Getting sporadic invalid URLs using structured outputs API

API Bugs
21

Ignoring what has been shared for a moment, if the request fails because it is routed to the wrong, non-existent, endpoint would you expect that non-existent endpoint, to still return a response_id? Could what was shared previously have been the follow on request if they implement retries? I could provide the successful response_id after the retry, but we currently do not get a response_id for the request that sees this error.

1 Like
22

I got the same error.
response.id: resp_6852d46e4e18819abb9bb1dfc48d868d0d81a5c4aa0baa8a
req_b4edbf4bb3e7b121c239a1091706d275
response._request_id: req_b4edbf4bb3e7b121c239a1091706d275

1 Like
23

We’re having the same issue, multiple times a day.
@OpenAI_Support request_id: req_6d54c93e7bb8602bf7cdaf0b006373ac
Full error log: 2025-06-18 16:13:44,531 - openai.agents - ERROR - Error getting response: Error code: 404 - {'error': {'message': 'Invalid URL (POST /v1/engines/gpt-4.1-mini-2025-04-14/completions)', 'type': 'invalid_request_error', 'param': None, 'code': None}}. (request_id: req_6d54c93e7bb8602bf7cdaf0b006373ac)
Edit: btw, we’re using openai-agents 0.0.17 and Python 3.13.3.

24

Getting the same error randomly.

25

This error often occurs when you send an image via URL to the OpenAI API. You’ll encounter it soon enough—go ahead and try it.

26

I’m also getting similar errors, but I’m seeing this error for multiple requests. Below is a sample request/response. For this request I’m using Structured Outputs and also linking to a previous_response_id. One of the previous requests in the “same chain of messages” includes images with the prompt.

Request: POST /v1/responses Body:

{
  "model" : "gpt-4.1-mini",
  "temperature" : 0.1,
  "previous_response_id" : "resp_685463d97084819aa8ddb56fe4f562da047c9e6054a143f7",
  "max_output_tokens" : 500,
  "instructions" : "You are an expert at classifying products into which category they fit best. Use the product title, description, features, attributes, and sample images of the product to make your decision. Do not add new special characters. Assign a confidence score between 0-10 on your choice, 10 if you are very confident in your decision and 0 if you are not confident.",
  "input" : [ {
    "role" : "user",
    "content" : [ {
      "type" : "input_text",
      "text" : "For the same product tell me which of the following categories best describes it. Choose from [Toy Action Figures, Action Figure Playsets, Collectible Figures, Animal Figures]"
    } ]
  } ],
  "text" : {
    "format" : {
      "type" : "json_schema",
      "name" : "CategoryResponse",
      "schema" : {
        "type" : "object",
        "properties" : {
          "confidence" : {
            "type" : "integer"
          },
          "choice" : {
            "type" : "string"
          }
        },
        "required" : [ "choice", "confidence" ],
        "additionalProperties" : false
      },
      "strict" : false
    }
  }
}

Response:

Response status: 404 NOT_FOUND Body:
{
  "error": {
    "message": "Invalid URL (POST /v1/engines/gpt-4.1-mini-2025-04-14/completions)",
    "type": "invalid_request_error",
    "param": null,
    "code": null
  }
}
2 Likes
27

OK it’s happening again:

ERROR:query_parser:Error parsing prompt '
I am going to give you a furniture e-commerce product.

You tell me which of the listed room it belongs to. Or if ambiguous, could fit in multiple rooms, or unclear, return 'No Room Fits'

Describe your reasoning. Default to 'No Room Fits' unless given compelling evidence

Product Name: octagon senoia ombre braided cotton beige area rug
Description: a cheerful ombre braided rug will highlight any room in your house , beautifully . soft cotton chenille . a rug pad is recommended , as these rugs do not have a non-slip backing . vacuum/spot clean .
        ': Error code: 404 - {'error': {'message': 'Invalid URL (POST /v1/engines/gpt-4.1-nano/completions)', 'type': 'invalid_request_error', 'param': None, 'code': None}}
28

I’m not sure how I get a response ID in this situation, but here’s are some request IDs:

req_987f4cb947dc48c6a785da22e5639762
req_ebe1d94b259fa7d99d9aa5a2055d19f9

29

Hello! We're actively working on fixing this issue. We'll be sure to update you as soon as the fix is live.

2 Likes
30

Hello! We've published a fix to this issue. We will mark this thread as resolved.

3 Likes