Ollama setup

Connect Odysseus to Ollama

Ollama is usually the fastest way to give Odysseus a local model. The main trick is remembering where Odysseus is running: on the host machine, inside Docker, or inside WSL2.

1. Test Ollama outside Odysseus

Start with a smaller model. If this step fails, fix Ollama before changing Odysseus settings.

ollama pull qwen2.5:7b
ollama run qwen2.5:7b
ollama list

2. Choose the right endpoint

If Odysseus runs natively on the same machine as Ollama, the local endpoint is usually:

http://localhost:11434/v1

If Odysseus runs in Docker while Ollama runs on the host machine, use Docker's host gateway name instead:

http://host.docker.internal:11434/v1

3. Save it as an OpenAI-compatible provider

Many local model tools expose an OpenAI-compatible API shape. In Odysseus, add or edit the model provider, set the base URL to the correct endpoint, and use the exact model name from ollama list.

Debug checklist

Does ollama list show the model?
Is Odysseus native, Docker, or WSL2?
Did you use localhost from inside Docker by mistake?
Does a firewall block port 11434?
Did you save the model name exactly, including tag?

Security note: Do not expose Ollama's API port publicly. Keep it local or behind strict private network rules.