Linux, NVIDIA GPU (8GB+ VRAM), CUDA 12.1+, Docker, Python 3.11+.
git clone https://github.com/SuiMotus/phage.git
cd phage
bash install.shThe installer checks your GPU, installs the node daemon, and writes a default config.
git clone https://github.com/SuiMotus/phage.git
cd phage
pip install -e node/
phage-node init$ phage-node register
detected: NVIDIA RTX 3090 (24 GB)
driver: 565.57.01
coordinator: https://kellphage.com
registering... done
node_id: ph_n_a7f3e2c1
gpu: RTX 3090 (24 GB)
status: registered
certs written to ~/.phage/certs/
start with: phage-node start
$ phage-node startLaunches vLLM, connects to the coordinator, starts pulling tasks.
Background: phage-node start --daemon
sudo cp contrib/phage-node.service /etc/systemd/system/
sudo systemctl enable --now phage-node~/.phage/config.toml
[coordinator]
url = "https://kellphage.com"
[gpu]
device = 0
max_vram_pct = 85
[sandbox]
runtime = "gvisor"
max_runtime_sec = 300
[vllm]
port = 8400
gpu_memory_utilization = 0.80
max_model_len = 8192
[heartbeat]
interval_sec = 10Multi-GPU: run one daemon per card with separate configs and CUDA_VISIBLE_DEVICES.
See config.example.toml for all options.
$ phage-node status
phage-node status
========================================
node_id: ph_n_a7f3e2c1
registered: 2026-06-18T08:14:00Z
gpu: NVIDIA RTX 3090 (24 GB)
driver: 565.57.01
coordinator: https://kellphage.com
tasks: 37 done, 2 failed
tokens: 148200
last: 2026-06-24T22:41:00Z