Max conversationToolCalls executed per LLM response. Each call is a run step (and
summarizeRange is a full LLM sub-call) — an unbounded Promise.all would let a single
response fan out arbitrary work/cost. Excess calls are reported back to the model as
skipped so it can re-request them next turn. Consumed by BaseAgent's executor.
Max conversationToolCalls executed per LLM response. Each call is a run step (and summarizeRange is a full LLM sub-call) — an unbounded Promise.all would let a single response fan out arbitrary work/cost. Excess calls are reported back to the model as skipped so it can re-request them next turn. Consumed by BaseAgent's executor.