[Unit]
Description=llama-server instance %i
After=network.target

[Service]
Type=simple
User=llamalink
WorkingDirectory=/opt/llamalink
ExecStart=/usr/local/bin/llama-server \
    --model /opt/llamalink/models/%i.gguf \
    --alias %i \
    --host 127.0.0.1 \
    --port 8080 \
    --ctx-size 8192 \
    --n-gpu-layers auto \
    --parallel 4 \
    --rope-scaling linear
Restart=on-failure
RestartSec=5
StandardOutput=journal
StandardError=journal
SyslogIdentifier=llama-server-%i

Environment="CUDA_VISIBLE_DEVICES=0"

# Security hardening
NoNewPrivileges=true
PrivateTmp=true
ProtectSystem=strict
ProtectHome=true
ReadOnlyPaths=/opt/llamalink/models
ReadWritePaths=/opt/llamalink/data

[Install]
WantedBy=multi-user.target