+1

Note that I only saw issue #2 (for adding proper Ollama support) after I created this PR; two things:

• obviously this one is for OpenAI compatible APIs instead, not the Ollama API proper
• this one does the bare minimum to allow pointing to a custom endpoint, with equally minimum functionality

IMHO this one still has value as an OAI compatible MVP, as it's minimally intrusive but it already affords access to basic self-hosted inference. Many other LLM clients also support this type of integration, sometimes in addition to Ollama.

Hey @ldmosquera! Thanks for the PR, but I'd rather we skip the half-measure and build a proper Ollama provider that speaks directly to their native API instead. The compatibility layer is just going to cause headaches - when I built the Gemini provider, I started with their OpenAI compatibility layer and ran into a ton of issues. Plus, Ollama's layer doesn't even support tool_choice. If you're up for it, I'd love to see a PR for that instead.

Fair enough; I can start a draft PR for that one.