docs-openwebui/docs/enterprise/deployment/python-pip.md

sidebar_position, title

sidebar_position	title
1	Python / Pip on VMs

Python / Pip on Auto-Scaling VMs

Deploy open-webui serve as a systemd-managed process on virtual machines in a cloud auto-scaling group (AWS ASG, Azure VMSS, GCP MIG).

:::info Prerequisites Before proceeding, ensure you have configured the shared infrastructure requirements: PostgreSQL, Redis, a vector database, shared storage, and content extraction. :::

When to Choose This Pattern

• Your organization has established VM-based infrastructure and operational practices
• Regulatory or compliance requirements mandate direct OS-level control
• Your team has limited container expertise but strong Linux administration skills
• You want a straightforward deployment without container orchestration overhead

Architecture

flowchart TB
    LB["Load Balancer"]

    subgraph ASG["Auto-Scaling Group"]
        VM1["VM 1"]
        VM2["VM 2"]
        VM3["VM N"]
    end

    subgraph Backend["Backing Services"]
        PG["PostgreSQL + PGVector"]
        Redis["Redis"]
        S3["Object Storage"]
        Tika["Tika"]
    end

    LB --> ASG
    ASG --> Backend

Installation

Install on each VM with the postgres extra, which adds the PostgreSQL and PGVector drivers:

python3.12 -m venv /opt/open-webui/venv
/opt/open-webui/venv/bin/pip install "open-webui[postgres]"

The all extra also adds the optional vector store clients, Playwright and Unstructured; install it only if you use one of them.

Create a systemd unit to manage the process:

[Unit]
Description=Open WebUI
After=network.target

[Service]
Type=simple
User=openwebui
EnvironmentFile=/etc/open-webui/env
ExecStart=/opt/open-webui/venv/bin/open-webui serve
Restart=always
RestartSec=5

[Install]
WantedBy=multi-user.target

Place your environment variables in /etc/open-webui/env (see Critical Configuration). Also set DATA_DIR, for example to /var/lib/open-webui, so local caches live outside the Python package directory, which is the default.

Scaling Strategy

• Horizontal scaling: Configure your auto-scaling group to add or remove VMs based on CPU utilization or request count.
• Health checks: Point your load balancer health check at /ready, which answers once startup has finished and the database and Redis respond. /health only shows the process is running.
• One process per VM: Keep UVICORN_WORKERS=1 and let the auto-scaler manage capacity.
• Sticky sessions: Optional; required only with ENABLE_WEBSOCKET_SUPPORT=false. See Session Affinity.

Key Considerations

Consideration	Detail
OS patching	You are responsible for OS updates, security patches, and Python runtime management.
Python environment	Use Python 3.11 or 3.12 (3.12 recommended) in a virtual environment.
Storage	Use object storage (such as S3) or a shared filesystem (such as NFS) since VMs in an auto-scaling group do not share a local filesystem.
Tika sidecar	Run a Tika server on each VM or as a shared service. A shared instance simplifies management.
Updates	Scale the group to one instance, set ENABLE_DB_MIGRATIONS=true on it, update the package (pip install --upgrade "open-webui[postgres]") and restart, wait for the migrations to finish, then scale back up with migrations off and the new version on every VM. See Updates and Migrations.

For pip installation basics, see the Quick Start guide.

---

Need help planning your enterprise deployment? Our team works with organizations worldwide to design and implement production Open WebUI environments.

Contact Enterprise Sales →