o1 image input doesn't work with custom instructions #6376
Unanswered
andresgit
asked this question in
Troubleshooting
Replies: 2 comments 1 reply
-
LLM issue? Provider/proxy issue? Can't reproduce it. Logs: {
"baseURL": "https://api.openai.com/v1",
"level": "debug",
"message": "[OpenAIClient] chatCompletion",
"modelOptions": {
"messages": [
{
"content": [
{
"text": "what do you see",
"type": "text"
},
{
"image_url": {
"detail": "auto",
"url": "data:image/png;base64,..."
},
"type": "image_url"
}
],
"role": "user"
}
],
"model": "o1",
"stream": true,
"user": "USER_ID"
},
"timestamp": "2025-03-17T20:44:50.300Z"
}
{"level":"debug","message":"Making request to https://api.openai.com/v1/chat/completions","timestamp":"2025-03-17T20:44:50.302Z"},
{
"choices": [
{
"finish_reason": "stop",
"index": 0,
"logprobs": null,
"message": {
"annotations": [],
"content": "It appears to be a stylized black-and-white logo made up of interlocking geometric shapes arranged in a circular pattern. The design resembles the well-known symbol often associated with OpenAI and ChatGPT.",
"refusal": null,
"role": "assistant"
}
}
],
"created": 1742244290,
"id": "chatcmpl-BCBcIPCD733e7UGS0CGhT4nN3FKoA",
"level": "debug",
"message": "[OpenAIClient] chatCompletion response",
"model": "o1-2024-12-17",
"object": "chat.completion",
"service_tier": "default",
"system_fingerprint": "fp_60442b0df0",
"timestamp": "2025-03-17T20:45:03.080Z",
"usage": {
"completion_tokens": 819,
"completion_tokens_details": {
"accepted_prediction_tokens": 0,
"audio_tokens": 0,
"reasoning_tokens": 768,
"rejected_prediction_tokens": 0
},
"prompt_tokens": 265,
"prompt_tokens_details": {
"audio_tokens": 0,
"cached_tokens": 0
},
"total_tokens": 1084
}
} |
Beta Was this translation helpful? Give feedback.
1 reply
-
It doesn't work if you use custom instructions, like in the first post screenshot. Without custom instructions it does work. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
What happened?
Using o1 with custom instructions and image input the model output ignores the image as if it doesn't see it, see the image below. But if I don't use custom instructions, it works. With 4o or 4.5 for example it works fine with custom instructions.
I think it is because with o1 for some reason it doesn't use the system message for custom instructions, but starts editing the actual user message and that somehow removes the image. With 4o or 4.5 it uses the system message for instructions, maybe o1 can also use the system message for instructions?
I am using fileStrategy: "firebase"
Version Information
ghcr.io/danny-avila/librechat-dev:latest@sha256:f39eca6a09e326235c3acc1283c114764bf0f3c8235e9fb4033d9e8f583a69ed
Steps to Reproduce
See above
What browsers are you seeing the problem on?
No response
Relevant log output
Screenshots
Code of Conduct
Beta Was this translation helpful? Give feedback.
All reactions