Computer Vision suddenly not working unattended

Help Studio
, ,
1

It was working fine both via Studio and on my machine as well as unattended on our QA server. Then all of a sudden during a login process it can’t find a simple button. I tried it via Studio on that server and it works fine, and “edit target” finds the button just fine. Resolution is the same on both machines, including when running unattended. I have tried two different robots on the same server, both have the same problem.

Is CV not reliable? What kind of things can cause this issue?

EDIT: more detail

The CV is used both in a custom activity and within the actual process. I opened the library in Studio and ran the login process and it worked fine. When I run the actual automation (attended or unattended) that uses the library, the login steps fail (both attended and unattended) because it doesn’t find a button.

2

Hello @postwick

  1. Verify consistent resolution settings between Studio and unattended robots.
  2. Ensure environment settings (display, font sizes) are the same.
  3. Double-check element selection methods.
  4. Confirm the application state is consistent.
  5. Adjust CV Scope settings (accuracy, timeout).
  6. Implement logging and debugging.
  7. Add retry mechanisms.
  8. Use the latest CV activity versions.
  9. Replicate the QA environment for testing.

Thanks & Cheers!!!

3

There is no CV Scope nor CV activities, this is the modern integrated CV selector. Resolution and everything else are the same, and as I said it worked fine for a while then suddenly won’t find any screen elements.

4

My organization has encountered an issue with Computer Vision that hinges on how the process was published. If you publish to Orchestrator through Studio, the Computer Vision works fine. If you publish using a DevOps pipeline, such as Azure, the Computer Vision will not work.

Manual publishing good, DevOps publishing bad… at least from our experience with ComputerVision.

5

It sounds like settings files aren’t being published when you do it with a DevOps pipeline.