Vertex AI Claude Adaptive Thinking

Use adaptive thinking with effort levels to control reasoning depth on Claude VertexAI.

Cookbook example demonstrating adaptive thinking with output_config on VertexAI.

adaptive_thinking.py

"""
VertexAI Claude Adaptive Thinking
=================================

Cookbook example demonstrating adaptive thinking with output_config on VertexAI.

For Claude 4.6 VertexAI models, use adaptive thinking with the effort parameter
to control thinking depth. Valid effort values:
- "low": Most efficient, significant token savings
- "medium": Balanced approach with moderate savings
- "high": Default, high capability for complex reasoning
- "max": Absolute maximum capability (Opus 4.6 only)

Prerequisites:
- Set GOOGLE_CLOUD_PROJECT and CLOUD_ML_REGION environment variables
- Authenticate with: gcloud auth application-default login
"""

from agno.agent import Agent
from agno.models.vertexai import Claude

# ---------------------------------------------------------------------------
# Create Agent with Adaptive Thinking
# ---------------------------------------------------------------------------

agent = Agent(
    model=Claude(
        id="claude-sonnet-4-6@20250514",
        max_tokens=4096,
        thinking={"type": "adaptive"},
        output_config={"effort": "high"},
    ),
    markdown=True,
)

# ---------------------------------------------------------------------------
# Run Agent
# ---------------------------------------------------------------------------
if __name__ == "__main__":
    # Complex reasoning task that benefits from extended thinking
    agent.print_response(
        "Explain the key differences between recursion and iteration, "
        "and when you would choose one over the other in software development."
    )

    # With streaming
    agent.print_response(
        "What are the trade-offs between microservices and monolithic architectures?",
        stream=True,
    )

Run the Example

1. Set up your virtual environment

   Mac

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

   uv venv --python 3.12
   .venv\Scripts\activate
   
2. Install dependencies

   uv pip install -U agno "anthropic[vertex]"
   
3. Export environment variables

   Mac/Linux

   export ANTHROPIC_VERTEX_PROJECT_ID="your_anthropic_vertex_project_id_here"
   export CLOUD_ML_REGION="your_cloud_ml_region_here"
   

   $Env:ANTHROPIC_VERTEX_PROJECT_ID="your_anthropic_vertex_project_id_here"
   $Env:CLOUD_ML_REGION="your_cloud_ml_region_here"
   
4. Authenticate with Google Cloud

   Sign in with Application Default Credentials:

   gcloud auth application-default login
   
5. Run the example

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

   python adaptive_thinking.py
   

Full source: cookbook/90_models/vertexai/claude/adaptive_thinking.py