chapman-sip 0.1.0

A CLI tool to run steerable models on the Chapman cluster with SSH port forwarding

Chapman-Sip CLI

A pip-installable Python command-line utility to connect to the Chapman cluster (dgx0.chapman.edu), pull the steerable-model-runner repository, find an open port, launch the model server, and tunnel the port back to your local machine.

Features

• Automated Host Setup: Checks if the remote repository is cloned at ~/.steerable-models on the cluster, cloning or pulling changes automatically.
• Dynamic Port Hunting: Generates a random port and verifies it is not currently bound on the cluster before initiating.
• SSH Port Forwarding: Automatically forwards the selected remote port to localhost:[PORT] locally.
• Interactive Log Streaming: Streams stderr and stdout from the model serving script in real-time.
• Safe Shutdown: Automatically cleans up the remote process, closes the port-forwarding server, and closes SSH channels on Ctrl+C.

Installation

Install in editable mode for development or directly from the source directory:

pip install -e .

Usage

Simply run:

sip

You will be prompted for:

• SSH username
• SSH password
• Hugging Face API token (for model download)
• Hugging Face repository name (e.g., the model repository to run)

Once connected, the utility will output the endpoint URL, e.g.:

https://localhost:[PORT]/v1/

Keep the CLI running while you query the endpoint. Press Ctrl+C to terminate the model server and close the tunnel.