Is there any existing work to remove the hardcoded formatting and debug styling in the agent printed/streamed responses?

[jspv]

printed/streamed responses from the agents include all sorts of extra messages and implied UI/debug formatting. For example:

In conversable_agent.py:

def _print_received_message(self, message: Union[Dict, str], sender: Agent):
    iostream = IOStream.get_default()
    # print the message received
    iostream.print(colored(sender.name, "yellow"), "(to", f"{self.name}):\n", flush=True)
    message = self._message_to_dict(message)

if num_code_blocks == 1:
    iostream.print(
        colored(
            f"\n>>>>>>>> EXECUTING CODE BLOCK (inferred language is {code_blocks[0].language})...",
            "red",
        ),
        flush=True,
    )
else:
     iostream.print(
         colored(
            f"\n>>>>>>>> EXECUTING {num_code_blocks} CODE BLOCKS (inferred languages are [{', '.join([x.language for x in code_blocks])}])...",
            "red",
        ),
        flush=True,
    )

This really makes for extremely messy streamed output (e.g. using Websockets to connect to a different UI framework) where I just want the returned text, not the pre-defined formatting and extra language.

I'm thinking of changing the behavior a bit to a structured output with json inclusive of the raw response, the content with the original formatting (for compatibility), metadata on the sender/receiver of the message, etc. but this seems like it would be a big undertaking.

Is anyone working on something similar or related?