OpenResponses.Adapters.Ollama (OpenResponses v0.1.1)

Adapter for Ollama — run any open-weight model locally.

Ollama exposes an OpenAI-compatible chat endpoint. No API key is required. GPU latency dominates for local models; run Ollama on the same host or a fast private network.

Configuration

No API key needed. Override the base URL if Ollama is running on a different host:

config :open_responses, :provider_config, %{
  ollama: [
    base_url: "http://gpu-server:11434"
  ]
}

Routing

Selected automatically for common local model name prefixes:

config :open_responses, :routing, %{
  ~r/^llama|^mistral|^phi|^qwen/ => OpenResponses.Adapters.Ollama
}

Add your own model name patterns to the routing table as needed.