Open WebUI

Open WebUI reads the model list from GET /v1/models and chats through /v1/chat/completions, both on the one ai.ml base URL. Every catalog model shows up in its model picker.

On first start

Pass the connection as environment variables, for example with docker run --env-file open-webui.env …:

# Environment for the Open WebUI container (docker run --env-file …).
# Open WebUI lists the models from GET /v1/models and chats through
# /v1/chat/completions; both are on the one ai.ml base URL.
OPENAI_API_BASE_URL=https://api.ai.ml/v1
OPENAI_API_KEY=aiml-live-…
# only ai.ml: no local Ollama connection
ENABLE_OLLAMA_API=False

On an existing install

These variables only seed a fresh install. After that, connections live in the admin panel: Admin Settings → Connections → OpenAI API → add a connection.

• URL: https://api.ai.ml/v1
• Key: your ai.ml key

What to expect

• The model picker shows catalog ids (anthropic/claude-sonnet-4.6), and Open WebUI only accepts models from that list. A request naming an alias (claude-sonnet-4-6) gets "model not found" from Open WebUI before it reaches ai.ml.
• Streaming, tool calls and images go through the same chat route, translated for whichever provider serves the model.
• Give the key a budget: everyone using the Open WebUI instance spends from it.

Verified with Open WebUI 0.11.4: model list, a chat, and a streamed reply.