chore: Remove model weights from container images

[YrrepNoj]

This PR is a proof of concept to how we can utilize Zarf Data Injections to push model weights to a PVC instead of directly building them into the Docker images themselves.

Pros:

• Much smaller image size (but similar Zarf Package sizes; with the exception of whisper which is now smaller as well)

Model Container	Original Size	New Size
llama-cpp-python	4.58 GB	0.21 GB
whisper	6 GB	5.86 GB
text-embeddings	10.4 GB	5.63 GB
vLLM	14.2 GB	9.62

• Faster build & deploy time
  • I did not quantify this bullet by testing the timings locally but I noticed significant speedups as I was performing other tests locally.
  • As a data-point, the e2e tests from this branch are running the model creation & deployment steps an average of 2 minutes faster than the tests branching from main. This is about a 30% speedup for this workflow step.

Cons:

• Adds a tiny bit of complexity to our deployment by introducing a new PVC. This seems pretty reasonable, but it is still worth noting that this is an added thing that delivery engineers would have to manage (for the storage class of the PVC).
• Adds a dependency on Python (and the dev optional-dependencies listed in the pyproject.toml being installed on the host machine during the zarf package create lifecycle.
  • This dependency is now introduced because we are now pulling down the model weights during the Zarf package creation time instead of pulling the model weights during the Docker container build time.

The base branch was changed.

[netlify]

✅ Deploy Preview for leapfrogai-docs ready!

Name	Link
🔨 Latest commit	742cf4c
🔍 Latest deploy log	https://app.netlify.com/sites/leapfrogai-docs/deploys/66aab5d55d38f80008e2bccd
😎 Deploy Preview	https://deploy-preview-786--leapfrogai-docs.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.
Lighthouse	1 paths audited Performance: 39 (🟢 up 2 from production) Accessibility: 98 (no change from production) Best Practices: 100 (no change from production) SEO: 92 (no change from production) PWA: - View the detailed breakdown and full score reports

---

To edit notification comments on pull requests, go to your Netlify site configuration.

[justinthelaw]

The Dockerfiles still contain a lot of extra ENVs and ARGs. I recommend removal and an update of the READMEs to add instructions for switching models and also any other mechanics that need to be explained to potential delivery engineers or users that use these.