RLHF using microsoft lightning

[PrestigeDevop]

As the title suggest , the agent lightning framework introduce a rlhf optimization rewards with AgentFlow — A modular multi-agent framework that combines planner, executor, verifier, and generator agents with the Flow-GRPO algorithm to tackle long-horizon, sparse-reward tasks.

I'm gonna assume it would , but the starting point will be from that repo.

second question is there any API similar to onnxGenAI local SemanticKernel.Connectors.OnnxRuntimeGenAI ? so no api required .

finally is there any advice for creating a polyglot MCP server with this library in mind ?

thanks