Open WebUI

Open WebUI is the interface most people put in front of their own models. It is self-hosted, open source, and deliberately complete — multi-user accounts, permissions, document chat, tools, and model management, all running on infrastructure you control.

Notable capabilities

• Any backend. Point it at Ollama, a local inference server, or any OpenAI-compatible endpoint, and mix several at once.
• Retrieval built in. Upload documents or point it at a URL and query them, with a local vector store handling the indexing.
• Multi-user with roles. Real accounts, groups, and permissions make it usable as a team deployment rather than a personal toy.
• Tools and functions. Extend the assistant with Python tools, pipelines, and custom model presets.
• Runs anywhere. A single Docker command gets a working deployment; Kubernetes charts exist for larger installs.

Who runs it

Teams that want a ChatGPT-style experience without sending data to a vendor, homelab users pairing it with local models, and organisations that need audit and access control around AI usage. Free and open source.