Open-WebUI

Open WebUI is a self-hosted AI interface that gives individuals and teams one chat workspace for Ollama, OpenAI-compatible APIs, and other model providers.

Why people choose it

It is a practical choice when you want a browser-based chat experience without tying the interface to one hosted provider. The official docs cover Ollama, OpenAI, Anthropic, vLLM, and other connections, along with model management, conversation history, and user administration.

What is included

The project supports Docker, Python, Kubernetes, and a desktop app. Depending on the image and configuration, it can support document uploads, retrieval-augmented generation (RAG), web search, tool calling, voice features, and Open Terminal. These capabilities can require an embedding provider, external services, or the standard image rather than the slim image.

Deployment requirements

Docker is the recommended path for most deployments. Persist `/app/backend/data` and set a stable `WEBUI_SECRET_KEY`; the official quick start warns that the volume contains chats, users, and settings. The standard image includes more local ML and document-processing dependencies, while `main-slim` is smaller but delegates several features to external providers.

Main tradeoffs

Open WebUI is an interface, not a model by itself. You still need Ollama, another local runtime, or a hosted API provider. RAG, voice, web search, and agent features add configuration and resource requirements, and a public deployment needs normal authentication, updates, backups, and network hardening.

Best fit

Choose it if you want a polished self-hosted ChatGPT-style workspace across local and hosted models, especially when you need more than a single-user command-line runtime. A simpler API-only setup may be better if you only need programmatic inference.