agents/metrics_agent.py

# agents/metrics_agent.py
import logging
import time
from typing import Dict, List, Optional, Tuple, Union, Any
import json
import matplotlib.pyplot as plt
import io
import base64
from datetime import datetime, timedelta

class MetricsAgent:
    def __init__(self, metrics_calculator, token_manager=None, cache_manager=None):
        """Initialize the MetricsAgent with required utilities."""
        self.logger = logging.getLogger(__name__)
        self.metrics_calculator = metrics_calculator
        self.token_manager = token_manager
        self.cache_manager = cache_manager
        
        # Track start time
        self.start_time = datetime.now()
        
        # Store historical metrics for trend analysis
        self.metrics_history = []
        self.history_interval = 10  # seconds between history points
        self.last_history_time = self.start_time
        
        # Agent name for logging
        self.agent_name = "metrics_agent"
        
    def track_agent_performance(self) -> Dict[str, Any]:
        """
        Track and report performance metrics for all agents.
        Returns a comprehensive metrics report.
        """
        if not self.metrics_calculator:
            return {"error": "No metrics calculator available"}
            
        # Get all metrics from calculator
        all_metrics = self.metrics_calculator.get_all_metrics()
        
        # Add additional derived metrics
        derived_metrics = self._calculate_derived_metrics(all_metrics)
        all_metrics["derived_metrics"] = derived_metrics
        
        # Add timestamp
        all_metrics["timestamp"] = datetime.now().isoformat()
        all_metrics["elapsed_time"] = (datetime.now() - self.start_time).total_seconds()
        
        # Add to history if enough time has passed
        if (datetime.now() - self.last_history_time).total_seconds() >= self.history_interval:
            self.metrics_history.append(all_metrics)
            self.last_history_time = datetime.now()
        
        return all_metrics
    
    def _calculate_derived_metrics(self, metrics: Dict[str, Any]) -> Dict[str, Any]:
        """Calculate additional derived metrics from raw metrics."""
        derived = {}
        
        # Calculate energy efficiency
        if metrics.get("token_metrics", {}).get("total_tokens", 0) > 0:
            energy_total = metrics.get("energy_usage", {}).get("total", 0.001)
            tokens_total = metrics.get("token_metrics", {}).get("total_tokens", 0)
            
            derived["energy_per_token"] = energy_total / tokens_total
            derived["tokens_per_watt_hour"] = tokens_total / energy_total
        
        # Calculate optimization effectiveness
        if metrics.get("optimization_gains", {}):
            opt_gains = metrics.get("optimization_gains", {})
            total_energy = metrics.get("energy_usage", {}).get("total", 0)
            
            if total_energy > 0:
                energy_saved = opt_gains.get("total_energy_saved", 0)
                derived["optimization_effectiveness"] = energy_saved / (total_energy + energy_saved)
            
            # Breakdown by optimization type
            derived["optimization_breakdown"] = {
                "caching": opt_gains.get("cache_energy_saved", 0),
                "model_selection": opt_gains.get("model_selection_energy_saved", 0)
            }
        
        # Calculate cache effectiveness
        if self.cache_manager:
            cache_stats = self.cache_manager.get_stats()
            derived["cache_effectiveness"] = {
                "hit_rate": cache_stats.get("hit_rate", 0),
                "memory_efficiency": cache_stats.get("current_size", 0) / max(cache_stats.get("max_size", 1), 1)
            }
        
        return derived
    
    def generate_sustainability_report(self) -> Dict[str, Any]:
        """
        Generate a comprehensive sustainability report.
        Includes environmental impact metrics and recommendations.
        """
        # Get current metrics
        metrics = self.track_agent_performance()
        
        # Calculate key sustainability indicators
        carbon_footprint = metrics.get("carbon_footprint_kg", 0)
        energy_usage = metrics.get("energy_usage", {}).get("total", 0)
        optimization_gains = metrics.get("optimization_gains", {})
        
