Does anyone know how to apply structured output when using batch API posting on \v1\responses, with gpt-5?
Welcome to the dev community @Hijkstuv
Here’s an example of a one line JSON for structured outputs over Responses API using the Batch API:
{"custom_id": "request001", "method": "POST", "url": "/v1/responses", "body": { "model": "gpt-5", "input": [ { "role": "system", "content": "You are an expert at structured data extraction. You will be given unstructured text from a research paper and should convert it into the given structure." }, { "role": "user", "content": "..." } ], "text": { "format": { "type": "json_schema", "name": "research_paper_extraction", "schema": {"type": "object","properties": {"title": { "type": "string" },"authors": { "type": "array", "items": { "type": "string" }},"abstract": { "type": "string" },"keywords": { "type": "array", "items": { "type": "string" }}},"required": ["title", "authors", "abstract", "keywords"],"additionalProperties": false }, "strict": true } }}}
This was put together simply by placing the body of the cURL request in the Structured output examples from docs into the body of a Batch API input request for the Responses API.
2 Likes
Here are some data employing the above line, prepared in Python:
import json
line = r'''that_line_copied'''
request_dict = json.loads(line)
pretty_json = json.dumps(request_dict, indent=2)
compact_line = json.dumps(request_dict, separators=(',', ':'))
print(f"{pretty_json}\n\n{compact_line}")
The request_dict.body is like you might already use in a normal request that serializes Python objects to a JSON-formatted string request body:
openai_client.responses.create(**request_dict["body"])
pretty_json is so we can see what is being communicated on the forum.
compact_line is a string to add to a JSONL, but without excess bytes.
{
"custom_id": "request001",
"method": "POST",
"url": "/v1/responses",
"body": {
"model": "gpt-5",
"input": [
{
"role": "system",
"content": "You are an expert at structured data extraction. You will be given unstructured text from a research paper and should convert it into the given structure."
},
{
"role": "user",
"content": "..."
}
],
"text": {
"format": {
"type": "json_schema",
"name": "research_paper_extraction",
"schema": {
"type": "object",
"properties": {
"title": {
"type": "string"
},
"authors": {
"type": "array",
"items": {
"type": "string"
}
},
"abstract": {
"type": "string"
},
"keywords": {
"type": "array",
"items": {
"type": "string"
}
}
},
"required": [
"title",
"authors",
"abstract",
"keywords"
],
"additionalProperties": false
},
"strict": true
}
}
}
}
{"custom_id":"request001","method":"POST","url":"/v1/responses","body":{"model":"gpt-5","input":[{"role":"system","content":"You are an expert at structured data extraction. You will be given unstructured text from a research paper and should convert it into the given structure."},{"role":"user","content":"..."}],"text":{"format":{"type":"json_schema","name":"research_paper_extraction","schema":{"type":"object","properties":{"title":{"type":"string"},"authors":{"type":"array","items":{"type":"string"}},"abstract":{"type":"string"},"keywords":{"type":"array","items":{"type":"string"}}},"required":["title","authors","abstract","keywords"],"additionalProperties":false},"strict":true}}}}
To ensure success, I’m going to directly consume from the compact_line meant for a batch:
import os, httpx, json
try:
with httpx.Client(timeout=1000) as client:
resp = client.post(
"https://api.openai.com/" + json.loads(compact_line).get('url'),
headers={
"Authorization": f"Bearer {os.getenv('OPENAI_API_KEY')}",
"Content-Type": "application/json",
"Accept": "application/json",
},
content=json.dumps(json.loads(compact_line).get('body')),
)
resp.raise_for_status()
except httpx.HTTPStatusError as e:
print(f"Request failed: {e}")
if e.response is not None:
try:
# print body error messages from OpenAI
print("Error response body:\n", e.response.text)
except Exception:
pass
raise
except httpx.RequestError as e:
print(f"Request error: {e}")
raise
<Response [200 OK]>
We see below the text.format with "type": "json_schema" echoed back at you in the large response object (that comes pretty-formatted), along with an AI forced to use that schema even though there is nothing to “say” about the user’s ellipsis input.
print(resp.text)
{
"id": "resp_1234",
"object": "response",
"created_at": 1760179057,
"status": "completed",
"background": false,
"billing": {
"payer": "openai"
},
"error": null,
"incomplete_details": null,
"instructions": null,
"max_output_tokens": null,
"max_tool_calls": null,
"model": "gpt-5-2025-08-07",
"output": [
{
"id": "rs_1234",
"type": "reasoning",
"summary": []
},
{
"id": "msg_1234",
"type": "message",
"status": "completed",
"content": [
{
"type": "output_text",
"annotations": [],
"logprobs": [],
"text": "{\"title\":\"\",\"authors\":[],\"abstract\":\"\",\"keywords\":[]}"
}
],
"role": "assistant"
}
],
"parallel_tool_calls": true,
"previous_response_id": null,
"prompt_cache_key": null,
"reasoning": {
"effort": "medium",
"summary": null
},
"safety_identifier": null,
"service_tier": "default",
"store": true,
"temperature": 1.0,
"text": {
"format": {
"type": "json_schema",
"description": null,
"name": "research_paper_extraction",
"schema": {
"type": "object",
"properties": {
"title": {
"type": "string"
},
"authors": {
"type": "array",
"items": {
"type": "string"
}
},
"abstract": {
"type": "string"
},
"keywords": {
"type": "array",
"items": {
"type": "string"
}
}
},
"required": [
"title",
"authors",
"abstract",
"keywords"
],
"additionalProperties": false
},
"strict": true
},
"verbosity": "medium"
},
"tool_choice": "auto",
"tools": [],
"top_logprobs": 0,
"top_p": 1.0,
"truncation": "disabled",
"usage": {
"input_tokens": 98,
"input_tokens_details": {
"cached_tokens": 0
},
"output_tokens": 536,
"output_tokens_details": {
"reasoning_tokens": 512
},
"total_tokens": 634
},
"user": null,
"metadata": {}
}
Verifying @sps gave us a working line (just with a few extra spaces). So the answer to the question is that, “Yes, user sps knows”.
2 Likes