Does Vanna2.0 support connecting with vLLM？

My environment uses an Oracle database, and the large model is deployed locally via vLLM. On the official website, I only found documentation for OpenAI or Ollama, but no instructions for a locally deployed vLLM model. I would like to know:

1. Does Vanna 2.0 support locally deployed vLLM, or large models that follow the OpenAI-compatible API specification?
2. If not, can I implement support by extending the code myself?
3. Could you provide some guidance and hints? Thank you.‍