ts:68:33)
    at /dev/app/node_modules/openai/src/core.ts:121:42
    at processTicksAndRejections (node:internal/process/task_queues:95:5)
    at /dev/app/dist/apps/api/main.js:4027:32

To Reproduce

You can reproduce by using the chat completions endpoint with structured outputs. I haven't tried this without Zod.

const completion = await this.client.beta.chat.completions.parse(
  {
    model: "gpt-4o-mini-2024-07-18",
    temperature: 0,
    messages: [
      {
        role: "system",
        content: prompt,
      },
      {
        role: "user",
        content: input,
      },
    ],
    response_format: zodResponseFormat(schema, schemaName),
  },
  { timeout: 60_000 },
);

Using this.client.chat.completions.create with response_format: { type: "json_object" } works just fine.

This is a regression introduced in v4.59.0

Code snippets

No response

OS

macOS

Node version

v20.13.1

Library version

4.59.0

[jreidgreer]

Looking at a comparison with v4.58.2 and v4.59.0: v4.58.2...v4.59.0

This seems like the most likely candidate: #1058

[RobertCraigie]

Thanks for the report @jreidgreer, I can't reproduce this, could you share more details about how you're using the SDK?

Here's the snippet I used to try to reproduce

import { zodResponseFormat } from 'openai/helpers/zod';
import OpenAI from 'openai/index';
import { z } from 'zod';

const Step = z.object({
  explanation: z.string(),
  output: z.string(),
});

const MathResponse = z.object({
  steps: z.array(Step),
  final_answer: z.string(),
});

const client = new OpenAI();

async function makeRequest() {
  const completion = await client.beta.chat.completions.parse({
    model: 'gpt-4o-2024-08-06',
    messages: [
      { role: 'system', content: 'You are a helpful math tutor.' },
      { role: 'user', content: 'solve 8x + 31 = 2' },
    ],
    response_format: zodResponseFormat(MathResponse, 'math_response'),
  });

  console.dir(completion, { depth: 5 });

  const message = completion.choices[0]?.message;
  if (message?.parsed) {
    console.log(message.parsed.steps);
    console.log(`answer: ${message.parsed.final_answer}`);
  }
}

async function main() {
  await makeRequest();
  await makeRequest();
}

main();

[rattrayalex]

@jreidgreer can you share a stackblitz or equivalent that reproduces the bug?