"""
Pattern 174: Performance Optimization Pattern with LangChain/LangGraph

Purpose: Optimize agent performance through profiling, tuning, and LLM-powered insights.
"""

from typing import Dict, Any, List, Optional, Callable, TypedDict
from dataclasses import dataclass, field
from datetime import datetime, timedelta
import time
from functools import wraps
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
import statistics


class OptimizationStrategy(Enum):
    """Performance optimization strategies."""
    CACHING = "caching"
    BATCHING = "batching"
    PARALLELIZATION = "parallelization"
    LAZY_LOADING = "lazy_loading"
    INDEX_OPTIMIZATION = "index_optimization"


class BottleneckType(Enum):
    """Types of performance bottlenecks."""
    CPU_BOUND = "cpu_bound"
    IO_BOUND = "io_bound"
    MEMORY_BOUND = "memory_bound"
    NETWORK_BOUND = "network_bound"
    DATABASE_BOUND = "database_bound"


class ProfileLevel(Enum):
    """Profiling detail levels."""
    BASIC = "basic"
    DETAILED = "detailed"
    COMPREHENSIVE = "comprehensive"


class PriorityLevel(Enum):
    """Optimization priority levels."""
    LOW = 1
    MEDIUM = 2
    HIGH = 3
    CRITICAL = 4


@dataclass
class PerformanceProfile:
    """Performance profile for an operation."""
    operation: str
    execution_count: int = 0
    total_time: float = 0.0
    min_time: float = float('inf')
    max_time: float = 0.0
    errors: int = 0
    last_execution: Optional[datetime] = None
    
    def avg_time(self) -> float:
        """Average execution time."""
        return self.total_time / self.execution_count if self.execution_count > 0 else 0.0
    
    def record(self, duration: float, error: bool = False):
        """Record an execution."""
        self.execution_count += 1
        self.total_time += duration
        self.min_time = min(self.min_time, duration)
        self.max_time = max(self.max_time, duration)
        self.last_execution = datetime.now()
        if error:
            self.errors += 1


@dataclass
class Bottleneck:
    """Identified performance bottleneck."""
    operation: str
    bottleneck_type: BottleneckType
    severity: float  # 0.0 to 1.0
    impact: str
    measured_time_ms: float
    expected_time_ms: float


@dataclass
class OptimizationSuggestion:
    """Performance optimization suggestion."""
    operation: str
    strategy: OptimizationStrategy
    issue: str
    suggestion: str
    priority: PriorityLevel
    expected_improvement: str


@dataclass
class OptimizationResult:
    """Result of applying an optimization."""
    operation: str
    strategy: OptimizationStrategy
    before_avg_ms: float
    after_avg_ms: float
    improvement_percent: float
    success: bool


@dataclass
class PerformanceReport:
    """Comprehensive performance report."""
    total_operations: int
    total_execution_time_ms: float
    avg_execution_time_ms: float
    bottlenecks: List[Bottleneck]
    suggestions: List[OptimizationSuggestion]
    top_slowest_operations: List[str]


class OptimizationState(TypedDict):
    """State for optimization workflow."""
    profiles: Dict[str, Dict[str, Any]]
    bottlenecks: List[Dict[str, Any]]
    analysis: str
    suggestions: List[str]
    strategies: List[str]


class PerformanceOptimizer:
    """
    Advanced performance optimizer with LangChain/LangGraph integration.
    """
    
    def __init__(self, model_name: str = "gpt-4", temperature: float = 0.3):
        self.llm = ChatOpenAI(model=model_name, temperature=temperature)
        self.profiles: Dict[str, PerformanceProfile] = {}
        self.bottlenecks: List[Bottleneck] = []
        self.optimizations: List[OptimizationResult] = []
        self.workflow = self._build_optimization_workflow()
        
        # Statistics
        self.stats = {
            "total_operations_profiled": 0,
            "total_optimizations_applied": 0,
            "bottlenecks_identified": 0,
            "avg_improvement_percent": 0.0,
            "strategies_used": {strategy: 0 for strategy in OptimizationStrategy}
        }
    
    def _build_optimization_workflow(self) -> StateGraph:
        """Build LangGraph workflow for performance optimization."""
        workflow = StateGraph(OptimizationState)
        
        # Add nodes
        workflow.add_node("analyze_profiles", self._analyze_profiles_node)
        workflow.add_node("identify_bottlenecks", self._identify_bottlenecks_node)
        workflow.add_node("generate_insights", self._generate_insights_node)
        workflow.add_node("recommend_strategies", self._recommend_strategies_node)
        workflow.add_node("prioritize_optimizations", self._prioritize_optimizations_node)
        
        # Define edges
        workflow.set_entry_point("analyze_profiles")
        workflow.add_edge("analyze_profiles", "identify_bottlenecks")
        workflow.add_edge("identify_bottlenecks", "generate_insights")
        workflow.add_edge("generate_insights", "recommend_strategies")
        workflow.add_edge("recommend_strategies", "prioritize_optimizations")
        workflow.add_edge("prioritize_optimizations", END)
        
        return workflow.compile()
    
    def _analyze_profiles_node(self, state: OptimizationState) -> OptimizationState:
        """Analyze performance profiles."""
        profiles_data = {}
        
        for operation, profile in self.profiles.items():
            profiles_data[operation] = {
                "execution_count": profile.execution_count,
                "avg_time_ms": profile.avg_time() * 1000,
                "min_time_ms": profile.min_time * 1000,
                "max_time_ms": profile.max_time * 1000,
                "error_rate": (profile.errors / profile.execution_count * 100) 
                             if profile.execution_count > 0 else 0.0
            }
        
        state["profiles"] = profiles_data
        return state
    
    def _identify_bottlenecks_node(self, state: OptimizationState) -> OptimizationState:
        """Identify performance bottlenecks."""
        profiles = state.get("profiles", {})
        bottlenecks_data = []
        
