Stream study guide to the screen

[yasonk]

Study guide will take some time to generate, would be nice if it didn't have to wait and would just stream.

for message_chunk, metadata in graph.stream(
    {"topic": "ice cream"},
    stream_mode="messages",
):
    if message_chunk.content:
        print(message_chunk.content, end="", flush=True)

Here it streams content of message chunk. The metadata contains which node is producing stream from LLM.
However, ToolNode for some reason streams additional metadata in a strange format. So for ToolNode streaming, need to tell LLM to wrap content in some markers like "::output::Here is your response from LLM::output"