fixed_upload.py

#!/usr/bin/env python3
"""
Fixed SmoLagent Data Analysis - Working Version
"""
import pandas as pd
from smolagents import CodeAgent, DuckDuckGoSearchTool
import warnings
warnings.filterwarnings('ignore')

# Your CSV file path
csv_file_path = "C:/Users/Cosmo/Desktop/NTU Peak Singtel/outsystems_sample_logs_6months.csv"

class FixedDataExplorer:
    """Working SmoLagent data explorer"""
    
    def __init__(self, csv_path=csv_file_path):
        self.csv_path = csv_path
        self.df = None
        self.agent = None
        self.load_data()
        self.setup_agent()
    
    def load_data(self):
        """Load the CSV data"""
        try:
            self.df = pd.read_csv(self.csv_path)
            print(f"✅ Data loaded: {self.df.shape[0]} rows, {self.df.shape[1]} columns")
            return True
        except Exception as e:
            print(f"❌ Data loading failed: {e}")
            return False
    
    def setup_agent(self):
        """Setup SmoLagent with proper model"""
        try:
            print("🤖 Setting up SmoLagent...")
            
            # Option 1: Try Ollama (if running locally)
            try:
                from smolagents import OllamaModel
                model = OllamaModel(model_id="llama2")
                self.agent = CodeAgent(
                    tools=[DuckDuckGoSearchTool()],
                    model=model
                )
                print("✅ SmoLagent configured with Ollama")
                return
            except:
                pass
            
            # Option 2: Use OpenAI (requires API key)
            try:
                from smolagents import OpenAIModel
                import os
                if os.getenv('OPENAI_API_KEY'):
                    model = OpenAIModel(model_id="gpt-3.5-turbo")
                    self.agent = CodeAgent(
                        tools=[DuckDuckGoSearchTool()],
                        model=model
                    )
                    print("✅ SmoLagent configured with OpenAI")
                    return
            except:
                pass
            
            # Option 3: Use smaller HuggingFace model (lighter download)
            try:
                from smolagents import TransformersModel
                model = TransformersModel(model_id="microsoft/DialoGPT-small")  # Smaller model
                self.agent = CodeAgent(
                    tools=[DuckDuckGoSearchTool()],
                    model=model
                )
                print("✅ SmoLagent configured with small Transformers model")
                return
            except Exception as e:
                print(f"❌ Model setup failed: {e}")
            
            print("❌ No AI model could be configured")
            print("💡 You can still use basic data analysis features")
            
        except Exception as e:
            print(f"❌ Agent setup failed: {e}")
    
    def basic_analysis(self):
        """Run basic data analysis without AI"""
        if self.df is None:
            print("❌ No data loaded")
            return
        
        print("\n📊 BASIC DATA ANALYSIS")
        print("=" * 40)
        
        # Basic stats
        print(f"📋 Dataset: {self.df.shape[0]} rows, {self.df.shape[1]} columns")
        print(f"📋 Columns: {list(self.df.columns)}")
        
        # Log level analysis
        if 'LogLevel' in self.df.columns:
            log_counts = self.df['LogLevel'].value_counts()
            print(f"\n📈 Log Level Distribution:")
            for level, count in log_counts.items():
                pct = count / len(self.df) * 100
                print(f"   {level}: {count} ({pct:.1f}%)")
        
        # Error analysis
        if 'LogLevel' in self.df.columns:
            errors = self.df[self.df['LogLevel'] == 'Error']
            if not errors.empty and 'Module' in errors.columns:
                print(f"\n🚨 Top Error Modules:")
                top_errors = errors['Module'].value_counts().head(3)
                for module, count in top_errors.items():
                    print(f"   • {module}: {count} errors")
        
        # Missing data
        missing = self.df.isnull().sum()
        print(f"\n❌ Missing Data:")
        for col, count in missing.items():
            if count > 0:
                pct = count / len(self.df) * 100
                print(f"   • {col}: {count} ({pct:.1f}%)")
    
    def ai_analysis(self, query):
        """Run AI-powered analysis"""
        if self.agent is None:
            print("❌ AI agent not available. Please configure a model first.")
            return
        
        if self.df is None:
            print("❌ No data loaded")
            return
        
        try:
            # Prepare data context for AI
            data_summary = f"""
            Dataset: {self.df.shape[0]} rows, {self.df.shape[1]} columns
            Columns: {list(self.df.columns)}
            Sample data: {self.df.head(2).to_string()}
            """
            
            full_query = f"""
            Analyze this OutSystems log data:
            {data_summary}
            
            User question: {query}
            """
            
            print(f"🤖 AI analyzing: {query}")
            response = self.agent.run(full_query)
            print(f"🤖 AI Response: {response}")
            
        except Exception as e:
            print(f"❌ AI analysis failed: {e}")

def main():
    """Main function"""
    print("🚀 FIXED SMOLAGENT DATA ANALYZER")
    print("=" * 50)
    
    # Create explorer
    explorer = FixedDataExplorer()
    
    # Run basic analysis
    explorer.basic_analysis()
    
    # Test AI if available
    if explorer.agent:
        print(f"\n🤖 AI FEATURES AVAILABLE!")
        print("   Try: explorer.ai_analysis('What are the main error types?')")
    else:
        print(f"\n💡 AI features not available - need model configuration")
    
    return explorer

if __name__ == "__main__":
    explorer = main()