Hello Fellow Automaters,
I'm running into an issue with calling the AI model (running a standard cats vs dogs detection using the Kaggle set).
When I open the file with the content to be detected with a manual trigger, it works fine
A very simple flow to open a file, and we see the contents of the file converted to base64, and passed on the AI Model. It runs without errors (except from detecting, need to add more samples ...)

Run the flow manually

The file is imported and the content is converted to base64, the content is passed on to the AI model

And succes!

However, when I tried to post this same image as base64 content through a HTTP requests, the model refuses to accept the input and fails. Imho it is exactly the same content ( opened the image and used base64encode.org to encode it).
This is the HTTP flow request

The HTTP POST request I send from Postman

The HTTP requests gets processed normally and base64 encoded image is stored in the Image attribute


However when passing the Image to the AI model, it fails with a bad request, even though the content of the image attribute is exactly the same as when I'm running it with a manual trigger.


What am I doing wrong? HELP!