Bad Request 400 : Reasoning model issue while store is True

[ashish-auditorsdesk]

A BadRequestError (400) occurs because the request includes a reasoning item (rs_68b14140...) without its required subsequent item. This indicates the input structure is incomplete or incorrectly formatted from the backend in Responses API.

## Issue Output
---------------------------------------------------------------------------
BadRequestError                           Traceback (most recent call last)
Cell In[273], [line 1](vscode-notebook-cell:?execution_count=273&line=1)
----> [1](vscode-notebook-cell:?execution_count=273&line=1) response = client.responses.create(
      2                 model="gpt-5",
      3                 input=[
      4                     {
      5                         "role": "user",
      6                         "content": [{"type": "input_text", "text": "Not ab,e to download the file you have shared"}],
      7                     }
      8                 ],
      9                 previous_response_id="resp_68b140cf2c9c8193a6417aabd97428b40b003345362caed7"
     10 
     11             )

File ~/Library/Python/3.9/lib/python/site-packages/openai/resources/responses/responses.py:795, in Responses.create(self, background, include, input, instructions, max_output_tokens, max_tool_calls, metadata, model, parallel_tool_calls, previous_response_id, prompt, prompt_cache_key, reasoning, safety_identifier, service_tier, store, stream, stream_options, temperature, text, tool_choice, tools, top_logprobs, top_p, truncation, user, extra_headers, extra_query, extra_body, timeout)
    759 def create(
    760     self,
    761     *,
   (...)
    793     timeout: float | httpx.Timeout | None | NotGiven = NOT_GIVEN,
    794 ) -> Response | Stream[ResponseStreamEvent]:
--> [795](https://file+.vscode-resource.vscode-cdn.net/Users/ashishsapra/Downloads/~/Library/Python/3.9/lib/python/site-packages/openai/resources/responses/responses.py:795)     return self._post(
    796         "/responses",
    797         body=maybe_transform(
    798             {
    799                 "background": background,
    800                 "include": include,
    801                 "input": input,
    802                 "instructions": instructions,
    803                 "max_output_tokens": max_output_tokens,
    804                 "max_tool_calls": max_tool_calls,
    805                 "metadata": metadata,
    806                 "model": model,
    807                 "parallel_tool_calls": parallel_tool_calls,
    808                 "previous_response_id": previous_response_id,
    809                 "prompt": prompt,
    810                 "prompt_cache_key": prompt_cache_key,
    811                 "reasoning": reasoning,
    812                 "safety_identifier": safety_identifier,
    813                 "service_tier": service_tier,
    814                 "store": store,
    815                 "stream": stream,
    816                 "stream_options": stream_options,
    817                 "temperature": temperature,
    818                 "text": text,
    819                 "tool_choice": tool_choice,
    820                 "tools": tools,
    821                 "top_logprobs": top_logprobs,
    822                 "top_p": top_p,
    823                 "truncation": truncation,
    824                 "user": user,
    825             },
    826             response_create_params.ResponseCreateParamsStreaming
    827             if stream
    828             else response_create_params.ResponseCreateParamsNonStreaming,
    829         ),
    830         options=make_request_options(
    831             extra_headers=extra_headers, extra_query=extra_query, extra_body=extra_body, timeout=timeout
    832         ),
    833         cast_to=Response,
    834         stream=stream or False,
    835         stream_cls=Stream[ResponseStreamEvent],
    836     )

File ~/Library/Python/3.9/lib/python/site-packages/openai/_base_client.py:[1259](https://file+.vscode-resource.vscode-cdn.net/Users/ashishsapra/Downloads/~/Library/Python/3.9/lib/python/site-packages/openai/_base_client.py:1259), in SyncAPIClient.post(self, path, cast_to, body, options, files, stream, stream_cls)
   1245 def post(
   1246     self,
   1247     path: str,
   (...)
   1254     stream_cls: type[_StreamT] | None = None,
   1255 ) -> ResponseT | _StreamT:
   1256     opts = FinalRequestOptions.construct(
   1257         method="post", url=path, json_data=body, files=to_httpx_files(files), **options
   1258     )
-> 1259     return cast(ResponseT, self.request(cast_to, opts, stream=stream, stream_cls=stream_cls))

File ~/Library/Python/3.9/lib/python/site-packages/openai/_base_client.py:[1047](https://file+.vscode-resource.vscode-cdn.net/Users/ashishsapra/Downloads/~/Library/Python/3.9/lib/python/site-packages/openai/_base_client.py:1047), in SyncAPIClient.request(self, cast_to, options, stream, stream_cls)
   1044             err.response.read()
   1046         log.debug("Re-raising status error")
-> 1047         raise self._make_status_error_from_response(err.response) from None
   1049     break
   1051 assert response is not None, "could not resolve response (should never happen)"

