LMDeploy Model Integration & Serving

Set up LMDeploy to serve at least two different LLMs (e.g., OpenAI o3 and a Llama variant). Implement the logic within your evaluation engine to dynamically switch between these models for benchmarking. Design the MCP-enabled tool to feed model outputs from LMDeploy into the DSPy evaluation pipeline.