        # Identify slow operations
        if profiles:
            avg_times = [p["avg_time_ms"] for p in profiles.values()]
            if len(avg_times) > 1:
                mean_time = statistics.mean(avg_times)
                stdev_time = statistics.stdev(avg_times)
                
                for operation, profile in profiles.items():
                    avg_time = profile["avg_time_ms"]
                    
                    # Operation is significantly slower than average
                    if avg_time > mean_time + 2 * stdev_time:
                        bottlenecks_data.append({
                            "operation": operation,
                            "avg_time_ms": avg_time,
                            "deviation": avg_time - mean_time,
                            "severity": min((avg_time - mean_time) / mean_time, 1.0)
                        })
        
        state["bottlenecks"] = bottlenecks_data
        return state
    
    def _generate_insights_node(self, state: OptimizationState) -> OptimizationState:
        """Generate insights using LLM."""
        profiles = state.get("profiles", {})
        bottlenecks = state.get("bottlenecks", [])
        
        # Use LLM to analyze performance data
        prompt = ChatPromptTemplate.from_template(
            "Analyze the following performance data and provide insights:\n"
            "Performance Profiles: {profiles_summary}\n"
            "Identified Bottlenecks: {bottlenecks_summary}\n\n"
            "Provide detailed insights about performance patterns and issues."
        )
        
        chain = prompt | self.llm
        response = chain.invoke({
            "profiles_summary": f"{len(profiles)} operations profiled",
            "bottlenecks_summary": f"{len(bottlenecks)} bottlenecks identified"
        })
        
        state["analysis"] = str(response.content)
        return state
    
    def _recommend_strategies_node(self, state: OptimizationState) -> OptimizationState:
        """Recommend optimization strategies using LLM."""
        bottlenecks = state.get("bottlenecks", [])
        analysis = state.get("analysis", "")
        
        suggestions = []
        
        if bottlenecks:
            # Use LLM to recommend strategies
            prompt = ChatPromptTemplate.from_template(
                "Based on the performance analysis, recommend specific optimization strategies:\n"
                "Analysis: {analysis}\n"
                "Bottlenecks: {bottlenecks}\n\n"
                "Provide 3-5 specific, actionable optimization strategies."
            )
            
            chain = prompt | self.llm
            response = chain.invoke({
                "analysis": analysis,
                "bottlenecks": str(bottlenecks)
            })
            
            # Parse suggestions
            suggestions = [
                line.strip()
                for line in str(response.content).split('\n')
                if line.strip() and len(line.strip()) > 10
            ][:5]
        
        state["suggestions"] = suggestions
        return state
    
    def _prioritize_optimizations_node(self, state: OptimizationState) -> OptimizationState:
        """Prioritize optimization strategies."""
        suggestions = state.get("suggestions", [])
        bottlenecks = state.get("bottlenecks", [])
        
        # Map common strategies
        strategies = []
        
        for bottleneck in bottlenecks:
            severity = bottleneck.get("severity", 0.0)
            if severity > 0.7:
                strategies.append(f"CRITICAL: Optimize '{bottleneck['operation']}'")
            elif severity > 0.4:
                strategies.append(f"HIGH: Improve '{bottleneck['operation']}'")
        
        # Add LLM suggestions
        strategies.extend(suggestions)
        
        state["strategies"] = strategies[:10]  # Top 10 strategies
        return state
    
    def profile_operation(self, operation: str) -> Callable:
        """Decorator to profile an operation."""
        def decorator(func: Callable) -> Callable:
            @wraps(func)
            def wrapper(*args, **kwargs):
                start_time = time.time()
                error_occurred = False
                
                try:
                    result = func(*args, **kwargs)
                    return result
                except Exception as e:
                    error_occurred = True
                    raise e
                finally:
                    duration = time.time() - start_time
                    
                    # Record profile
                    if operation not in self.profiles:
                        self.profiles[operation] = PerformanceProfile(operation=operation)
                    
                    self.profiles[operation].record(duration, error_occurred)
                    self.stats["total_operations_profiled"] += 1
            
            return wrapper
        return decorator
    
    def record_execution(
        self,
        operation: str,
        duration_ms: float,
        error: bool = False
    ):
        """Manually record an operation execution."""
        if operation not in self.profiles:
            self.profiles[operation] = PerformanceProfile(operation=operation)
        
        self.profiles[operation].record(duration_ms / 1000.0, error)
        self.stats["total_operations_profiled"] += 1
    
    def identify_bottleneck(
        self,
        operation: str,
        bottleneck_type: BottleneckType,
        measured_time_ms: float,
        expected_time_ms: float
    ) -> Bottleneck:
        """Manually identify a bottleneck."""
        severity = min((measured_time_ms - expected_time_ms) / expected_time_ms, 1.0)
        
        bottleneck = Bottleneck(
            operation=operation,
            bottleneck_type=bottleneck_type,
            severity=max(0.0, severity),
            impact=f"{measured_time_ms - expected_time_ms:.2f}ms slower than expected",
            measured_time_ms=measured_time_ms,
            expected_time_ms=expected_time_ms
        )
        
        self.bottlenecks.append(bottleneck)
        self.stats["bottlenecks_identified"] += 1
        
        return bottleneck
    
    def apply_optimization(
        self,
        operation: str,
        strategy: OptimizationStrategy,
        before_avg_ms: float,
        after_avg_ms: float
    ) -> OptimizationResult:
        """Record an applied optimization."""
        improvement = ((before_avg_ms - after_avg_ms) / before_avg_ms * 100) if before_avg_ms > 0 else 0.0
        
        result = OptimizationResult(
            operation=operation,
            strategy=strategy,
            before_avg_ms=before_avg_ms,
            after_avg_ms=after_avg_ms,
            improvement_percent=improvement,
            success=improvement > 0
        )
        
        self.optimizations.append(result)
        self.stats["total_optimizations_applied"] += 1
        self.stats["strategies_used"][strategy] += 1
        
