Vertex AI Basic Reasoning Stream
Stream Gemini 2.5 Flash reasoning on Vertex AI with a fixed 1,024-token thinking budget and thought summaries.
"""
Basic Reasoning Stream
======================
Demonstrates this reasoning cookbook example.
"""
import asyncio
from agno.agent import Agent
from agno.models.google import Gemini
from agno.run.agent import RunEvent # noqa
# ---------------------------------------------------------------------------
# Create Example
# ---------------------------------------------------------------------------
def run_example() -> None:
async def streaming_reasoning():
"""Test streaming reasoning with a VertexAI with Gemini model."""
# Create an agent with reasoning enabled
agent = Agent(
reasoning_model=Gemini(
id="gemini-2.5-flash",
vertexai=True,
thinking_budget=1024, # Required to enable thinking mode
include_thoughts=True, # Include thought summaries in response
),
instructions="Think step by step about the problem.",
)
prompt = "What is 25 * 37? Show your reasoning."
await agent.aprint_response(prompt, stream=True, stream_events=True)
# Use manual event loop to see all events
# async for run_output_event in agent.arun(
# prompt,
# stream=True,
# stream_events=True,
# ):
# if run_output_event.event == RunEvent.run_started:
# print(f"\nEVENT: {run_output_event.event}")
# elif run_output_event.event == RunEvent.reasoning_started:
# print(f"\nEVENT: {run_output_event.event}")
# print("Reasoning started...\n")
# elif run_output_event.event == RunEvent.reasoning_content_delta:
# # This is the NEW streaming event for reasoning content
# print(run_output_event.reasoning_content, end="", flush=True)
# elif run_output_event.event == RunEvent.reasoning_step:
# print(f"\nEVENT: {run_output_event.event}")
# elif run_output_event.event == RunEvent.reasoning_completed:
# print(f"\n\nEVENT: {run_output_event.event}")
# elif run_output_event.event == RunEvent.run_content:
# if run_output_event.content:
# print(run_output_event.content, end="", flush=True)
# elif run_output_event.event == RunEvent.run_completed:
# print(f"\n\nEVENT: {run_output_event.event}")
if __name__ == "__main__":
asyncio.run(streaming_reasoning())
# ---------------------------------------------------------------------------
# Run Example
# ---------------------------------------------------------------------------
if __name__ == "__main__":
run_example()Before authenticating, choose a Google Cloud project with billing enabled, enable the Vertex AI API, and grant the calling identity the required Vertex AI permissions (for example, roles/aiplatform.user). Select a location where the model is available. Application Default Credentials select an identity; login alone does not provision the project or grant access. Follow the Google project setup.
Google lists this hosted gemini-2.5-flash version for retirement on October 20, 2026. Check the hosted model lifecycle before migrating. The 1,024-token setting controls the thinking budget; this model supports dynamic thinking without an explicit budget. include_thoughts=True requests summaries, not the complete internal trace.
The program omits the final model, so the answer uses Agno's default OpenAI model and still needs the separate OPENAI_API_KEY.
Run the Example
Set up your virtual environment
uv venv --python 3.12
source .venv/bin/activateInstall dependencies
uv pip install -U agno google-genai openaiExport environment variables
export GOOGLE_CLOUD_LOCATION="your_google_cloud_location_here"
export GOOGLE_CLOUD_PROJECT="your_google_cloud_project_here"
export OPENAI_API_KEY="your_openai_api_key_here"Authenticate with Google Cloud
Install the Google Cloud CLI, then sign in with Application Default Credentials:
gcloud auth application-default loginRun the example
Save the code above as basic_reasoning_stream.py, then run:
python basic_reasoning_stream.pyFull source: cookbook/10_reasoning/models/vertex_ai/basic_reasoning_stream.py