openai
#10006
response_format_error
"messages" must include "json" to use "response_format" as "json_object".
This error has been identified and solved.
Reason
The error you are seeing, where it states that "messages" must contain the word "json" in some form to use "response_format" of type "json_object", is likely due to a specific requirement in the OpenAI API. Here’s a concise explanation:
The OpenAI API has strict criteria for when the
response_formatcan be set tojson_object. Specifically, the API expects the input or the instructions provided to the model to explicitly indicate that the response should be in JSON format. If this indication is missing, the API will reject the request with a 400 error, as it does not meet the required conditions for generating a JSON response.
Solution
To fix the error where "messages" must include "json" to use "response_format" as "json_object", you need to ensure the following:
Include the word "json" in your prompt or instructions to the model.
Ensure that the model you are using supports the
response_formatparameter set tojson_object. Currently, this is supported for models likegpt-3.5-turboandgpt-4-turbo, but not forgpt-4-visionmodels.
Key points to check:
Modify your prompt to explicitly mention JSON.
Verify the model you are using is compatible with the
json_objectresponse format.
