Where and how are the secrets stored in AI Workbench?

The secrets in the AI Workbench are stored locally, if I understand it correctly. I would like to know, how (encrypted or cleartext) and where (/home or elsewhere) these credentials are stored. Thanks for any information. I also wonder, how this works, when we use a remote cloud instead of our local machine.

Please tick the appropriate box to help us categorize your post
Bug or Error
Feature Request
Documentation Issue
Other

Hi, thanks for reaching out.

Yes, secrets are managed automatically by AI Workbench and are stored locally to the system that the project is running on, agnostic of local or remote location. You can typically find them under the ~/.nvwb/project-runtime-info/<project-name-and-hash> directory. Hope this helps!

Best,
Edward

1 Like

Thank you, Edward. Appreciate the quick answer. I always am struggling a bit with the whole secrets management. Will check out my options. I like the AI Workbench and will get around with it. Have a good time.

Current docs are here: Environment Variables — NVIDIA AI Workbench User Guide

Please let us know if these aren’t sufficiently clear.

Also, it would help to know what you don’t like about the current secrets management.

What is the blocker for you?

1 Like

Hi Tyler. Thanks for linking to the documentation. I am pretty careful, because I need to share my machine with other people and I had some pretty bad experience with .env files in the past. After reading the doc, I am ready to go through the onboarding project and then get more experience and ease in using the tool that seems to be a blessing for a dev. I need to grasp the concept by using the tool first. Thanks for your patience and support. Kind regards, Markus.

Hi again. I almost feel the need to apologize for my question. RTFM. All good now and I have finally got it, how convenient and practical the whole application is. I will make sure to rotate my secrets regularly and keep my stuff secure, knowing that the secrets will never leave my machine. Thanks again.

all good. it’s a new approach and the details aren’t well explained by us.