Tool Calls with Streaming

[Igrium]

What's the best approach for handling tool calls & saving message history when streaming chat completion responses from OpenAI (via flow)

Normally, you look at the returned message and see if it has any tool calls, but streamed completions use chunks and stuff, which I still don't fully understand.