Ollama

Runtime provider id: ollama. Talks to a local (or remote) Ollama instance — no API key required.

Provider options

Field	Type	Notes
baseURL	string	Optional. Defaults to the local Ollama instance.
reasoning	object	Optional. See Reasoning.

Normalized temperature on the LLM entry is mapped into Ollama’s options.temperature.

Reasoning

Setting reasoning.effort enables Ollama thinking, mapped to the think level (minimal/low -> "low", medium -> "medium", high -> "high"); omit to leave thinking off. summary/display are not used by Ollama.

Example configs

Local instance with the default Ollama endpoint:

{
  "name": "Ollama",
  "provider": "ollama",
  "options": {}
}

{
  "name": "Default",
  "provider": "Ollama",
  "options": {
    "model": "gemma4:e4b"
  },
  "default": true
}

Remote instance, with thinking enabled on a reasoning-capable local model:

{
  "name": "Ollama Remote",
  "provider": "ollama",
  "options": {
    "baseURL": "http://ollama.internal:11434",
    "reasoning": { "effort": "medium" }
  }
}

{
  "name": "Qwen3 (thinking)",
  "provider": "Ollama Remote",
  "options": {
    "model": "qwen3:32b",
    "temperature": 0.7
  }
}