epic: Cortex has a better CLI UI for downloading

[dan-menlo]

Goal

• We should have a better CLI UI for downloading progress
• We are currently using the libcurl default, which may be confusing to users used to Github or Docker
• We need to think of this from first principles (i.e. does not map onto Github, Docker or even Ollama's current UX may not be correct)

Task list

• planning: Cortex CLI should be a layer on top of the API Server #1492

Inspirations

Github has a status indicator (for objects)

• Models tend to not have many objects, just several large ones
  

Docker pulls layers (shows progress while layer is downloading)

• Models do not have "layers"
  

Ollama has a progress bar

• Similar to Docker, they use layers (vs. our approach with normal GGUF files)
  

[vansangpfiev]

@dan-homebrew I would like to propose a new CLI UI for downloading as below:

Cortex will show the file names, download progress (both percentage and downloaded/total bytes, time remaining).
How do you think?

[dan-menlo]

@vansangpfiev This looks good and a clear improvement. Questions:

• Will it download each file in parallel or sequential?
• Is there a CLI Library that we can use for this?

Thank you for optimizing for UX.

[vansangpfiev]

@vansangpfiev This looks good and a clear improvement. Questions:

• Will it download each file in parallel or sequential?
• Is there a CLI Library that we can use for this?

Thank you for optimizing for UX.

@dan-homebrew

• Basically, CLI requests the Server to download a model and open a websocket to pull the progress.
  Server supports download files in parallel (thanks @namchuai)
• For progress indicators, I think can use https://github.com/p-ranav/indicators (MIT License)

[vansangpfiev]

Unicode char set does not work well on Windows, so we need to simplify the UI a little bit

This progress bar is similar to wget

[gabrielle-ong]

Thanks Sang! Marking as complete

QA: ✅ v1.0.1-218
Mac:

Windows:

Linux:

[dan-menlo]

@gabrielle-ong @vansangpfiev Can you link the PR for this issue, for better discoverability in future