Hi,
First, my DGX Spark is running smoothly - thanks for nice setup experience.
I tried the playbook “RAG application in AI Workbench” and followed the step-by-step instructions provided via this link and both API keys are supplied but AI Workbench gave this container build error message. See screenshot below. Pls fix.
Environment → ProjectContainer → Scripts and edit preBuild.bash script and add the following:
pip install --upgrade pip
What will frustrate you next is the step that says, “launch the application”. (from your image, it clearly did not add the web app.)
“Navigate to Environment > Project Container > Apps > Chat and start the web application.”
Hi @conor.sands - Thanks for pointing out the 2nd issue.
Hi nVidia - Pls fix the 2nd issue… Step 5 of Playbook.
Thank you for pointing this out, we will look into improving the playbook
When you clone it, add the preBuild.script line.
Update your base image in the Project Container. (This may be what broke your web app in your screenshot)
go to Git > Changes, select the .project/spec.yaml file and discard changes. The app will show back up.
Modify your compose.yaml to ensure you have the correct NIM: nvcr.io/nim/meta/llama-3.1-8b-instruct-dgx-spark:latest
Once you launch the Chat app, navigate to the Models tab on the top right of the web app. You will be able to change the model to your local NIM. (Just change the tab from API Endpoints to Self-Hosted Endpoint.
Hope that works for you and you can start playing around with the prompts instead of abstraction layers.
I created a fork with these fixes:
and a pull request:
Hi all,
Apologies for the issue, just pushed a fix as described in the thread above. Simple pip version bump.
I recommend re-cloning your project repo down on the Spark system and trying again. Verified as working on my end, but please let me know if there are additional issues.
You can also follow along with any updates to this project repo in the support thread for this project here.
Thanks for the catch and for your patience.
conor, ed & edwli - thank you.
I just cloned in AI Workbench today and it is still not working for me. URLs do not get loaded and 403 errors come all the time.
