Limit LLM Calls Per Agent Turn in Moderated Strategy

[dspencej]

I'm running a task with multiple agents using a moderated turn strategy. Some models get stuck in a loop, repeatedly making the same LLM call. More often, they keep making tool calls instead of ending their turn when they need more information.

I'd like to set a per-turn limit on LLM calls for each agent. Once this threshold is reached, the agent should end its turn and pass control back to the moderator agent.

I've tried max_llm_calls and max_agent_turns, but these stop the entire run rather than just enforcing a per-turn limit. Is there a built-in way to handle this?

Thanks!