        # Update average improvement
        if self.optimizations:
            self.stats["avg_improvement_percent"] = statistics.mean(
                opt.improvement_percent for opt in self.optimizations
            )
        
        return result
    
    def analyze_performance(self) -> PerformanceReport:
        """Analyze performance using workflow."""
        initial_state: OptimizationState = {
            "profiles": {},
            "bottlenecks": [],
            "analysis": "",
            "suggestions": [],
            "strategies": []
        }
        
        # Run workflow
        result = self.workflow.invoke(initial_state)
        
        # Create report
        total_ops = len(self.profiles)
        total_time = sum(p.total_time for p in self.profiles.values()) * 1000
        avg_time = total_time / total_ops if total_ops > 0 else 0.0
        
        # Get top slowest operations
        sorted_profiles = sorted(
            self.profiles.items(),
            key=lambda x: x[1].avg_time(),
            reverse=True
        )
        top_slowest = [op for op, _ in sorted_profiles[:5]]
        
        # Convert suggestions to proper format
        suggestions = [
            OptimizationSuggestion(
                operation="general",
                strategy=OptimizationStrategy.CACHING,
                issue="Performance improvement needed",
                suggestion=sugg,
                priority=PriorityLevel.MEDIUM,
                expected_improvement="TBD"
            )
            for sugg in result["suggestions"]
        ]
        
        return PerformanceReport(
            total_operations=total_ops,
            total_execution_time_ms=total_time,
            avg_execution_time_ms=avg_time,
            bottlenecks=self.bottlenecks,
            suggestions=suggestions,
            top_slowest_operations=top_slowest
        )
    
    def get_profile(self, operation: str) -> Optional[PerformanceProfile]:
        """Get profile for an operation."""
        return self.profiles.get(operation)
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get optimization statistics."""
        return {
            "total_operations_profiled": self.stats["total_operations_profiled"],
            "unique_operations": len(self.profiles),
            "total_optimizations_applied": self.stats["total_optimizations_applied"],
            "bottlenecks_identified": self.stats["bottlenecks_identified"],
            "avg_improvement_percent": round(self.stats["avg_improvement_percent"], 2),
            "strategies_used": {
                strategy.value: count
                for strategy, count in self.stats["strategies_used"].items()
                if count > 0
            },
            "successful_optimizations": len([o for o in self.optimizations if o.success])
        }


class PerformanceOptimizationPattern:
    """
    Comprehensive performance optimization pattern coordinator with LangChain/LangGraph.
    """
    
    def __init__(self, model_name: str = "gpt-4"):
        self.optimizer = PerformanceOptimizer(model_name=model_name)
    
    def profile(self, operation: str) -> Callable:
        """Profile an operation using decorator."""
        return self.optimizer.profile_operation(operation)
    
    def record(
        self,
        operation: str,
        duration_ms: float,
        error: bool = False
    ):
        """Record operation execution."""
        self.optimizer.record_execution(operation, duration_ms, error)
    
    def identify_bottleneck(
        self,
        operation: str,
        bottleneck_type: BottleneckType,
        measured_ms: float,
        expected_ms: float
    ) -> Bottleneck:
        """Identify a performance bottleneck."""
        return self.optimizer.identify_bottleneck(
            operation, bottleneck_type, measured_ms, expected_ms
        )
    
    def optimize(
        self,
        operation: str,
        strategy: OptimizationStrategy,
        before_ms: float,
        after_ms: float
    ) -> OptimizationResult:
        """Record an optimization."""
        return self.optimizer.apply_optimization(
            operation, strategy, before_ms, after_ms
        )
    
    def analyze(self) -> Dict[str, Any]:
        """Analyze performance."""
        report = self.optimizer.analyze_performance()
        return {
            "total_operations": report.total_operations,
            "total_time_ms": report.total_execution_time_ms,
            "avg_time_ms": report.avg_execution_time_ms,
            "bottlenecks_count": len(report.bottlenecks),
            "suggestions_count": len(report.suggestions),
            "top_slowest": report.top_slowest_operations
        }
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get statistics."""
        return self.optimizer.get_statistics()



if __name__ == "__main__":
    print("=== Performance Optimization Pattern with LangChain/LangGraph Demo ===\n")
    
    # Note: This demo shows the structure. In production, you'd need:
    # - export OPENAI_API_KEY="your-key-here"
    # - Actual LLM calls would be made
    
    print("Scenario 1: Operation Profiling")
    print("-" * 50)
    optimizer = PerformanceOptimizationPattern(model_name="gpt-4")
    
    # Record multiple operations
    operations = [
        ("database_query", 45.5, False),
        ("api_call", 120.3, False),
        ("data_processing", 230.8, False),
        ("cache_lookup", 5.2, False),
        ("database_query", 52.1, False),
        ("api_call", 115.7, False),
        ("data_processing", 245.2, False),
        ("database_query", 48.9, True),  # Error
    ]
    
    print("Recording operation executions:")
    for operation, duration, error in operations:
        optimizer.record(operation, duration, error)
        status = "ERROR" if error else "OK"
        print(f"  {operation}: {duration:.1f}ms [{status}]")
    
    print("\nScenario 2: Bottleneck Identification")
    print("-" * 50)
    
