For example, when I am attaching an image file to my request “Translate File” it wants to call function `extract_text_from_image` or `translateText`. We started to experience this behavior very recently, it used to work fine before. Expected behavior: it should not try to call extract_text_from_image/translateText functions, but should recognize text on images using model internal multi modal capabilities. It is not reproducible all the time, but happens now in 50-60% percent of cases.

Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.