Playground UI and Debug Terminal not giving same response when the tool call output is huge.

[samyogdhital]

Here is a simple code.

    name="Huge Chunk returning agent",
    model=get_model(),
    agent_id="test_agent",
    tools=[HugeTextChunkReturnTool()],
    instructions=[
        "Your job is to return the huge text chunk given by HugeTextChunkReturnTool class."
    ],
    markdown=True,
    debug_mode=True,
)

Tool Code

from phi.tools import Toolkit


class HugeTextChunkReturnTool(Toolkit):
    def __init__(self):
        super().__init__(name="huge_text_chunk_return_tool")
        self.register(self.get_huge_text_chunk)

    # The end function must accept str from llm and give str to llm.
    def get_huge_text_chunk(self) -> str:
        """
        Returns a huge text chunk.
        """
    return ".............................................................. static long 100k+ token text."

Here is the video showing the output of the prompt in Agentic UI vs in the debug terminal.
See the output on playground frontend is not clear and not same as the output shown in the debug terminal with debug turned True
The debug terminal gives correct output, but Agentic UI is not able to show that. I think the connection cuts from your localhost to agentic UI before the code execution finishes.

Couldn't upload video here directly. Please check this YouTube video.

I have been noticing this lately particularly if we are dealing with text returned by tool call that is close to 100K token.

[pritipsingh]

Thanks for pointing this out, @samyogdhital. We're already aware that large chunks break in the playground, and we're actively working to fix this as soon as possible.

We really appreciate your time and effort in providing such a detailed description—thank you!