    # Identify bottlenecks
    bottlenecks = [
        ("data_processing", BottleneckType.CPU_BOUND, 240.0, 100.0),
        ("api_call", BottleneckType.NETWORK_BOUND, 118.0, 50.0),
        ("database_query", BottleneckType.DATABASE_BOUND, 50.0, 20.0),
    ]
    
    print("Identifying bottlenecks:")
    for operation, bt_type, measured, expected in bottlenecks:
        bottleneck = optimizer.identify_bottleneck(
            operation, bt_type, measured, expected
        )
        print(f"  {operation} ({bt_type.value}):")
        print(f"    Measured: {measured:.1f}ms, Expected: {expected:.1f}ms")
        print(f"    Severity: {bottleneck.severity:.2f}, Impact: {bottleneck.impact}")
    
    print("\nScenario 3: Optimization Application")
    print("-" * 50)
    
    # Apply optimizations
    optimizations = [
        ("data_processing", OptimizationStrategy.PARALLELIZATION, 240.0, 85.0),
        ("api_call", OptimizationStrategy.CACHING, 118.0, 12.0),
        ("database_query", OptimizationStrategy.INDEX_OPTIMIZATION, 50.0, 18.0),
    ]
    
    print("Applying optimizations:")
    for operation, strategy, before, after in optimizations:
        result = optimizer.optimize(operation, strategy, before, after)
        print(f"  {operation} ({strategy.value}):")
        print(f"    Before: {before:.1f}ms, After: {after:.1f}ms")
        print(f"    Improvement: {result.improvement_percent:.1f}%")
        print(f"    Success: {result.success}")
    
    print("\nScenario 4: Performance Analysis")
    print("-" * 50)
    
    print("Analyzing performance with LLM insights...")
    analysis = optimizer.analyze()
    
    print(f"\nAnalysis Results:")
    print(f"  Total Operations: {analysis['total_operations']}")
    print(f"  Total Execution Time: {analysis['total_time_ms']:.1f}ms")
    print(f"  Average Time per Operation: {analysis['avg_time_ms']:.1f}ms")
    print(f"  Bottlenecks Identified: {analysis['bottlenecks_count']}")
    print(f"  Optimization Suggestions: {analysis['suggestions_count']}")
    
    if analysis['top_slowest']:
        print(f"\nTop Slowest Operations:")
        for i, operation in enumerate(analysis['top_slowest'], 1):
            print(f"    {i}. {operation}")
    
    print("\nScenario 5: Using Profile Decorator")
    print("-" * 50)
    
    # Example of using decorator
    @optimizer.profile("heavy_computation")
    def heavy_computation(n: int) -> int:
        """Simulate heavy computation."""
        result = 0
        for i in range(n):
            result += i * i
        return result
    
    print("Profiling decorated function:")
    for i in range(3):
        result = heavy_computation(1000)
        print(f"  Iteration {i+1}: computed {result}")
    
    # Get profile for decorated function
    profile = optimizer.optimizer.get_profile("heavy_computation")
    if profile:
        print(f"\nProfile for 'heavy_computation':")
        print(f"  Executions: {profile.execution_count}")
        print(f"  Avg Time: {profile.avg_time()*1000:.2f}ms")
        print(f"  Min Time: {profile.min_time*1000:.2f}ms")
        print(f"  Max Time: {profile.max_time*1000:.2f}ms")
    
    print("\nScenario 6: Statistics")
    print("-" * 50)
    
    stats = optimizer.get_statistics()
    print(f"Total Operations Profiled: {stats['total_operations_profiled']}")
    print(f"Unique Operations: {stats['unique_operations']}")
    print(f"Optimizations Applied: {stats['total_optimizations_applied']}")
    print(f"Bottlenecks Identified: {stats['bottlenecks_identified']}")
    print(f"Average Improvement: {stats['avg_improvement_percent']:.1f}%")
    print(f"Successful Optimizations: {stats['successful_optimizations']}")
    
    if stats['strategies_used']:
        print(f"\nOptimization Strategies Used:")
        for strategy, count in stats['strategies_used'].items():
            print(f"  {strategy}: {count}")
    
    print("\n" + "=" * 50)
    print("Performance optimization pattern demonstration complete!")
    print("=" * 50)

    

Agentic Architecture Pattern: Performance Optimization

Pattern 174: 174 Performance Optimization – Line-by-Line Explanation

Overview

Pattern 174: Performance Optimization Pattern with LangChain/LangGraph

Purpose: Optimize agent performance through profiling, tuning, and LLM-powered insights.

Code Structure Analysis

File Information

• Pattern Number: 174
• Pattern Name: 174 Performance Optimization
• File Path: patterns\quality_performance\174_performance_optimization.py
• Total Lines: 648

Detailed Code Explanation

Module Docstring

"""
Pattern 174: Performance Optimization Pattern with LangChain/LangGraph

Purpose: Optimize agent performance through profiling, tuning, and LLM-powered insights.
"""

Explanation: This module docstring describes the purpose and key features of the 174 Performance Optimization pattern.

Imports Section

from typing import Dict, Any, List, Optional, Callable, TypedDict
from dataclasses import dataclass, field
from datetime import datetime, timedelta
import time
from functools import wraps
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
import statistics  # First 20 imports shown

Explanation: The module imports necessary dependencies: – from typing import Dict, Any, List, Optional, Callable, TypedDict: Type hints for better code documentation – from dataclasses import dataclass, field: Data class decorators for clean data structures – from datetime import datetime, timedelta: Date and time utilities – from enum import Enum: Enumeration support for fixed sets of values – from langchain_core.prompts import ChatPromptTemplate: LangChain framework integration – from langchain_openai import ChatOpenAI: LangChain framework integration – from langgraph.graph import StateGraph, END: LangGraph for workflow orchestration

Enumerations

OptimizationStrategy (Line 19)

class OptimizationStrategy(Enum):
    CACHING
    BATCHING
    PARALLELIZATION
    LAZY_LOADING
    INDEX_OPTIMIZATION

Purpose: Performance optimization strategies.

