- `_processToolCalls` runs the `For each` sequentially - A model can return independent `tool_calls` in one message - dispatching them concurrently (e.g. via`4D.SystemWorker`?) cuts latency.
_processToolCallsruns theFor eachsequentiallytool_callsin one message4D.SystemWorker?) cuts latency.