Ollama API adapter implementation.
Handles API communication with local Ollama models through the configured HTTP adapter.
Ollama doesn't require authentication, so we mark the OpenAI-compatible backend explicitly for ReqLLM instead of sending placeholder credentials.