cllmhub-cli

README

cLLMHub CLI

The command-line interface for cLLMHub — publish local LLMs to the cLLMHub network.

What it does

• Publish models from any local inference backend to the hub so anyone with an API key can use them
• Connect external backends (Ollama, vLLM, LM Studio, llama.cpp, MLX) to the hub
• Chat completions with multimodal support (text + images) across all backends
• Authenticate via OAuth device flow, manage credentials

Quick start

# Install (pick one)
npm install -g cllmhub          # npm
brew install cllmhub/tap/cllmhub # Homebrew
curl -fsSL https://raw.githubusercontent.com/cllmhub/cllmhub-cli/main/install.sh | sh  # shell script

# Authenticate
cllmhub login

# Publish a model (interactive selection)
cllmhub publish

Installation

npm / npx

# Install globally
npm install -g cllmhub

# Or run without installing
npx cllmhub --help

The npm package automatically downloads the correct pre-built binary for your platform on install.

Homebrew

brew tap cllmhub/tap
brew install cllmhub

Pre-built binaries

Download from your hub's Settings > Downloads page, or grab from GitHub Releases. Available for:

Platform	Architecture
macOS	Apple Silicon (arm64), Intel (amd64)
Linux	x86_64 (amd64), ARM64
Windows	x86_64 (amd64)

Install script

curl -fsSL https://raw.githubusercontent.com/cllmhub/cllmhub-cli/main/install.sh | sh

Build from source

Requires Go 1.22+.

git clone https://github.com/cllmhub/cllmhub-cli.git
cd cllmhub-cli
make build
# Binary is at bin/cllmhub

Cross-compile for all platforms:

make build-all

Commands

Publishing

cllmhub publish

Publish models to the cLLMHub network. All publishing goes through the background daemon.

Use flags to specify a model and backend directly, or run without flags for interactive selection from detected backends.

# Direct publish
cllmhub publish -m llama3-70b -b ollama
cllmhub publish -m mixtral-8x7b -b vllm
cllmhub publish -m my-model -b mlx --api-key sk-xxx

# Interactive selection
cllmhub publish

Flags:
  --model,          -m   Model name to publish
  --backend,        -b   Backend type: ollama | vllm | lmstudio | llamacpp | mlx (default: ollama)
  --backend-url          Backend endpoint URL (overrides default for the backend type)
  --api-key              API key for the backend server
  --description,    -d   Model description
  --max-concurrent, -c   Maximum concurrent requests (0 = auto-detect, default: 0)

cllmhub unpublish [model...]

Stop serving one or more published models. Run without arguments to interactively select from currently published models.

# Unpublish by name
cllmhub unpublish llama3-70b

# Interactive selection
cllmhub unpublish

Daemon

The daemon runs in the background and manages model publishing bridges.

cllmhub start

Start the cLLMHub daemon.

cllmhub start

cllmhub stop

Stop the running cLLMHub daemon.

cllmhub status

Show daemon status, including PID, uptime, and currently published models.

cllmhub logs

Show daemon logs.

Flags:
  --follow, -f   Follow log output
  --lines,  -n   Number of lines to show (default: 50)

Account

cllmhub login

Authenticate with cLLMHub using OAuth 2.0 device flow. Opens a browser to complete authorization.

cllmhub whoami

Show the currently logged-in user.

cllmhub logout

Revoke credentials on the server and remove the local credentials file.

cllmhub update

Update the CLI to the latest version. The CLI also checks for updates automatically after each command.

Supported backends

Backend	Default endpoint	Notes
ollama	http://localhost:11434	Default backend, most common. Native chat API with image support
vllm	http://localhost:8000	High throughput, GPU optimized. OpenAI-compatible chat API
lmstudio	http://localhost:1234	Desktop app for running local LLMs. OpenAI-compatible chat API
llamacpp	http://localhost:8080	CPU-friendly, quantized models. OpenAI-compatible chat API
mlx	http://localhost:8080	Apple Silicon optimized via mlx-lm. OpenAI-compatible chat API

All backends support both text completions and chat completions (OpenAI-compatible /v1/chat/completions format). Multimodal messages with image content parts are supported — Ollama automatically converts OpenAI-format image parts to its native base64 image format.

License

Apache License 2.0 — see LICENSE.