How can I configure the Assistant API to return JSON output without triggering a validation error?

Below is an explanation with a suggested solution:

Background

According to the API documentation, you can enable "JSON mode" by specifying the response format as either a string or an object. The docs show that setting this parameter to an object like

JSONCopied
{ "type": "json_object" }

should guarantee that the assistant's output is valid JSON. However, when using JSON mode you must also tell the model (via a system or user message) to produce JSON output. Otherwise, the model might not behave as expected.

The Issue

The error you’re encountering:

BadRequestError: 400 1 validation error for Request  
body → response_format  
extra fields not permitted (type=value_error.extra)

suggests that the API is rejecting extra keys in the object passed as the value of response_format.

Recommended Workaround

Even though the API docs indicate that a response format can be specified as an object, your error indicates that, at least in this beta endpoint, the API might not accept the object form as provided.

So the workaround is to pass the response format as a string:

JAVASCRIPTCopied
const myAssistant = await openai.beta.assistants.create({
  instructions: "Cool instruction",
  name: "Cool Assistant",
  tools: [{ type: "retrieval" }],
  model: "gpt-4-1106-preview",
  response_format: "json_object",
  file_ids: [],
});

Additional Advice

• Model Instruction: When using JSON mode, remember to instruct your model (typically via a system or user prompt) to produce valid JSON. Without this additional instruction, even setting "json_object" may not result in valid JSON output.

• Review Documentation: Since beta endpoints can sometimes have slightly different expectations or limitations, double-check the latest documentation if you run into similar issues.

This adjustment should resolve the extra field error while still enabling JSON mode for the assistant's output.