        # Generate environmental equivalents
        env_equivalents = metrics.get("environmental_equivalents", {})
        
        # Calculate potential future savings
        potential_savings = self._calculate_potential_savings(metrics)
        
        # Generate recommendations
        recommendations = self._generate_optimization_recommendations(metrics)
        
        # Prepare report
        report = {
            "timestamp": datetime.now().isoformat(),
            "runtime_seconds": metrics.get("runtime_seconds", 0),
            "sustainability_metrics": {
                "energy_usage_wh": energy_usage,
                "carbon_footprint_kg": carbon_footprint,
                "environmental_equivalents": env_equivalents
            },
            "optimization_results": {
                "energy_saved_wh": optimization_gains.get("total_energy_saved", 0),
                "tokens_saved": optimization_gains.get("tokens_saved", 0),
                "tokens_saved_percent": optimization_gains.get("tokens_saved_pct", 0)
            },
            "potential_future_savings": potential_savings,
            "recommendations": recommendations
        }
        
        return report
    
    def _calculate_potential_savings(self, metrics: Dict[str, Any]) -> Dict[str, Any]:
        """Calculate potential future energy savings based on current patterns."""
        # Get current optimization effectiveness
        derived = metrics.get("derived_metrics", {})
        current_effectiveness = derived.get("optimization_effectiveness", 0)
        
        # Calculate potential improvements
        cache_hit_rate = derived.get("cache_effectiveness", {}).get("hit_rate", 0)
        potential_hit_rate = min(cache_hit_rate + 0.2, 0.95)  # Assume we can improve hit rate by up to 20%
        
        # Calculate energy that could be saved with improved caching
        energy_usage = metrics.get("energy_usage", {}).get("total", 0)
        potential_cache_savings = energy_usage * 0.3 * (potential_hit_rate - cache_hit_rate) / max(cache_hit_rate, 0.01)
        
        # Calculate potential model selection improvements
        model_selection_current = derived.get("optimization_breakdown", {}).get("model_selection", 0)
        potential_model_selection = model_selection_current * 1.5  # Assume 50% improvement
        
        return {
            "improved_caching_wh": potential_cache_savings,
            "improved_model_selection_wh": potential_model_selection - model_selection_current,
            "total_potential_wh": potential_cache_savings + (potential_model_selection - model_selection_current),
            "percent_improvement": (potential_cache_savings + (potential_model_selection - model_selection_current)) / max(energy_usage, 0.001) * 100
        }
    
    def _generate_optimization_recommendations(self, metrics: Dict[str, Any]) -> List[Dict[str, Any]]:
        """Generate actionable recommendations for improving efficiency."""
        recommendations = []
        
        # Check token usage by model
        token_metrics = metrics.get("token_metrics", {})
        by_model = token_metrics.get("by_model", {})
        
        # Find the most token-heavy model
        if by_model:
            heaviest_model = max(by_model.items(), key=lambda x: x[1])
            if heaviest_model[1] > 1000:
                recommendations.append({
                    "type": "token_optimization",
                    "target": heaviest_model[0],
                    "description": f"Optimize prompts for {heaviest_model[0]} to reduce token usage",
                    "potential_impact": "Medium"
                })
        
        # Check cache effectiveness
        derived = metrics.get("derived_metrics", {})
        cache_effectiveness = derived.get("cache_effectiveness", {})
        
        if cache_effectiveness.get("hit_rate", 1) < 0.5:
            recommendations.append({
                "type": "cache_optimization",
                "description": "Increase cache size or improve cache key generation for better hit rates",
                "potential_impact": "High"
            })
        
        # Check model selection
        model_selection = derived.get("optimization_breakdown", {}).get("model_selection", 0)
        energy_usage = metrics.get("energy_usage", {}).get("total", 0)
        
        if model_selection < 0.1 * energy_usage:
            recommendations.append({
                "type": "model_selection",
                "description": "Refine complexity detection to use smaller models more frequently",
                "potential_impact": "High"
            })
        
        return recommendations
    
    def generate_metrics_dashboard(self, format: str = "json") -> Dict[str, Any]:
        """
        Generate a visual or structured representation of system metrics.
        Returns data in the specified format (json, html, image_base64).
        """
        # Get current metrics
        metrics = self.track_agent_performance()
        
        if format == "json":
            return metrics
        
        elif format == "image_base64":
            # Generate visualization
            img_data = self._generate_dashboard_visualization(metrics)
            return {
                "format": "image_base64",
                "data": img_data,
                "timestamp": datetime.now().isoformat()
            }
        
