feat: add vLLM LXC container script

ct/vllm.sh

@@ -0,0 +1,69 @@
+#!/usr/bin/env bash
+source <(curl -fsSL https://raw.githubusercontent.com/community-scripts/ProxmoxVE/main/misc/build.func)
+# Copyright (c) 2021-2026 community-scripts ORG
+# Author: piotrlaczykowski
+# License: MIT | https://github.com/community-scripts/ProxmoxVE/raw/main/LICENSE
+# Source: https://github.com/vllm-project/vllm
+
+APP="vLLM"
+var_tags="${var_tags:-ai;llm}"
+var_cpu="${var_cpu:-8}"
+var_ram="${var_ram:-16384}"
+var_disk="${var_disk:-40}"
+var_os="${var_os:-ubuntu}"
+var_version="${var_version:-24.04}"
+var_unprivileged="${var_unprivileged:-0}"
+var_gpu="${var_gpu:-yes}"
+
+header_info "$APP"
+variables
+color
+catch_errors
+
+function update_script() {
+  header_info
+  check_container_storage
+  check_container_resources
+
+  if [[ ! -d /opt/vllm ]]; then
+    msg_error "No ${APP} Installation Found!"
+    exit
+  fi
+
+  RELEASE=$(get_latest_gh_tag "vllm-project/vllm")
+  if [[ ! -f /opt/vLLM_version.txt ]] || [[ "${RELEASE}" != "$(cat /opt/vLLM_version.txt)" ]]; then
+    if [[ ! -f /opt/vLLM_version.txt ]]; then
+      touch /opt/vLLM_version.txt
+    fi
+
+    msg_info "Stopping Service"
+    systemctl stop vllm
+    msg_ok "Stopped Service"
+
+    msg_info "Updating ${APP} to ${RELEASE}"
+    $STD uv pip install --python /opt/vllm/.venv/bin/python --upgrade "vllm==${RELEASE#v}"
+    echo "${RELEASE}" >/opt/vLLM_version.txt
+    msg_ok "Updated ${APP} to ${RELEASE}"
+
+    msg_info "Starting Service"
+    systemctl start vllm
+    msg_ok "Started Service"
+    msg_ok "Updated successfully!"
+  else
+    msg_ok "No update required. ${APP} is already at ${RELEASE}"
+  fi
+  exit
+}
+
+start
+build_container
+description
+
+msg_ok "Completed successfully!\n"
+echo -e "${CREATING}${GN}${APP} setup has been successfully initialized!${CL}"
+echo -e "${INFO}${YW} Access it using the following URL:${CL}"
+echo -e "${TAB}${GATEWAY}${BGN}http://${IP}:8000${CL}"
+echo -e "${INFO}${YW} OpenAI-compatible API endpoint:${CL}"
+echo -e "${TAB}${GATEWAY}${BGN}http://${IP}:8000/v1${CL}"
+echo -e "${INFO}${YW} Swagger docs:${CL}"
+echo -e "${TAB}${GATEWAY}${BGN}http://${IP}:8000/docs${CL}"

