How does one process images passed to open-webui's ui through pipelines?

[thistleknot]

I have an openai compatible api endpoint that handles text messages just fine, but I'm not sure how files uploaded as images are translated to be send to an openai api endpoint.

I've tried base64, but keep getting a 422 unprocessable entity. Looking at openai's sdk, it looks like they expect url's, but also read url key value can be a base64.

Was hoping to see pipeline code that accepts an inputted image from an upload sent to an openai compatible api endpoint within the pipeline script.