BadRequestError: Error code: 400 - {'error': {'message': "Item 'rs_68b14140d5848193a4a6d51d06d2ff9d0b003345362caed7' of type 'reasoning' was provided without its required following item.", 'type': 'invalid_request_error', 'param': 'input', 'code': None}}

Debug information

• Python 3.12.3
• OpenAI 1.99.9

Repro steps

Use this code with any file with any prompt:

response = client.responses.create(
  model="gpt-5",
  input=[
    {
      "role": "user",
      "content": [
        {
          "type": "input_text",
          "text": "Give me items which are in books but not in 2B"
        }
      ]
    }
  ],
  text={
    "format": {
      "type": "text"
    },
    "verbosity": "medium"
  },
  reasoning={
    "effort": "medium"
  },
  tools=[
    {
      "type": "code_interpreter",
      "container": {
        "type": "auto",
        "file_ids": [
          "file-4vLY8h****",
          "file-3xAy2*****"
        ]
      }
    }
  ],
  store=True,
  include=[
    "code_interpreter_call.outputs",
    "reasoning.encrypted_content",
    "web_search_call.action.sources"
  ]
)

• Store the send the response_id as previous_response_id in next call, input_text it can be anything

response = client.responses.create(
                model="gpt-5",
                input=[
                    {
                        "role": "user",
                        "content": [{"type": "input_text", "text": "Not ab,e to download the file you have shared"}],
                    }
                ],
                previous_response_id="resp_68b1****"
            )

• This will trigger this error

Expected behavior

It should give explanation

[seratch]

Could you share complete repro steps? Just using previous_response_id does not cause this, so we need more details.

[ashish-auditorsdesk]

It will trigger this error anything if I use the previous response_id with code_interpreter.
Using excel files as code_interpreter's file, day before yesterday was working fine for me, did for many excel files. From yesterday i.e. 28 August. Giving me this error!

Let me know what more I can give you

[seratch]

I created a tiny script to see the same error but this simple one does not cause it (I know the code interpreter sandbox does not have the two uploaded files in this code snippet tho):

import logging

from openai import OpenAI

logging.basicConfig(level=logging.DEBUG)

client = OpenAI()

response = client.responses.create(
    model="gpt-5",
    input=[
        {
            "role": "user",
            "content": [
                {"type": "input_text", "text": "Give me items which are in books but not in 2B"}
            ],
        }
    ],
    text={"format": {"type": "text"}, "verbosity": "medium"},
    reasoning={"effort": "medium"},
    tools=[
        {
            "type": "code_interpreter",
            "container": {"type": "auto"},
        }
    ],
    store=True,
    include=[
        "code_interpreter_call.outputs",
        "reasoning.encrypted_content",
        "web_search_call.action.sources",
    ],
)

response = client.responses.create(
    model="gpt-5",
    input=[
        {
            "role": "user",
            "content": [
                {"type": "input_text", "text": "Not ab,e to download the file you have shared"}
            ],
        }
    ],
    previous_response_id=response.id,
)

One possible scenario would be your stored conversation is quite long and the older reasoning items than the allowed context window could be redacted under the hood. If that's the case, we do understand developers expect the stored conversation side does better job but for now you may need to optimize the conversation size to meet your business needs.

If the conversation is not long at all, I am still not sure about what's happening.

[ashish-auditorsdesk]

I've used the code that you have provided above and just add the files using this code:
{ "type": "code_interpreter", "container": { "type": "auto", "file_ids": ["file-N4kEyk*****wi","file-RFejp33t****2toudpH"], } }

but giving me this error, it's due to reasoning provided without type, pram and code

BadRequestError: Error code: 400 - {'error': {'message': "Item 'rs_68b184c2ed08819e93e54428ad7187af09822ebd2d57ed93' of type 'reasoning' was provided without its required following item.", 'type': 'invalid_request_error', 'param': 'input', 'code': None}}

{'id': 'rs_68b184c2ed08819e93******7187af09822ebd2d57ed93', 'summary': [], 'type': 'reasoning', 'content': None, 'encrypted_content': 'gAAAAA****************oJVn1NsXxQzFmMMP2rN8V6bcny3BsCaSIDihICfBPC03MRZ6kFEcFM=', 'status': None}

[ashish-auditorsdesk]

This error I got on the Playground also, it's same as we are getting in the python

Item 'rs_68b18af501bc8191977c1956754bc29905b05753f8f2b5d3' of type 'reasoning' was provided without its required following item.

[ashish-auditorsdesk]

@seratch Any update on the Bugs?

[hacastro22]

Exact same issue. It seems the problem happens when there's a reasoning item in the preceding message and previous_response_id is then used.

[Tadhgus]

I think this is the same issue as reported in issue 985.

For my own testing this can be reproduced by getting the agent to produce two or more consecutive CI calls, and the following up with an additional non-ci step. It seems the API returns references in one of the CI items to a RS item that is not returned from the API.

[github-actions]

This issue is stale because it has been open for 7 days with no activity.