Values: – CACHING: Represents caching – BATCHING: Represents batching – PARALLELIZATION: Represents parallelization – LAZY_LOADING: Represents lazy loading – INDEX_OPTIMIZATION: Represents index optimization

BottleneckType (Line 28)

class BottleneckType(Enum):
    CPU_BOUND
    IO_BOUND
    MEMORY_BOUND
    NETWORK_BOUND
    DATABASE_BOUND

Purpose: Types of performance bottlenecks.

Values: – CPU_BOUND: Represents cpu bound – IO_BOUND: Represents io bound – MEMORY_BOUND: Represents memory bound – NETWORK_BOUND: Represents network bound – DATABASE_BOUND: Represents database bound

ProfileLevel (Line 37)

class ProfileLevel(Enum):
    BASIC
    DETAILED
    COMPREHENSIVE

Purpose: Profiling detail levels.

Values: – BASIC: Represents basic – DETAILED: Represents detailed – COMPREHENSIVE: Represents comprehensive

PriorityLevel (Line 44)

class PriorityLevel(Enum):
    LOW
    MEDIUM
    HIGH
    CRITICAL

Purpose: Optimization priority levels.

Values: – LOW: Represents low – MEDIUM: Represents medium – HIGH: Represents high – CRITICAL: Represents critical

Classes

OptimizationStrategy (Line 19)

Description: Performance optimization strategies.

BottleneckType (Line 28)

Description: Types of performance bottlenecks.

ProfileLevel (Line 37)

Description: Profiling detail levels.

PriorityLevel (Line 44)

Description: Optimization priority levels.

PerformanceProfile (Line 53)

Description: Performance profile for an operation.

Methods:

avg_time (Line 63)

def avg_time(self):
    """Average execution time."""

Purpose: Average execution time.

Parameters: – self: Instance reference

record (Line 67)

def record(self, duration, error):
    """Record an execution."""

Purpose: Record an execution.

Parameters: – self: Instance reference – duration: Parameter for duration – error: Parameter for error

Bottleneck (Line 79)

Description: Identified performance bottleneck.

OptimizationSuggestion (Line 90)

Description: Performance optimization suggestion.

OptimizationResult (Line 101)

Description: Result of applying an optimization.

PerformanceReport (Line 112)

Description: Comprehensive performance report.

OptimizationState (Line 122)

Description: State for optimization workflow.

PerformanceOptimizer (Line 131)

Description: Advanced performance optimizer with LangChain/LangGraph integration.

Methods:

__init__ (Line 136)

def __init__(self, model_name, temperature):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – model_name: LLM model to use (e.g., “gpt-4”) – temperature: Randomness in LLM responses (0.0-1.0)

_build_optimization_workflow (Line 152)

def _build_optimization_workflow(self):
    """Build LangGraph workflow for performance optimization."""

Purpose: Build LangGraph workflow for performance optimization.

Parameters: – self: Instance reference

_analyze_profiles_node (Line 173)

def _analyze_profiles_node(self, state):
    """Analyze performance profiles."""

Purpose: Analyze performance profiles.

Parameters: – self: Instance reference – state: Current workflow state

_identify_bottlenecks_node (Line 190)

def _identify_bottlenecks_node(self, state):
    """Identify performance bottlenecks."""

Purpose: Identify performance bottlenecks.

Parameters: – self: Instance reference – state: Current workflow state

_generate_insights_node (Line 217)

def _generate_insights_node(self, state):
    """Generate insights using LLM."""

Purpose: Generate insights using LLM.

Parameters: – self: Instance reference – state: Current workflow state

_recommend_strategies_node (Line 239)

def _recommend_strategies_node(self, state):
    """Recommend optimization strategies using LLM."""

Purpose: Recommend optimization strategies using LLM.

Parameters: – self: Instance reference – state: Current workflow state

_prioritize_optimizations_node (Line 271)

def _prioritize_optimizations_node(self, state):
    """Prioritize optimization strategies."""

Purpose: Prioritize optimization strategies.

Parameters: – self: Instance reference – state: Current workflow state

profile_operation (Line 292)

def profile_operation(self, operation):
    """Decorator to profile an operation."""

Purpose: Decorator to profile an operation.

Parameters: – self: Instance reference – operation: Parameter for operation

record_execution (Line 319)

def record_execution(self, operation, duration_ms, error):
    """Manually record an operation execution."""

Purpose: Manually record an operation execution.

Parameters: – self: Instance reference – operation: Parameter for operation – duration_ms: Parameter for duration_ms – error: Parameter for error

identify_bottleneck (Line 332)

def identify_bottleneck(self, operation, bottleneck_type, measured_time_ms, expected_time_ms):
    """Manually identify a bottleneck."""

Purpose: Manually identify a bottleneck.

Parameters: – self: Instance reference – operation: Parameter for operation – bottleneck_type: Parameter for bottleneck_type – measured_time_ms: Parameter for measured_time_ms – expected_time_ms: Parameter for expected_time_ms

apply_optimization (Line 356)

def apply_optimization(self, operation, strategy, before_avg_ms, after_avg_ms):
    """Record an applied optimization."""

Purpose: Record an applied optimization.

Parameters: – self: Instance reference – operation: Parameter for operation – strategy: Parameter for strategy – before_avg_ms: Parameter for before_avg_ms – after_avg_ms: Parameter for after_avg_ms

analyze_performance (Line 387)

def analyze_performance(self):
    """Analyze performance using workflow."""

Purpose: Analyze performance using workflow.

