Hi all,
I’m using the new /v1/responses endpoint for classification tasks that typically take two file inputs:
- A taxonomy file (PDF)
- A project document (PDF.)
In the standard (non-batch) flow, I use client.responses.create(...) and pass both files using input_file objects along with the prompt — and this works perfectly.
Now I want to batch this logic using the /v1/responses batch endpoint. My batch requests look like this:
{
“custom_id”: “some-id”,
“method”: “POST”,
“url”: “/v1/responses”,
“body”: {
“model”: “gpt-4.1”,
“input”: [
{
“type”: “input_text”,
“text”: “Prompt with instructions”
},
{
“type”: “input_file”,
“file_id”: “file-abc123”
},
{
“type”: “input_file”,
“file_id”: “file-xyz456”
}
]
}
}
But I receive the following error:
Invalid value: ‘text’. Supported values are: ‘code_interpreter_call’, ‘function_call’, etc.
Some threads and helpers suggest file IDs might not be supported in batch mode and that file content must be inlined, but I haven’t found this confirmed in the official documentation. Could someone from the OpenAI team confirm:
- Are file uploads supported in batch requests to
/v1/responses? - Is there any limit to prompt length or total request size per line in the batch?
Thanks in advance!