Run Open WebUI as a production local AI appliance

Receive Open WebUI ready to use rather than merely installed? The AI Cube Pro combines Open WebUI, a local model, hardening, functional testing, and initial setup into a prepared business AI platform. Explore the AI Cube Pro

Open WebUI is the interface through which staff actually use local AI. It resembles familiar AI chats, can present several models, and manages knowledge, users, groups, and other workspace resources. Production requires the interface to be embedded into a complete platform.

What users receive

• chat with one or more approved models;
• history and workspaces on the organisation's platform;
• file uploads and personal knowledge collections;
• shared team knowledge after approval;
• models, prompts, and tools according to role;
• one interface for local and deliberately approved external models.

These capabilities do not mean an external source is already integrated. Users can maintain knowledge in Open WebUI; automatic Nextcloud, SharePoint, or DMS synchronisation needs a separate pipeline.

The guide to RAG with Nextcloud, SharePoint, and DMS explains how such a connection is built.

The platform underneath

Model endpoint

Ollama, vLLM, or another compatible runtime serves the local model. Model release, quantisation, context, and concurrency are documented. The endpoint is not exposed directly.

Database and storage

Chat history, users, configuration, and metadata need persistent storage. Files and knowledge artefacts use defined volumes or object storage. Backup and recovery must cover both together.

Identity and permissions

Business operation uses local accounts or OIDC/SSO, groups, and minimal defaults. Access to models, knowledge, and tools is deliberate. Administration is separated from normal use.

Network

TLS, internal DNS, firewall, and potentially VPN or NetBird constrain access. Web interface, model server, database, and administration do not need to share one exposed segment.

Minimum standard for production

These decisions turn an installation into an operable service. They should be documented before several teams store production data on the platform.

Updates without uncontrolled downtime

Open WebUI and the model runtime evolve independently. Updates require release review, backup, test, maintenance window, and rollback. Unreviewed automatic upgrades are unsuitable for a production platform.

Model changes also require evaluation against the same work tasks. A higher version number does not replace quality testing.

Monitoring and backup

Monitor availability, latency, errors, memory, storage, services, and backup status. Content prompts do not need to be sent wholesale to external monitoring. Logs should be limited to what operations require.

A restore test verifies that database, files, keys, and configuration recover together. A backup without a tested restart is not reliable evidence of recoverability.

AI Cube Pro as a prepared foundation

For the AI Cube Pro, WZ-IT installs Open WebUI, local runtime, and an agreed model, configures and hardens the base system, and tests function. Initial setup and five hours of support are included. The platform is usually ready within two weeks after configuration approval.

SSO, automated sources, custom RAG pipelines, MCP servers, special backup targets, and secure site connectivity are added to the existing environment. Optional managed operations start at EUR 149.90 excluding VAT per AI Cube and month.

Sources

• Open WebUI: Knowledge
• Open WebUI: authentication and access
• Open WebUI: groups and resource access
• Open WebUI: permissions
• WZ-IT: Open WebUI expertise