Comprehensive comparison for Prompt Engineering technology in AI applications

import agenta as ag
from openai import OpenAI
import logging
from typing import List, Dict, Optional

# Initialize Agenta configuration for prompt engineering
ag.init()

# Configure logging for production monitoring
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

# Define configuration schema for A/B testing different prompts
ag.config.default(
    temperature=ag.FloatParam(default=0.7, minval=0.0, maxval=1.0),
    max_tokens=ag.IntParam(default=500, minval=100, maxval=2000),
    model=ag.MultipleChoiceParam(
        default="gpt-4",
        choices=["gpt-3.5-turbo", "gpt-4", "gpt-4-turbo"]
    ),
    system_prompt=ag.TextParam(
        default="You are an expert customer support agent for an e-commerce platform. Provide helpful, professional, and empathetic responses."
    ),
    response_format=ag.MultipleChoiceParam(
        default="detailed",
        choices=["concise", "detailed", "step-by-step"]
    )
)

@ag.entrypoint
def generate_customer_support_response(
    customer_query: str,
    order_id: Optional[str] = None,
    customer_history: Optional[List[Dict]] = None
) -> str:
    """
    Generate AI-powered customer support responses with configurable prompts.
    
    Args:
        customer_query: The customer's question or issue
        order_id: Optional order ID for context
        customer_history: Optional list of previous interactions
    
    Returns:
        AI-generated support response
    """
    try:
        # Initialize OpenAI client
        client = OpenAI()
        
        # Build context from customer history
        context = ""
        if order_id:
            context += f"\nOrder ID: {order_id}"
        
        if customer_history and len(customer_history) > 0:
            context += "\n\nPrevious interactions:\n"
            for interaction in customer_history[-3:]:  # Last 3 interactions
                context += f"- {interaction.get('summary', '')}\n"
        
        # Construct user message based on response format preference
        format_instructions = {
            "concise": "Provide a brief, direct answer in 2-3 sentences.",
            "detailed": "Provide a comprehensive response with all relevant details.",
            "step-by-step": "Break down your response into clear, numbered steps."
        }
        
        user_message = f"{customer_query}\n{context}\n\nResponse format: {format_instructions[ag.config.response_format]}"
        
        # Log request for monitoring
        logger.info(f"Processing customer query with model: {ag.config.model}")
        
        # Make API call with configured parameters
        response = client.chat.completions.create(
            model=ag.config.model,
            messages=[
                {"role": "system", "content": ag.config.system_prompt},
                {"role": "user", "content": user_message}
            ],
            temperature=ag.config.temperature,
            max_tokens=ag.config.max_tokens
        )
        
        # Extract and validate response
        ai_response = response.choices[0].message.content
        
        if not ai_response or len(ai_response.strip()) == 0:
            raise ValueError("Empty response received from AI model")
        
        logger.info("Successfully generated customer support response")
        return ai_response.strip()
    
    except Exception as e:
        logger.error(f"Error generating response: {str(e)}")
        # Return fallback message for production resilience
        return "We apologize for the inconvenience. Our team will review your query and respond within 24 hours. Please contact [email protected] for urgent matters."

# Example usage for testing
if __name__ == "__main__":
    test_query = "I received a damaged product and want to return it. What's the process?"
    test_order_id = "ORD-12345"
    test_history = [
        {"summary": "Customer inquired about shipping times"},
        {"summary": "Order was delayed, customer notified"}
    ]
    
    result = generate_customer_support_response(
        customer_query=test_query,
        order_id=test_order_id,
        customer_history=test_history
    )
    
    print(f"AI Response:\n{result}")