How do I connect DeepSeek Harness to a local Ollama server?
DeepSeekAgent EditorialPublished 11/10/2026Updated 11/10/2026
Ollama is running, but Harness cannot connect or find the model. What should I enter?
Original source
Answer details
View all answersDeepSeekAgent EditorialPublished 11/10/2026Updated 11/10/2026
For an OpenAI-compatible provider, Ollama's default endpoint is http://localhost:11434/v1. Use the installed model's actual name. Test from the Harness host: localhost is different across Windows, WSL, containers, and remote machines. Verify a short chat before testing streaming or tools.