install/vllm-install.sh

@@ -0,0 +1,148 @@
+#!/usr/bin/env bash
+
+# Copyright (c) 2021-2026 community-scripts ORG
+# Author: piotrlaczykowski
+# License: MIT | https://github.com/community-scripts/ProxmoxVE/raw/main/LICENSE
+# Source: https://github.com/vllm-project/vllm
+
+source /dev/stdin <<<"$FUNCTIONS_FILE_PATH"
+color
+verb_ip6
+catch_errors
+setting_up_container
+network_check
+update_os
+
+msg_info "Installing Dependencies"
+$STD apt install -y \
+  python3 \
+  python3-dev \
+  build-essential \
+  pkg-config \
+  libssl-dev \
+  libffi-dev
+msg_ok "Installed Dependencies"
+
+setup_hwaccel "vllm"
+setup_uv
+
+msg_info "Setting up Python Environment"
+$STD uv venv /opt/vllm/.venv
+msg_ok "Set up Python Environment"
+
+RELEASE=$(get_latest_gh_tag "vllm-project/vllm")
+VLLM_VERSION="${RELEASE#v}"
+
+msg_info "Installing ${APP} ${RELEASE} (Patience — this takes 5-15 minutes)"
+if nvidia-smi &>/dev/null; then
+  msg_info "GPU detected — installing vLLM with CUDA support"
+  $STD uv pip install --python /opt/vllm/.venv/bin/python "vllm==${VLLM_VERSION}"
+else
+  msg_info "No GPU detected — installing vLLM with CPU/OpenVINO backend"
+  $STD uv pip install --python /opt/vllm/.venv/bin/python "vllm==${VLLM_VERSION}" --extra-index-url https://download.pytorch.org/whl/cpu
+fi
+echo "${RELEASE}" >/opt/vLLM_version.txt
+msg_ok "Installed ${APP} ${RELEASE}"
+
+msg_info "Configuring ${APP}"
+mkdir -p /etc/vllm /opt/vllm-models
+cat <<EOF >/etc/vllm/vllm.env
+MODEL="Qwen/Qwen2.5-1.5B-Instruct"
+HOST="0.0.0.0"
+PORT="8000"
+GPU_MEM_UTIL="0.90"
+MAX_MODEL_LEN="4096"
+TENSOR_PARALLEL_SIZE="1"
+QUANTIZATION=""
+# HF_TOKEN=""
+HF_HOME="/opt/vllm-models"
+EOF
+msg_ok "Configured ${APP}"
+
+msg_info "Creating vLLM Server Wrapper"
+cat <<'EOF' >/usr/local/bin/vllm-server
+#!/usr/bin/env bash
+set -a
+source /etc/vllm/vllm.env
+set +a
+
+EXTRA_ARGS=()
+[[ -n "${QUANTIZATION}" ]] && EXTRA_ARGS+=(--quantization "${QUANTIZATION}")
+[[ -n "${MAX_MODEL_LEN}" && "${MAX_MODEL_LEN}" != "0" ]] && EXTRA_ARGS+=(--max-model-len "${MAX_MODEL_LEN}")
+[[ -n "${HF_TOKEN}" ]] && export HUGGING_FACE_HUB_TOKEN="${HF_TOKEN}"
+
+exec /opt/vllm/.venv/bin/python -m vllm.entrypoints.openai.api_server \
+  --model "${MODEL}" \
+  --host "${HOST}" \
+  --port "${PORT}" \
+  --gpu-memory-utilization "${GPU_MEM_UTIL}" \
+  --tensor-parallel-size "${TENSOR_PARALLEL_SIZE}" \
+  --trust-remote-code \
+  "${EXTRA_ARGS[@]}"
+EOF
+chmod +x /usr/local/bin/vllm-server
+msg_ok "Created vLLM Server Wrapper"
+
+msg_info "Creating Service"
+cat <<'EOF' >/etc/systemd/system/vllm.service
+[Unit]
+Description=vLLM OpenAI-Compatible Inference Server
+After=network-online.target
+Wants=network-online.target
+
+[Service]
+Type=simple
+User=root
+WorkingDirectory=/opt/vllm
+ExecStart=/usr/local/bin/vllm-server
+Restart=on-failure
+RestartSec=10
+TimeoutStartSec=300
+StandardOutput=journal
+StandardError=journal
+SyslogIdentifier=vllm
+
+[Install]
+WantedBy=multi-user.target
+EOF
+systemctl enable -q vllm
+msg_ok "Created Service (not started — configure /etc/vllm/vllm.env first)"
+
+msg_info "Installing vllm-cli helper"
+cat <<'EOF' >/usr/local/bin/vllm-cli
+#!/usr/bin/env bash
+case "$1" in
+  start)
+    systemctl start vllm
+    echo "vLLM starting... check logs with: journalctl -u vllm -f"
+    ;;
+  stop) systemctl stop vllm ;;
+  restart) systemctl restart vllm ;;
+  status) systemctl status vllm ;;
+  logs) journalctl -u vllm -f --no-pager ;;
+  config) "${EDITOR:-nano}" /etc/vllm/vllm.env ;;
+  models)
+    /opt/vllm/.venv/bin/python -c "
+import os, pathlib
+hf_home = os.environ.get('HF_HOME', '/opt/vllm-models')
+models = [d for d in pathlib.Path(hf_home).glob('models--*') if d.is_dir()]
+if models:
+    print('Cached models:')
+    for m in models:
+        print(' ' + m.name.replace('models--', '').replace('--', '/'))
+else:
+    print('No cached models found in', hf_home)
+"
+    ;;
+  version)
+    /opt/vllm/.venv/bin/python -c "import vllm; print('vLLM', vllm.__version__)"
+    ;;
+  *) echo "Usage: vllm-cli {start|stop|restart|status|logs|config|models|version}" ;;
+esac
+EOF
+chmod +x /usr/local/bin/vllm-cli
+msg_ok "Installed vllm-cli helper"
+
+motd_ssh
+customize
+cleanup_lxc

json/vllm.json

@@ -0,0 +1,42 @@
+{
+  "name": "vLLM",
+  "slug": "vllm",
+  "categories": [
+    3
+  ],
+  "description": "vLLM is a high-throughput and memory-efficient inference and serving engine for LLMs. It provides an OpenAI-compatible API server for local LLM inference with NVIDIA GPU support.",
+  "type": "ct",
+  "privileged": true,
+  "updateable": true,
+  "interface_port": 8000,
+  "documentation": "https://docs.vllm.ai",
+  "website": "https://github.com/vllm-project/vllm",
+  "logo": "https://raw.githubusercontent.com/vllm-project/vllm/main/docs/source/assets/logos/vllm-logo-text-light.png",
+  "install_methods": [
+    {
+      "type": "default",
+      "script": "ct/vllm.sh",
+      "resources": {
+        "cpu": 2,
+        "ram": 4096,
+        "hdd": 40,
+        "os": "Ubuntu",
+        "version": "24.04"
+      }
+    }
+  ],
+  "default_credentials": {
+    "username": null,
+    "password": null
+  },
+  "notes": [
+    {
+      "text": "GPU passthrough requires a privileged LXC container with NVIDIA drivers installed on the Proxmox host. Edit /etc/vllm/vllm.env to configure model and settings before starting the service.",
+      "type": "info"
+    },
+    {
+      "text": "Default resources (2 CPU, 4GB RAM) are sufficient for installation only. Running models requires significantly more: small models (1-3B) need 4+ CPU and 8GB+ RAM; large models (7B+) require a GPU with 8GB+ VRAM and 16GB+ RAM.",
+      "type": "warning"
+    }
+  ]
+}