import pandas as pd
from ai_agent_sdk import DataProcessingAgent
from datetime import datetime

# Initialize the data processing agent
agent = DataProcessingAgent(
    name="sales_data_processor",
    model="gpt-4-turbo",
    capabilities=["cleaning", "enrichment", "validation"]
)

# Configure processing pipeline
pipeline_config = {
    "data_sources": [
        {"type": "database", "connection": "postgresql://sales_db"},
        {"type": "api", "endpoint": "https://api.company.com/products"},
        {"type": "file", "path": "s3://bucket/customer_data/*.csv"}
    ],
    "processing_rules": {
        "missing_values": "smart_impute",
        "outliers": "flag_and_investigate",
        "duplicates": "merge_with_latest",
        "validation": "business_rules.yaml"
    },
    "output": {
        "format": "parquet",
        "destination": "s3://processed-data/",
        "partitioning": ["date", "region"]
    }
}

# Execute automated processing
async def process_sales_data():
    # Agent analyzes data structure and quality
    analysis = await agent.analyze_data_quality(pipeline_config["data_sources"])
    
    print(f"Data Quality Score: {analysis['quality_score']}")
    print(f"Issues Found: {len(analysis['issues'])}")
    
    # Agent creates optimized processing plan
    processing_plan = await agent.create_processing_plan(
        sources=pipeline_config["data_sources"],
        quality_report=analysis,
        business_context="Prepare data for quarterly sales analysis"
    )
    
    # Execute the processing with monitoring
    result = await agent.execute_pipeline(
        plan=processing_plan,
        config=pipeline_config,
        monitor_callback=lambda status: print(f"Progress: {status}")
    )
    
    # Generate data quality report
    quality_report = await agent.generate_quality_report(result)
    
    return {
        "records_processed": result["total_records"],
        "processing_time": result["duration"],
        "quality_metrics": quality_report,
        "output_location": result["output_path"]
    }

# Natural language data exploration
async def explore_processed_data():
    # Agent can answer questions about the data
    insights = await agent.analyze(
        query="What are the top factors affecting sales decline in Q4?",
        data_path="s3://processed-data/latest/",
        include_visualizations=True
    )
    
    print(f"Key Findings: {insights['summary']}")
    for viz in insights['visualizations']:
        print(f"Chart: {viz['title']} - {viz['url']}")

# Run the processing
if __name__ == "__main__":
    import asyncio
    
    # Process data
    results = asyncio.run(process_sales_data())
    print(f"\nProcessed {results['records_processed']:,} records in {results['processing_time']}")
    
    # Explore insights
    asyncio.run(explore_processed_data())