bug: App shell conflicts between `cortex-nightly` and `cortex-stable`
Cortex version
v172
Describe the Bug
App shell conflicts between cortex-nightly and cortex-stable:
when I load a model on one version, the model is loaded on both cortex and cortex-nightly
when I unload the model on one version, the model is unloaded on both cortex and cortex-nightly
Not sure if this is a cortex ps issue or installation/data folder conflict with cortex and cortex-nightly
Steps to Reproduce
No response
Screenshots / Logs
What is your OS?
- [X] MacOS
- [ ] Windows
- [ ] Linux
What engine are you running?
- [X] cortex.llamacpp (default)
- [ ] cortex.tensorrt-llm (Nvidia GPUs)
- [ ] cortex.onnx (NPUs, DirectML)
Our CLI and Server communicate by TCP IP socket for now and the Server does not know about cortex variant, so it is expected.
This will be solved after we support Unix domain socket
cc: @namchuai @nguyenhoangthuan99
@vansangpfiev do we have plans to move to UDS, and should this be done earlier rather than later?
cc @dan-homebrew @0xSage
Decision:
- move Bug to icebox (Sprints: 2 sprints away)
- create new Epic on moving to Unix Domain Sockets (Size: Large, Status: Planning)
Dumb/nonurgent question: If we switch to UDS, what about devs directly embedding Cortex in their apps and only using the API? Will we still support TCP/IP (the most common protocol)?
IMO, we will support both UDS and TCP/IP. Clients (CLI, Jan, ...) can switch to which protocol they want.
@gabrielle-ong A temporary solution for TCP/IP is to use different ports by configuring apiServerPort in .cortexrc*.
Thanks @vansangpfiev, Ill use that workaround iceboxing this bug and creating a planning ticket #1517