Access Dependencies in Tool

How to access dependencies passed to an agent in a tool

This example demonstrates how tools can access dependencies passed to the agent, allowing tools to utilize dynamic context like user profiles and current time information for enhanced functionality.

Code

access_dependencies_in_tool.py

from typing import Dict, Any, Optional
from datetime import datetime

from agno.agent import Agent
from agno.models.openai import OpenAIChat


def get_current_context() -> dict:
    """Get current contextual information like time, weather, etc."""
    return {
        "current_time": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
        "timezone": "PST",
        "day_of_week": datetime.now().strftime("%A"),
    }

def analyze_user(user_id: str, dependencies: Optional[Dict[str, Any]] = None) -> str:
    """
    Analyze a specific user's profile and provide insights.
    
    This tool analyzes user behavior and preferences using available data sources.
    Call this tool with the user_id you want to analyze.
    
    Args:
        user_id: The user ID to analyze (e.g., 'john_doe', 'jane_smith')
        dependencies: Available data sources (automatically provided)
    
    Returns:
        Detailed analysis and insights about the user
    """
    if not dependencies:
        return "No data sources available for analysis."
    
    print(f"--> Tool received data sources: {list(dependencies.keys())}")
    
    results = [f"=== USER ANALYSIS FOR {user_id.upper()} ==="]
    
    # Use user profile data if available
    if "user_profile" in dependencies:
        profile_data = dependencies["user_profile"]
        results.append(f"Profile Data: {profile_data}")
        
        # Add analysis based on the profile
        if profile_data.get("role"):
            results.append(f"Professional Analysis: {profile_data['role']} with expertise in {', '.join(profile_data.get('preferences', []))}")
    
    # Use current context data if available
    if "current_context" in dependencies:
        context_data = dependencies["current_context"]
        results.append(f"Current Context: {context_data}")
        results.append(f"Time-based Analysis: Analysis performed on {context_data['day_of_week']} at {context_data['current_time']}")

    print(f"--> Tool returned results: {results}")
    
    return "\n\n".join(results)

# Create an agent with the analysis tool function
agent = Agent(
    model=OpenAIChat(id="gpt-4o"),
    tools=[analyze_user],
    name="User Analysis Agent",
    description="An agent specialized in analyzing users using integrated data sources.",
    instructions=[
        "You are a user analysis expert with access to user analysis tools.",
        "When asked to analyze any user, use the analyze_user tool.",
        "This tool has access to user profiles and current context through integrated data sources.",
        "After getting tool results, provide additional insights and recommendations based on the analysis.",
        "Be thorough in your analysis and explain what the tool found."
    ],
)

print("=== Tool Dependencies Access Example ===\n")

response = agent.run(
    input="Please analyze user 'john_doe' and provide insights about their professional background and preferences.",
    dependencies={
        "user_profile": {
            "name": "John Doe",
            "preferences": ["AI/ML", "Software Engineering", "Finance"],
            "location": "San Francisco, CA",
            "role": "Senior Software Engineer",
        },
        "current_context": get_current_context,
    },
    session_id="test_tool_dependencies",
)

print(f"\nAgent Response: {response.content}")

Usage

1. Create a virtual environment

   Open the Terminal and create a python virtual environment.

   Mac

   python3 -m venv .venv
   source .venv/bin/activate
   
2. Install libraries

   pip install -U agno openai
   
3. Export your OpenAI API key

   Mac/Linux

     export OPENAI_API_KEY="your_openai_api_key_here"
   
4. Create a Python file

   Create a Python file and add the above code.

   touch access_dependencies_in_tool.py
   
5. Run Agent

   Mac

   python access_dependencies_in_tool.py
   
6. Find All Cookbooks

   Explore all the available cookbooks in the Agno repository. Click the link below to view the code on GitHub:

   Agno Cookbooks on GitHub