Current limitation: Using vLLM (or other OpenAI-compatible local endpoints) currently requires manually overriding the OpenAI provider base URL, which is not discoverable.
Proposed solution: Add a Provider::Vllm variant or a dedicated configuration helper to streamline connecting to vLLM instances.
Current limitation: Using vLLM (or other OpenAI-compatible local endpoints) currently requires manually overriding the OpenAI provider base URL, which is not discoverable.
Proposed solution: Add a
Provider::Vllmvariant or a dedicated configuration helper to streamline connecting to vLLM instances.