Analyze image from open ai storage

API
,
1

How can I analyze an image for availability if it is in open ai storage for the purpose of vision?
I’m trying to do this in java using the com library.theokanning.openai

1 Like
2

The only endpoint that can perform vision analysis on files uploaded to OpenAI storage, needing purpose:vision set on them, is assistants. A user file attachment within the message contents can be an image ID.

Chat completions, on the other hand, furnishes either base64 upload when sending a user message as a block of content, or you can specify an internet URL to be retrieved for image.

You can see if the tedious use of assistants, more for maintaining a conversation server-side than for specific vision features, would meet your need, or you can rethink why you need the additional step of uploading and waiting on a file ID to be returned when you could have already sent the image in an API language request to chat completions.

1 Like
3

Right now it works like this that I first upload the image to the open ai storage, and then I need to analyze it at the user’s request.

Will this solution from other similar topics help?

thread_message = client.beta.threads.messages.create(
    "thread_abc123",
    role="user",
    content=[
        {
            "type": "text",
            "text": "Here's an image example using an uploaded file."
        },
        {
            "type": "image file",
            "file_id": "file_abc123",
            "detail": "auto"  # optional, can be 'low', 'high', or 'auto'
        }
    ]
)
4

Yes, that is how, in Assistants, that you add a message to a thread that provides images for any vision analysis task the user may request.

Also a reminder that as long as you want that thread to still work, you must keep the image file in storage.