        elif format == "html":
            # Generate HTML representation
            html_data = self._generate_html_dashboard(metrics)
            return {
                "format": "html",
                "data": html_data,
                "timestamp": datetime.now().isoformat()
            }
        
        else:
            return {"error": f"Unsupported format: {format}"}
    
    def _generate_dashboard_visualization(self, metrics: Dict[str, Any]) -> str:
        """Generate a visualization of metrics and return as base64 image."""
        # Create figure with subplots
        fig, axs = plt.subplots(2, 2, figsize=(12, 10))
        
        # 1. Energy Usage by Model
        energy_by_model = metrics.get("energy_usage", {}).get("by_model", {})
        if energy_by_model:
            models = list(energy_by_model.keys())
            energy_values = list(energy_by_model.values())
            
            axs[0, 0].bar(models, energy_values)
            axs[0, 0].set_title('Energy Usage by Model')
            axs[0, 0].set_ylabel('Watt-hours')
            axs[0, 0].tick_params(axis='x', rotation=45)
        
        # 2. Token Usage by Agent
        token_by_agent = metrics.get("token_metrics", {}).get("by_agent", {})
        if token_by_agent:
            agents = list(token_by_agent.keys())
            token_values = list(token_by_agent.values())
            
            axs[0, 1].bar(agents, token_values)
            axs[0, 1].set_title('Token Usage by Agent')
            axs[0, 1].set_ylabel('Tokens')
            axs[0, 1].tick_params(axis='x', rotation=45)
        
        # 3. Optimization Gains
        opt_gains = metrics.get("optimization_gains", {})
        if opt_gains:
            gain_types = ['Tokens Saved', 'Cache Energy Saved', 'Model Selection Energy Saved']
            gain_values = [
                opt_gains.get("tokens_saved", 0) / 100,  # Scale down for visualization
                opt_gains.get("cache_energy_saved", 0),
                opt_gains.get("model_selection_energy_saved", 0)
            ]
            
            axs[1, 0].bar(gain_types, gain_values)
            axs[1, 0].set_title('Optimization Gains')
            axs[1, 0].set_ylabel('Value')
            axs[1, 0].tick_params(axis='x', rotation=45)
        
        # 4. Environmental Impact
        env_equiv = metrics.get("environmental_equivalents", {})
        if env_equiv:
            impact_types = list(env_equiv.keys())
            impact_values = list(env_equiv.values())
            
            axs[1, 1].bar(impact_types, impact_values)
            axs[1, 1].set_title('Environmental Equivalents')
            axs[1, 1].set_ylabel('Value')
            axs[1, 1].tick_params(axis='x', rotation=45)
        
        # Adjust layout and convert to base64
        plt.tight_layout()
        
        # Save to bytes buffer
        buffer = io.BytesIO()
        plt.savefig(buffer, format='png')
        buffer.seek(0)
        
        # Convert to base64
        img_str = base64.b64encode(buffer.getvalue()).decode('utf-8')
        
        # Close plot to free memory
        plt.close(fig)
        
        return img_str
    
    def _generate_html_dashboard(self, metrics: Dict[str, Any]) -> str:
        """Generate an HTML representation of the metrics dashboard."""
        # Simple HTML template
        html_template = """
        <!DOCTYPE html>
        <html>
        <head>
            <title>Sustainability Metrics Dashboard</title>
            <style>
                body { font-family: Arial, sans-serif; margin: 20px; }
                .dashboard { display: grid; grid-template-columns: 1fr 1fr; gap: 20px; }
                .card { border: 1px solid #ddd; border-radius: 5px; padding: 15px; }
                .card h2 { margin-top: 0; color: #333; }
                .metric { margin: 10px 0; }
                .metric-name { font-weight: bold; }
                .metric-value { float: right; }
                .highlight { color: #2a6496; font-weight: bold; }
                .recommendations { grid-column: span 2; }
            </style>
        </head>
        <body>
            <h1>Sustainability Metrics Dashboard</h1>
            <p>Generated on: {timestamp}</p>
            
