Added model loading Feedback

[ParamThakkar123]

Fixes: #879

[codecov-commenter]

Codecov Report

❌ Patch coverage is 0% with 22 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
api/transformerlab/shared/shared.py	0.00%	22 Missing ⚠️

📢 Thoughts on this report? Let us know!

[deep1401]

Just realized I didnt migrate my review on here. This is what I had posted on API:

tail field in job data stored things correctly for me, but I think the initial startup of the subprocess is the one that takes time and there is no output in console then. Which is why you'll only see Starting model... for most plugins while only some whose actual process also takes time, those will show these things.

Maybe @dadmobile can help us resolve but I think this wont solve the problem right?

Also side note, we wouldn't need to add a new server logs route in API right?
I thought you were just reading from job data directly on the frontend already (THIS IS DONE as per a follow-up from @ParamThakkar123)

Just for reference this is what tail had in job data:

{
  "tail": [
    "2025-11-17 07:55:03 | INFO | model_worker | Loading the model ['Qwen3-0.6B'] on workerb617e619, worker type: MLX worker...",
    "2025-11-17 07:55:03 | INFO | model_worker | Model architecture: Qwen3ForCausalLM",
    "2025-11-17 07:55:04 | ERROR | stderr | \rFetching 7 files:   0%|          | 0/7 [00:00<?, ?it/s]",
    "2025-11-17 07:55:04 | ERROR | stderr | \rFetching 7 files: 100%|██████████| 7/7 [00:00<00:00, 30551.64it/s]",
    "2025-11-17 07:55:04 | ERROR | stderr |",
    "2025-11-17 07:55:05 | INFO | stdout | Context length:  40960"
  ]
}

[deep1401]

Also for another comment I think Ali and I had also worked on the cause of the "plugin taking too much time to start" problem earlier once, it takes a long time just for the python imports to work themselves out because they're in a venv and pycache may not be developed yet