Open WebUI Can't Connect to Ollama in Docker: The Fixes

Description and use

Open WebUI fails to connect to Ollama in Docker due to incorrect URL and network isolation. Set OLLAMA_BASE_URL to the correct address, add host.docker.internal on Linux hosts, and ensure Ollama listens on a non‑localhost interface. Then the UI lists models and passes connection checks.

Problem

Open WebUI cannot reach Ollama when both run in Docker, resulting in an empty model picker or failed admin connection check.

Outcome

After configuring the correct Ollama URL and ensuring network visibility, Open WebUI can list models and pass admin connection checks.

Workflow

• 1. Identify where Ollama runs (same Compose file, host, or another machine).
• 2. Set OLLAMA_BASE_URL to the appropriate URL: http://ollama:11434 for same Compose, http://host.docker.internal:11434 for host (with extra_hosts on Linux), or http://<host-ip>:11434 for another machine.
• 3. If using host, add --add-host host.docker.internal:host-gateway (or in Compose - host.docker.internal:host-gateway). Ensure Ollama listens on an address other than 127.0.0.1 (e.g., 0.0.0.0).
• 4. Test connectivity from inside the Open WebUI container: docker exec open-webui curl -s http://ollama:11434 should return "Ollama is running" and curl -s http://host.docker.internal:11434/api/version should return a JSON version string.
• 5. If OLLAMA_BASE_URL change appears ineffective, update the value in Open WebUI Admin Panel > Settings > Connections, as the first value is persisted in the database.
• 6. Optionally, monitor the connection with an HTTP monitor like Uptime Kuma from a container with the same network access.

Evidence

Open WebUI shows no models and admin connection check reports "Ollama: Network Problem".

Source

docker exec open-webui curl -s http://ollama:11434 should return "Ollama is running".

Source

OLLAMA_BASE_URL set to http://ollama:11434 when both run in one Compose file.

Source