Same problem appears today for me with my personal text generation model.
- Originally, I used old (2.8.1) version of the @huggingface/inference JS package - “Error fetching from Hugging Face API: An error occurred while fetching the blob” while trying to send text input.
- Updated the package to the last version (3.12.1) - the error have changed - “No Inference Provider available for model …”
- Turn back to the old version, tried to specified the blob format (as this solution earlier worked with whisper audio during this API hell of the month) - “Unexpected token ‘N’, “Not Found” is not valid JSON”
- Tried to send raw HTTP request - Error 404: Model not found.
Yeah, something is going very wrong since 15 of April. Probably, the only solutions are just wait or switch to a different platform (which is not the case for my friends-only model experiments, so I just believe)