Parameters: – self: Instance reference

get_profile (Line 435)

def get_profile(self, operation):
    """Get profile for an operation."""

Purpose: Get profile for an operation.

Parameters: – self: Instance reference – operation: Parameter for operation

get_statistics (Line 439)

def get_statistics(self):
    """Get optimization statistics."""

Purpose: Get optimization statistics.

Parameters: – self: Instance reference

PerformanceOptimizationPattern (Line 456)

Description: Comprehensive performance optimization pattern coordinator with LangChain/LangGraph.

Methods:

__init__ (Line 461)

def __init__(self, model_name):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – model_name: LLM model to use (e.g., “gpt-4”)

profile (Line 464)

def profile(self, operation):
    """Profile an operation using decorator."""

Purpose: Profile an operation using decorator.

Parameters: – self: Instance reference – operation: Parameter for operation

record (Line 468)

def record(self, operation, duration_ms, error):
    """Record operation execution."""

Purpose: Record operation execution.

Parameters: – self: Instance reference – operation: Parameter for operation – duration_ms: Parameter for duration_ms – error: Parameter for error

identify_bottleneck (Line 477)

def identify_bottleneck(self, operation, bottleneck_type, measured_ms, expected_ms):
    """Identify a performance bottleneck."""

Purpose: Identify a performance bottleneck.

Parameters: – self: Instance reference – operation: Parameter for operation – bottleneck_type: Parameter for bottleneck_type – measured_ms: Parameter for measured_ms – expected_ms: Parameter for expected_ms

optimize (Line 489)

def optimize(self, operation, strategy, before_ms, after_ms):
    """Record an optimization."""

Purpose: Record an optimization.

Parameters: – self: Instance reference – operation: Parameter for operation – strategy: Parameter for strategy – before_ms: Parameter for before_ms – after_ms: Parameter for after_ms

analyze (Line 501)

def analyze(self):
    """Analyze performance."""

Purpose: Analyze performance.

Parameters: – self: Instance reference

get_statistics (Line 513)

def get_statistics(self):
    """Get statistics."""

Purpose: Get statistics.

Parameters: – self: Instance reference

Design Patterns Used

• Strategy Pattern: Different execution strategies for various scenarios
• State Pattern: Managing different states in the workflow
• Chain of Responsibility: Sequential processing through workflow nodes
• Observer Pattern: Tracking and monitoring pattern execution
• Factory Pattern: Creating instances with proper configuration

Usage Example

# Initialize the pattern
pattern = OptimizationStrategy(name="example", model_name="gpt-4")

# Configure and execute
result = pattern.execute(data={"key": "value"})

# Check results
print(result)

Key Concepts

• LLM Integration: Uses LangChain/OpenAI for intelligent decision-making
• Workflow Orchestration: LangGraph manages execution flow through nodes
• State Management: TypedDict defines workflow state structure
• Error Handling: Comprehensive error handling and fallback mechanisms
• Extensibility: Designed for easy customization and extension

Extension Points

This pattern can be extended by: 1. Adding custom workflow nodes 2. Implementing additional validation logic 3. Integrating with external services 4. Customizing LLM prompts and responses 5. Adding monitoring and logging capabilities

    """
Pattern 173: Telemetry Pattern with LangChain/LangGraph

Purpose: Collect detailed operational data for analysis and optimization with LLM insights.
"""

from typing import Dict, Any, List, Optional, TypedDict
from dataclasses import dataclass, field
from datetime import datetime, timedelta
from collections import defaultdict
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
import statistics
import uuid


class EventType(Enum):
    """Telemetry event types."""
    USER_ACTION = "user_action"
    SYSTEM_EVENT = "system_event"
    ERROR = "error"
    PERFORMANCE = "performance"
    BUSINESS = "business"


class TraceLevel(Enum):
    """Trace detail levels."""
    CRITICAL = "critical"
    ERROR = "error"
    WARNING = "warning"
    INFO = "info"
    DEBUG = "debug"


class AggregationPeriod(Enum):
    """Data aggregation periods."""
    MINUTE = "minute"
    HOUR = "hour"
    DAY = "day"
    WEEK = "week"


@dataclass
class TelemetryEvent:
    """Telemetry event."""
    event_id: str
    event_type: EventType
    agent_id: str
    properties: Dict[str, Any]
    severity: TraceLevel = TraceLevel.INFO
    timestamp: datetime = field(default_factory=datetime.now)


@dataclass
class Span:
    """Distributed tracing span."""
    span_id: str
    parent_id: Optional[str]
    operation: str
    start_time: datetime
    end_time: Optional[datetime] = None
    tags: Dict[str, Any] = field(default_factory=dict)
    status: str = "in_progress"
    error: Optional[str] = None
    
    def duration_ms(self) -> float:
        """Get span duration in milliseconds."""
        if self.end_time:
            return (self.end_time - self.start_time).total_seconds() * 1000
        return 0.0


@dataclass
class MetricPoint:
    """Custom metric data point."""
    metric_name: str
    value: float
    dimensions: Dict[str, str]
    unit: str = ""
    timestamp: datetime = field(default_factory=datetime.now)


@dataclass
class Trace:
    """Distributed trace containing multiple spans."""
    trace_id: str
    root_span_id: str
    spans: List[Span]
    start_time: datetime
    end_time: Optional[datetime] = None
    service_name: str = ""


