Reasoning Agent Events

Capture lifecycle and content-delta events from a separate reasoning stage.

Capture reasoning events during streaming.

reasoning_agent_events.py
"""
Reasoning Agent Events
=============================

Demonstrates how to capture reasoning events during streaming.
"""

import asyncio

from agno.agent import RunEvent
from agno.agent.agent import Agent
from agno.models.openai import OpenAIResponses

# ---------------------------------------------------------------------------
# Create Agent
# ---------------------------------------------------------------------------
finance_agent = Agent(
    model=OpenAIResponses(id="gpt-5.6"),
    reasoning_model=OpenAIResponses(id="o3-mini"),
)


async def run_agent_with_events(prompt: str):
    content_started = False
    async for run_output_event in finance_agent.arun(
        prompt,
        stream=True,
        stream_events=True,
    ):
        if run_output_event.event in [RunEvent.run_started, RunEvent.run_completed]:
            print(f"\nEVENT: {run_output_event.event}")

        if run_output_event.event in [RunEvent.reasoning_started]:
            print(f"\nEVENT: {run_output_event.event}")

        if run_output_event.event in [RunEvent.reasoning_step]:
            print(f"\nEVENT: {run_output_event.event}")
            print(f"REASONING CONTENT: {run_output_event.reasoning_content}")  # type: ignore

        if run_output_event.event in [RunEvent.reasoning_completed]:
            print(f"\nEVENT: {run_output_event.event}")

        if run_output_event.event in [RunEvent.run_content]:
            if not content_started:
                print("\nCONTENT:")
                content_started = True
            else:
                print(run_output_event.content, end="")


# ---------------------------------------------------------------------------
# Run Agent
# ---------------------------------------------------------------------------
if __name__ == "__main__":
    task = (
        "Analyze the key factors that led to the signing of the Treaty of Versailles in 1919. "
        "Discuss the political, economic, and social impacts of the treaty on Germany and how it "
        "contributed to the onset of World War II. Provide a nuanced assessment that includes "
        "multiple historical perspectives."
    )
    asyncio.run(
        run_agent_with_events(
            task,
        )
    )

Handle native reasoning content

The configured native reasoning stage emits ReasoningContentDelta; ReasoningStep belongs to a different structured reasoning path. Replace run_agent_with_events in the saved script with this consumer:

async def run_agent_with_events(prompt: str):
    content_started = False
    async for event in finance_agent.arun(prompt, stream=True, stream_events=True):
        if event.event in (
            RunEvent.run_started, RunEvent.run_completed,
            RunEvent.reasoning_started, RunEvent.reasoning_completed,
        ):
            print(f"\nEVENT: {event.event}")
        elif event.event == RunEvent.reasoning_content_delta:
            print(event.reasoning_content, end="", flush=True)
        elif event.event == RunEvent.reasoning_step:
            print(event.reasoning_content, end="", flush=True)
        elif event.event == RunEvent.run_content and event.content:
            if not content_started:
                print("\nCONTENT:")
                content_started = True
            print(event.content, end="", flush=True)
        elif event.event == RunEvent.run_error:
            print(f"\nERROR: {event.content}")

The intermediate text can be the separate reasoning model's answer. It is not guaranteed access to a model's private chain of thought. As of September 7, 2026, OpenAI schedules o3-mini retirement for October 23, 2026; check the deprecation table before selecting that model for a new deployment.

Run the Example

Set up your virtual environment

uv venv --python 3.12
source .venv/bin/activate

Install dependencies

uv pip install -U agno openai

Export your OpenAI API key

export OPENAI_API_KEY="your_openai_api_key_here"

Run the example

Save the code above as reasoning_agent_events.py, then run:

python reasoning_agent_events.py

Full source: cookbook/02_agents/14_advanced/reasoning_agent_events.py