Problem

Currently jupyterlite-ai includes explicit and built-in support for Ollama.

However Ollama is compatible with the OpenAI API: https://ollama.com/blog/openai-compatibility

So maybe it could make sense to drop support for Ollama as built-in-provider, and instead document how to use the generic OpenAI-compatible provider?

This would also mean dropping the dependency on the third-party provider:

Proposed Solution

Additional context

The main downside would be to manually provide the base URL for Ollama when using the generic provider.

Unless the UI could suggest some common URLs, for example Ollama, Openrouter and LiteLLM proxy.