janhq
feat: Cortex model importing
Goal
- Cortex can import a model that is already on the user's computer
- Question: by symlink or cp?
- CLI:
cortex model import <path>, but how do we specify engine etc? (or scope to only GGUF?)
Original Post
Problem Even though I already have a GGUF model, I still need to re-pull the model, which seems not very friendly. Could Cortex add a feature to import models?
It would be even better if the models command could directly load a model file, although loading a model not listed in the model ID list might cause an error, but this way users can try using more models.
@cloudherder This is an important feature. We will be expanding this to solve for generic model importing in cortex.cpp.
@cloudherder This is an important feature. We will be expanding this to solve for generic model importing in cortex.cpp.
I am honored to receive your affirmative response. I will continue to monitor the development of cortex.cpp.
@namchuai @nguyenhoangthuan99 Are we doing this in Sprint 21? (I will move it to Sprint 22 otherwise)
