Support slashes in model names when using local LLMs via custom base_url

[WismutHansen]

First of all thank you for building DSRs, I love the idea of DSPy but prefer building most of my tools in Rust.

When using e.g. LMstudio, some models have a slash in their name by default like "qwen/qwen3-coder-30b". This trips up the build in provider detection even when setting a custom base_url. It's of course possible to serve the model under a name without a slash but maybe there is a way to include better support for local LLM engines/models.

Thank you!

[krypticmouse]

Thank you for the issue! From #37 and #39 onwards we support this via : instead of /, you can do: openai:gpt-4o-mini style declarations now, for known providers. You can refer the docs for this:

• OAI compat
• Local Models
• Known provider

Unfortunately, we don't support Qwen Provider natively but you can pass the oai compatible base url and make it work!