Hosted Inference Credits rising beyond 1 expected usage counter for 1 image test

We are having issues with the Usage counts rising more than a 1 when testing a model and only uploading 1 image on the Model test screen.

Note that my usage count was at 1,008 before I started these steps.

  1. I revoked my API key just to be sure this was clear.
  2. I went to the Model page and uploaded/dragged/dropped a 1 page PNG image to look for barcodes in my images.
  3. I revoked the API key again right away, to be sure nothing else would call our service after I submitted the above image.
  4. When I go to the Usage counts, it went up slowly to 1,009, then up to 1,017 - every 5-10 minutes until it finally rose to 1,017.

Shouldn’t my one PNG upload only increment the counter to from 1,008 to 1,009 ?

I reached out to support 3 days ago, because my account was adding 1 usage count every 5 minutes for hours over night (even when the API key was deleted.) They seemed to have fix that problem yesterday - but the have never contacted me. I reset my account password to a very strong password so I know someone isn’t online dropping images in the Model testing area.

Any feedback on why this might be happening would be great

1 Like

Hi @tarcham33 - we had issues with inference tracking at the beginning of the month; it seems like the issues you are seeing are related.

We’re working on a long-term fix.


Thanks Jacob. Our trial ends in 3 days, is it possible to extend this until its resolved because i want to keep our dataset private. I didn’t want to go live with the paid plan until I know we won’t have our usage counts incremented in error.

Happy to help; I’ll reach out over DMs to confirm some info.

Hi @tarcham33,

I want to let you know that we’ve resolved the issues around our hosted inference API tracking. The number showing in your workspace should now be correct.

Our product team is working on controls to ensure we don’t experience the issue again, though please let me know if you see anything out of the ordinary.

We’re really sorry about the inconvenience here, and appreciate you for building with Roboflow. Give us a shout if we can help with anything.