feat(oci): support OCI images and Ollama models

[mudler]

Description

This PR fixes #2527
This PR fixes #1028

Now models URL can be specified also as oci:// or ollama:// prefix. Note that when oci:// is used, it will default to unpack the image content to the model path, so the container have to be a naked one (e.g. a FROM scratch built with docker). When using ollama://, it can be used as for e.g. ollama://llama3 or ollama://gemma:2b and so on so forth.

My plan is to further make the installation more flexible in a way that local-ai models install ... can be used directly with URL/oci/ollama.

How it works

Models URLs can be specified as usual, in the YAML config file for example:

parameters:
  model: ollama://...

or directly start the ollama model:

local-ai run ollama://gemma:2b

Alternatively, the model can be pre-populated with local-ai models install too:

local-ai models install ollama://gemma:2b

For OCI images, same applies but the prefix is oci://. To build a container image which works in this way, you must have a Dockerfile like this:

FROM scratch
COPY ./my_gguf_file.gguf /

Push the image in some registry and reference it.

Notes for Reviewers

It also includes a small nuances in fixing watcher message that displays always as an error 🤷

Signed commits

• Yes, I signed my commits.

[netlify]

✅ Deploy Preview for localai ready!

Name	Link
🔨 Latest commit	8952511
🔍 Latest deploy log	https://app.netlify.com/sites/localai/deploys/6675e6bf725617000820224b
😎 Deploy Preview	https://deploy-preview-2628--localai.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

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