Skip to main content

CLI Command Reference

CSGHub-Lite provides a rich set of command-line tools for users to manage and run Large Language Models (LLMs) directly in the terminal. Model names generally follow the format namespace/model_name, e.g., Qwen/Qwen3-0.6B-GGUF.

Basic Commands

CommandDescription
csghub-lite run <model>Pull model, start background service, and begin chat (fully automatic)
csghub-lite chat <model>Chat with a locally downloaded model
csghub-lite psList currently running models (loaded in memory) and their keep-alive duration
csghub-lite stop <model>Stop and unload the currently running model
csghub-lite serveManually start the API server (automatically handled by the run command)
csghub-lite pull <model>Download the specified model file from CSGHub
csghub-lite list / lsList all locally downloaded models
csghub-lite show <model>Show detailed information for a specified model (format, size, specific files)
csghub-lite rm <model>Remove specified local model files
csghub-lite loginSet the access token for the CSGHub platform (needed for private models)
csghub-lite search <query>Search for specified models on the CSGHub platform
csghub-lite uninstallCompletely remove csghub-lite, llama-server, and related data
csghub-lite --versionView current version information

Configuration Commands

CommandDescription
csghub-lite config set <key> <value>Set a specific configuration item
csghub-lite config get <key>Get the value of a specified configuration item
csghub-lite config showView all current configuration information

Difference Between run and chat

  • run — Highest degree of automation. If the model is not found locally, it will download it automatically; it also starts the background server and enters interactive chat.
  • chat — Specifically for chatting with models already downloaded. If you want to use a custom system prompt, you can use the --system parameter.
# Download, run, and chat (fully automatic)
csghub-lite run Qwen/Qwen3-0.6B-GGUF

# After exiting chat, the model stays in memory for instant restart next time
csghub-lite run Qwen/Qwen3-0.6B-GGUF

# Start a task with a custom system prompt for a downloaded model
csghub-lite chat Qwen/Qwen3-0.6B-GGUF --system "You are a professional coding assistant."