#o1 preview response missing

1 messages · Page 1 of 1 (latest)

winter kettle
#

I'm not getting a response from o1 preview models from within my app. The generation details on OR activity dashboard shows a 200 status from openAI but the response isn't there (see 0 tokens_completion below). Finish reason is "error". I am getting a successful response from other models from within my app and I am even able to get a successful response from o1 for the same prompt via postman.

Any idea what's happening here? 😅

{
  "id": ******,
  "generation_id": "******",
  "provider_name": "OpenAI",
  "app_id": ******,
  "streamed": true,
  "cancelled": false,
  "generation_time": 0,
  "latency": 853,
  "moderation_latency": 691,
  "created_at": "2024-12-18T19:34:57.180394+00:00",
  "tokens_prompt": 1698,
  "tokens_completion": 0,
  "native_tokens_prompt": 1698,
  "native_tokens_completion": 0,
  "native_tokens_reasoning": 0,
  "num_media_prompt": null,
  "num_media_completion": null,
  "origin": "******",
  "usage": 0,
  "finish_reason": "error",
  "usage_cache": null,
  "usage_data": 0,
  "provider_responses": [
    {
      "status": 200,
      "latency": 853,
      "provider_name": "OpenAI"
    }
  ],
  "app": {
    "id": 188514,
    "created_at": "2024-09-10T14:02:23.781907+00:00",
    "description": null,
    "title": "******",
    "main_url": null,
    "origin_url": "******",
    "source_code_url": null
  }
}
sterile zephyr
#

This is an error response, see here ->

"finish_reason": "error"
winter kettle
#

Yes, saw that. But any idea why? Similar api call through postman to the same model is getting a response

sterile zephyr
winter kettle
#

Found the issue. Writing here for others' reference - OpenAI was flagging my input. The input itself was related to web design, nothing that violates their policies BUT the prompt also asked it to explicitly output its thoughts in a thoughts section. Once I removed that, it worked lol. I guess there's no need to force CoT on o1 models as it does it internally anyway.

magic echo