@dataclass
class AggregatedMetrics:
    """Aggregated metrics summary."""
    period: AggregationPeriod
    start_time: datetime
    end_time: datetime
    metrics: Dict[str, Dict[str, float]]
    event_counts: Dict[EventType, int]
    error_rate: float


class TelemetryState(TypedDict):
    """State for telemetry analysis workflow."""
    events: List[Dict[str, Any]]
    traces: List[Dict[str, Any]]
    metrics: Dict[str, Any]
    insights: str
    recommendations: List[str]
    anomalies: List[str]


class TelemetryCollector:
    """
    Advanced telemetry collector with LangChain/LangGraph integration.
    """
    
    def __init__(self, model_name: str = "gpt-4", temperature: float = 0.3):
        self.llm = ChatOpenAI(model=model_name, temperature=temperature)
        self.events: List[TelemetryEvent] = []
        self.spans: Dict[str, Span] = {}
        self.traces: Dict[str, Trace] = {}
        self.metrics: Dict[str, List[MetricPoint]] = defaultdict(list)
        self.active_spans: Dict[str, Span] = {}
        self.workflow = self._build_telemetry_workflow()
        
        # Statistics
        self.stats = {
            "total_events": 0,
            "total_spans": 0,
            "total_traces": 0,
            "total_metrics": 0,
            "events_by_type": {event_type: 0 for event_type in EventType},
            "errors_tracked": 0,
            "avg_span_duration_ms": 0.0
        }
    
    def _build_telemetry_workflow(self) -> StateGraph:
        """Build LangGraph workflow for telemetry analysis."""
        workflow = StateGraph(TelemetryState)
        
        # Add nodes
        workflow.add_node("collect_data", self._collect_data_node)
        workflow.add_node("analyze_patterns", self._analyze_patterns_node)
        workflow.add_node("detect_anomalies", self._detect_anomalies_node)
        workflow.add_node("generate_insights", self._generate_insights_node)
        workflow.add_node("provide_recommendations", self._provide_recommendations_node)
        
        # Define edges
        workflow.set_entry_point("collect_data")
        workflow.add_edge("collect_data", "analyze_patterns")
        workflow.add_edge("analyze_patterns", "detect_anomalies")
        workflow.add_edge("detect_anomalies", "generate_insights")
        workflow.add_edge("generate_insights", "provide_recommendations")
        workflow.add_edge("provide_recommendations", END)
        
        return workflow.compile()
    
    def _collect_data_node(self, state: TelemetryState) -> TelemetryState:
        """Collect and organize telemetry data."""
        # Get recent events
        recent_events = self.events[-50:]
        state["events"] = [
            {
                "event_id": e.event_id,
                "type": e.event_type.value,
                "agent_id": e.agent_id,
                "severity": e.severity.value,
                "timestamp": e.timestamp.isoformat()
            }
            for e in recent_events
        ]
        
        # Get recent traces
        recent_traces = list(self.traces.values())[-20:]
        state["traces"] = [
            {
                "trace_id": t.trace_id,
                "spans": len(t.spans),
                "duration_ms": sum(s.duration_ms() for s in t.spans),
                "service": t.service_name
            }
            for t in recent_traces
        ]
        
        # Get metrics summary
        metrics_summary = {}
        for metric_name, points in self.metrics.items():
            recent_points = points[-20:]
            if recent_points:
                metrics_summary[metric_name] = {
                    "count": len(recent_points),
                    "avg": statistics.mean(p.value for p in recent_points),
                    "min": min(p.value for p in recent_points),
                    "max": max(p.value for p in recent_points)
                }
        
        state["metrics"] = metrics_summary
        return state
    
    def _analyze_patterns_node(self, state: TelemetryState) -> TelemetryState:
        """Analyze telemetry patterns using LLM."""
        events = state.get("events", [])
        traces = state.get("traces", [])
        metrics = state.get("metrics", {})
        
        # Use LLM to analyze patterns
        prompt = ChatPromptTemplate.from_template(
            "Analyze the following telemetry data and identify patterns:\n"
            "Recent Events: {events_summary}\n"
            "Trace Data: {traces_summary}\n"
            "Metrics: {metrics_summary}\n\n"
            "Identify key patterns, trends, and noteworthy observations."
        )
        
        chain = prompt | self.llm
        response = chain.invoke({
            "events_summary": f"{len(events)} events tracked",
            "traces_summary": f"{len(traces)} traces with avg {statistics.mean([t['duration_ms'] for t in traces]):.2f}ms" if traces else "No traces",
            "metrics_summary": str(metrics)
        })
        
        state["insights"] = str(response.content)
        return state
    
    def _detect_anomalies_node(self, state: TelemetryState) -> TelemetryState:
        """Detect anomalies in telemetry data."""
        anomalies = []
        
        # Detect error spikes
        recent_events = state.get("events", [])
        error_events = [e for e in recent_events if e.get("type") == EventType.ERROR.value]
        if len(error_events) > len(recent_events) * 0.1:
            anomalies.append(f"High error rate: {len(error_events)}/{len(recent_events)} events are errors")
        
        # Detect slow traces
        traces = state.get("traces", [])
        if traces:
            durations = [t["duration_ms"] for t in traces]
            avg_duration = statistics.mean(durations)
            stdev_duration = statistics.stdev(durations) if len(durations) > 1 else 0
            
            slow_traces = [t for t in traces if t["duration_ms"] > avg_duration + 2 * stdev_duration]
            if slow_traces:
                anomalies.append(f"Detected {len(slow_traces)} slow traces (>2 std dev)")
        
