go-dls

Local NVIDIA vGPU license server. Runs fully offline — no internet connection or purchased license required.

Driver Compatibility

Driver Version	Support
17.x (535.x)	go-dls only
18.x (570.x)	go-dls + gridd-unlock-patcher

Usage

1. Run the license server

Run on any host reachable from your guest VM (Proxmox host, separate VM, or inside the guest itself).

docker run -d \
  -e DLS_URL=192.168.1.x \
  -p 443:443 \
  -v dls-cert:/app/cert \
  -v dls-db:/app/db \
  injundev/go-dls

Set DLS_URL to the IP or hostname that your guest VM can reach.

To run inside the guest VM itself:

docker run -d \
  -e DLS_URL=127.0.0.1 \
  -p 443:443 \
  -v dls-cert:/app/cert \
  -v dls-db:/app/db \
  injundev/go-dls

2. Install the client token

Run inside the guest VM after installing the vGPU guest driver.

mkdir -p /etc/nvidia/ClientConfigToken
curl -sk https://<DLS_URL>/-/client-token \
  -o /etc/nvidia/ClientConfigToken/client_configuration_token.tok
systemctl restart nvidia-gridd

3. Verify

nvidia-smi -q | grep "License"

Licensed confirms the license is active.

Management UI

Visit https://<DLS_URL>/-/manage to view connected VMs and active leases.

Environment Variables

Variable	Default	Description
DLS_URL	localhost	IP or hostname reachable from guest VMs (embedded in .tok)
DLS_PORT	443	HTTPS listen port
CERT_DIR	/app/cert	TLS certificate directory
DB_DSN	/app/db/db.sqlite	SQLite database path
LEASE_EXPIRE_DAYS	90	Lease validity in days
LEASE_RENEWAL_PERIOD	0.15	Renewal threshold (fraction of lease duration remaining)
TOKEN_EXPIRE_DAYS	1	Access token validity in days

Build

docker build -t go-dls .