Frameworks

swm includes declarative framework definitions for common AI tools. Each framework is installed with swm setup install and managed with start/stop.

Built-in frameworks

Framework	Slug	Category	Default port
vLLM	vllm	LLM inference	8000
Open WebUI	open-webui	LLM chat UI	8080
Ollama	ollama	LLM engine	11434
ComfyUI	comfyui	Image/video gen	8188
SwarmUI	swarmui	Generation UI	7801
Axolotl	axolotl	LLM fine-tuning	—
H2O LLM Studio	llm-studio	No-code fine-tuning	10101

Usage

swm setup list                          # see all frameworks
swm setup install vllm runpod:abc123    # install
swm setup start vllm runpod:abc123      # start (background)
swm setup stop vllm runpod:abc123       # stop

Auto-detection

• Tensor parallelism: vLLM auto-detects GPU count and sets --tensor-parallel-size
• SSH tunnels: If a port isn’t externally mapped, swm opens a background SSH tunnel
• Health probing: After starting, swm probes the health endpoint with retries

Custom frameworks

For tools not in the built-in list, use swm run to install and manage manually:

swm run runpod:abc123 "pip install my-framework"
swm run runpod:abc123 "nohup my-server --port 8000 > /workspace/server.log 2>&1 &"