            <div class="dashboard">
                <div class="card">
                    <h2>Energy Usage</h2>
                    {energy_metrics}
                </div>
                
                <div class="card">
                    <h2>Token Usage</h2>
                    {token_metrics}
                </div>
                
                <div class="card">
                    <h2>Optimization Gains</h2>
                    {optimization_metrics}
                </div>
                
                <div class="card">
                    <h2>Environmental Impact</h2>
                    {environmental_metrics}
                </div>
                
                <div class="card recommendations">
                    <h2>Recommendations</h2>
                    {recommendations}
                </div>
            </div>
        </body>
        </html>
        """
        
        # Generate energy metrics HTML
        energy_usage = metrics.get("energy_usage", {})
        energy_metrics_html = ""
        if energy_usage:
            energy_metrics_html += self._format_metric("Total Energy", f"{energy_usage.get('total', 0):.6f} Wh")
            energy_metrics_html += "<h3>By Model</h3>"
            for model, usage in energy_usage.get("by_model", {}).items():
                energy_metrics_html += self._format_metric(model, f"{usage:.6f} Wh")
        
        # Generate token metrics HTML
        token_metrics = metrics.get("token_metrics", {})
        token_metrics_html = ""
        if token_metrics:
            token_metrics_html += self._format_metric("Total Tokens", token_metrics.get("total_tokens", 0))
            token_metrics_html += self._format_metric("Tokens Saved", token_metrics.get("tokens_saved", 0))
            token_metrics_html += "<h3>By Agent</h3>"
            for agent, tokens in token_metrics.get("by_agent", {}).items():
                token_metrics_html += self._format_metric(agent, tokens)
        
        # Generate optimization metrics HTML
        opt_gains = metrics.get("optimization_gains", {})
        optimization_metrics_html = ""
        if opt_gains:
            optimization_metrics_html += self._format_metric("Tokens Saved", opt_gains.get("tokens_saved", 0))
            optimization_metrics_html += self._format_metric("Tokens Saved %", f"{opt_gains.get('tokens_saved_pct', 0):.2f}%")
            optimization_metrics_html += self._format_metric("Cache Energy Saved", f"{opt_gains.get('cache_energy_saved', 0):.6f} Wh")
            optimization_metrics_html += self._format_metric("Model Selection Saved", f"{opt_gains.get('model_selection_energy_saved', 0):.6f} Wh")
            optimization_metrics_html += self._format_metric("Total Energy Saved", f"{opt_gains.get('total_energy_saved', 0):.6f} Wh", highlight=True)
        
        # Generate environmental metrics HTML
        env_equiv = metrics.get("environmental_equivalents", {})
        environmental_metrics_html = ""
        if env_equiv:
            environmental_metrics_html += self._format_metric("Carbon Footprint", f"{metrics.get('carbon_footprint_kg', 0):.6f} kg CO₂")
            environmental_metrics_html += "<h3>Equivalents</h3>"
            for impact_type, value in env_equiv.items():
                label = impact_type.replace('_', ' ').title()
                environmental_metrics_html += self._format_metric(label, f"{value:.2f}")
        
        # Generate recommendations HTML
        recommendations = metrics.get("recommendations", [])
        recommendations_html = "<ul>"
        if recommendations:
            for rec in recommendations:
                impact = rec.get("potential_impact", "")
                impact_class = "highlight" if impact == "High" else ""
                recommendations_html += f"<li><span class='{impact_class}'>{rec.get('description', '')}</span> (Impact: {impact})</li>"
        else:
            recommendations_html += "<li>No recommendations available</li>"
        recommendations_html += "</ul>"
        
        # Format final HTML
        html = html_template.format(
            timestamp=datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
            energy_metrics=energy_metrics_html,
            token_metrics=token_metrics_html,
            optimization_metrics=optimization_metrics_html,
            environmental_metrics=environmental_metrics_html,
            recommendations=recommendations_html
        )
        
        return html

    def _format_metric(self, name: str, value: Any, highlight: bool = False) -> str:
        """Helper method to format a metric as HTML."""
        highlight_class = "highlight" if highlight else ""
        return f"<div class='metric'><span class='metric-name'>{name}:</span> <span class='metric-value {highlight_class}'>{value}</span></div>"