Use local LLM via Ollama on desktop

[orlando0309constellation]

Question

I try to connect my local LLM on the OpenCode Desktop, but it shows only Ollama Cloud, can anyone help me?

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• Local Ollama models are not agentic #5694: Local Ollama models are not agentic
• Why is opencode not working with local llms via Ollama? #4428: Why is opencode not working with local llms via Ollama?
• Nemotron seems to ignore my request when using ollama #6327: Nemotron seems to ignore my request when using ollama
• Ollama: User message content arrives as empty array - model cannot see user input #5187: Ollama: User message content arrives as empty array - model cannot see user input

Feel free to ignore if none of these address your specific case.

[TomasHubelbauer]

There is some documentation here (local and cloud sections) but I didn't really have any success following it, especially such that the local model would do tool calls right:
https://opencode.ai/docs/providers/#ollama

[Fastidious]

There is some documentation here (local and cloud sections) but I didn't really have any success following it, especially such that the local model would do tool calls right: https://opencode.ai/docs/providers/#ollama

Are those instructions for desktop, or CLI? As it is not specified, I assume it is CLI only (considering that desktop is beta). I am having the same issue as OP.

[Hronom]

Same here

[Insighttful]

Same issue.

Openchamber Version 1.9.5 app shows only Ollama local (which is defined in the opencode.json config) but does not show any other provider (or allow adding any other provider). Openchamber VS Code extension DOES show the other providers (OpenAI, Opencode Zen, Opencode Go).

This makes the app extremely limited in use.

APP

VS CODE EXT.

[JinkaiNiu]

To use a local Ollama model in OpenCode Desktop, you need to configure Ollama as a local provider in your config. The "Ollama Cloud" you see is the default hosted option — you need to add a local Ollama provider manually.

Add this to your opencode.json (or ~/.config/opencode/opencode.jsonc):

{
  "provider": {
    "ollama": {
      "npm": "@ai-sdk/openai-compatible",
      "options": { "baseURL": "http://localhost:11434/v1" },
      "models": {
        "llama3.2-vision": { "name": "llama3.2-vision" }
      }
    }
  }
}

Then run /models in the TUI to refresh the model list, and you should see your local Ollama models.

For more details, see the docs:
https://opencode.ai/docs/providers/