[SUPPORT] Workbench Example Blueprint: PDF-to-Podcast

AI & Data Science NVIDIA AI Workbench
1

(The other support thread was accidently closed. Starting a new thread here.)

Great news! NVIDIA Blueprints are coming to AI Workbench!

This is the support thread for the PDF-to-Podcast Blueprint on GitHub. Any major updates we push to the project will be announced here. Further, feel free to discuss, raise issues, and ask for assistance in this thread.

Please keep discussion in this thread project-related. Any issues with the Workbench application should be raised as a standalone thread. Thanks!

3

I am trying to do this using a DGX Spark but get an error when composing the docker file which chatgpt indicated was an architecture mismatch error from my compose logs in AI workbench. will the DGX Spark ever be supported for this project?

4

I am also facing same issue.

5

so i did manage to get gemini 3.1 pro in gemini cli to help me launch it but it seemed extremely glitchy ( i think theres defiantely a way to be able to have them do this - im thinking of seeing if i can have my agent on my spark recompile it for us )

6

With the help of AI, I managed to get the Compose to build and run.

If you’re getting error messages specifically regarding trying to build pdf-api and celery-worker the following worked for me:

The following advice came from AI when I gave it my error messages:

Use tonistiigi/binfmt

  1. The tonistiigi/binfmt image is multi-arch and works on ARM64 hosts. It is the modern, maintained alternative recommended by Docker BuildKit.

docker run --privileged --rm tonistiigi/binfmt --install all

  1. Verify it works:

Test that AMD64 emulation is now registered on your ARM64 host

docker run --rm --platform linux/amd64 alpine uname -m

Expected output: x86_64