agent/rica_agent.py

"""
RICA Agent optimized for Hugging Face Spaces
Fixed for latest smolagents API
"""

import os
from smolagents import CodeAgent
from smolagents.models import OpenAIServerModel
from agent_tools.ml_tools import predict_customer_churn_hf, get_model_status

def create_rica_agent_hf():
    """Create RICA agent using correct smolagents API"""
    
    # Check API key availability
    api_key = os.getenv("OPENAI_API_KEY")
    if not api_key:
        raise ValueError("OpenAI API key not configured")
    
    # Initialize OpenAI model using smolagents model class
    model = OpenAIServerModel(
        model_id="gpt-3.5-turbo",
        api_key=api_key
    )
    
    # HF Spaces optimized tools
    hf_tools = [
        predict_customer_churn_hf,
        get_model_status
    ]
    
    try:
        # Create agent with correct parameters (no max_iterations)
        agent = CodeAgent(
            tools=hf_tools,
            model=model,
            add_base_tools=False
        )
        return agent
    except Exception as e:
        raise Exception(f"Agent creation failed: {str(e)}")

def execute_rica_analysis_hf(analysis_type: str, parameters: dict = None):
    """Execute RICA analysis optimized for HF Spaces"""
    
    # Check API key first
    if not os.getenv("OPENAI_API_KEY"):
        return "Error: OpenAI API key not configured. Please set your API key in the sidebar."
    
    try:
        agent = create_rica_agent_hf()
        
        # Simplified goals for HF Spaces
        hf_goals = {
            "comprehensive": f"""
            Execute business intelligence analysis:
            1) Check model status with get_model_status()
            2) Predict customer churn with predict_customer_churn_hf()
            3) Provide executive summary with key insights and recommendations
            
            Focus on actionable insights for business decision-making.
            Parameters: {parameters}
            """,
            
            "churn_focus": f"""
            Focus on customer churn analysis:
            1) Predict customer churn with predict_customer_churn_hf(risk_threshold={parameters.get('risk_threshold', 0.6) if parameters else 0.6})
            2) Identify high-risk customers requiring immediate attention
            3) Provide specific intervention strategies
            
            Parameters: {parameters}
            """,
            
            "quick_insights": f"""
            Provide quick business insights:
            1) Get model status with get_model_status()
            2) Run limited churn analysis with predict_customer_churn_hf()
            3) Summarize top 3 business priorities
            
            Parameters: {parameters}
            """
        }
        
        goal = hf_goals.get(analysis_type, hf_goals["comprehensive"])
        
        result = agent.run(goal)
        return result
        
    except Exception as e:
        return f"Analysis failed: {str(e)}. Please check your API key and model status."