        # Detect metric anomalies
        metrics = state.get("metrics", {})
        for metric_name, stats in metrics.items():
            if stats["max"] > stats["avg"] * 3:
                anomalies.append(f"Metric '{metric_name}' has outlier: max={stats['max']:.2f}, avg={stats['avg']:.2f}")
        
        state["anomalies"] = anomalies
        return state
    
    def _generate_insights_node(self, state: TelemetryState) -> TelemetryState:
        """Generate insights from telemetry data."""
        insights = state.get("insights", "")
        anomalies = state.get("anomalies", [])
        
        if anomalies:
            # Use LLM to interpret anomalies
            prompt = ChatPromptTemplate.from_template(
                "Based on the following telemetry anomalies, provide detailed insights:\n"
                "Anomalies: {anomalies}\n"
                "Context: {insights}\n\n"
                "Provide actionable insights about what these anomalies might indicate."
            )
            
            chain = prompt | self.llm
            response = chain.invoke({
                "anomalies": "\n".join(anomalies),
                "insights": insights
            })
            
            state["insights"] = str(response.content)
        
        return state
    
    def _provide_recommendations_node(self, state: TelemetryState) -> TelemetryState:
        """Provide recommendations based on telemetry analysis."""
        insights = state.get("insights", "")
        anomalies = state.get("anomalies", [])
        
        recommendations = []
        
        if anomalies or insights:
            # Use LLM to generate recommendations
            prompt = ChatPromptTemplate.from_template(
                "Based on the telemetry analysis, provide 3-5 specific recommendations:\n"
                "Insights: {insights}\n"
                "Anomalies: {anomalies}\n\n"
                "Provide actionable recommendations for improvement."
            )
            
            chain = prompt | self.llm
            response = chain.invoke({
                "insights": insights,
                "anomalies": "\n".join(anomalies) if anomalies else "No anomalies detected"
            })
            
            # Parse recommendations
            recommendations = [
                line.strip()
                for line in str(response.content).split('\n')
                if line.strip() and len(line.strip()) > 10
            ][:5]
        
        state["recommendations"] = recommendations
        return state
    
    def track_event(
        self,
        event_type: EventType,
        agent_id: str,
        properties: Dict[str, Any],
        severity: TraceLevel = TraceLevel.INFO
    ) -> TelemetryEvent:
        """Track a telemetry event."""
        event = TelemetryEvent(
            event_id=str(uuid.uuid4()),
            event_type=event_type,
            agent_id=agent_id,
            properties=properties,
            severity=severity
        )
        
        self.events.append(event)
        self.stats["total_events"] += 1
        self.stats["events_by_type"][event_type] += 1
        
        if event_type == EventType.ERROR:
            self.stats["errors_tracked"] += 1
        
        return event
    
    def start_span(
        self,
        operation: str,
        parent_id: Optional[str] = None,
        trace_id: Optional[str] = None,
        tags: Optional[Dict[str, Any]] = None
    ) -> Span:
        """Start a new tracing span."""
        span = Span(
            span_id=str(uuid.uuid4()),
            parent_id=parent_id,
            operation=operation,
            start_time=datetime.now(),
            tags=tags or {}
        )
        
        self.spans[span.span_id] = span
        self.active_spans[span.span_id] = span
        self.stats["total_spans"] += 1
        
        # Create or update trace
        if trace_id:
            if trace_id in self.traces:
                self.traces[trace_id].spans.append(span)
            else:
                self.traces[trace_id] = Trace(
                    trace_id=trace_id,
                    root_span_id=span.span_id,
                    spans=[span],
                    start_time=span.start_time
                )
                self.stats["total_traces"] += 1
        
        return span
    
    def end_span(self, span_id: str, status: str = "success", error: Optional[str] = None):
        """End a tracing span."""
        if span_id in self.active_spans:
            span = self.active_spans[span_id]
            span.end_time = datetime.now()
            span.status = status
            span.error = error
            
            # Update average duration
            total_durations = sum(s.duration_ms() for s in self.spans.values() if s.end_time)
            completed_spans = sum(1 for s in self.spans.values() if s.end_time)
            if completed_spans > 0:
                self.stats["avg_span_duration_ms"] = total_durations / completed_spans
            
            del self.active_spans[span_id]
    
    def record_metric(
        self,
        metric_name: str,
        value: float,
        dimensions: Optional[Dict[str, str]] = None,
        unit: str = ""
    ):
        """Record a custom metric."""
        metric = MetricPoint(
            metric_name=metric_name,
            value=value,
            dimensions=dimensions or {},
            unit=unit
        )
        
        self.metrics[metric_name].append(metric)
        self.stats["total_metrics"] += 1
    
    def get_aggregated_metrics(
        self,
        period: AggregationPeriod,
        start_time: Optional[datetime] = None,
        end_time: Optional[datetime] = None
    ) -> AggregatedMetrics:
        """Get aggregated metrics for a period."""
        if not end_time:
            end_time = datetime.now()
        if not start_time:
            # Default to last hour
            start_time = end_time - timedelta(hours=1)
        
        # Filter events in time range
        period_events = [
            e for e in self.events
            if start_time <= e.timestamp  0 else 0.0
        
        # Aggregate metrics
        metrics_summary = {}
        for metric_name, points in self.metrics.items():
            period_points = [
                p for p in points
                if start_time <= p.timestamp  1 else 0.0
                }
        
        return AggregatedMetrics(
            period=period,
            start_time=start_time,
            end_time=end_time,
            metrics=metrics_summary,
            event_counts=event_counts,
            error_rate=error_rate
        )
    
    def analyze_telemetry(self) -> Dict[str, Any]:
        """Analyze telemetry data using workflow."""
        initial_state: TelemetryState = {
            "events": [],
            "traces": [],
            "metrics": {},
            "insights": "",
            "recommendations": [],
            "anomalies": []
        }
        
        # Run workflow
        result = self.workflow.invoke(initial_state)
        
        return {
            "insights": result["insights"],
            "anomalies": result["anomalies"],
            "recommendations": result["recommendations"],
            "events_analyzed": len(result["events"]),
            "traces_analyzed": len(result["traces"])
        }
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get telemetry statistics."""
        return {
            "total_events": self.stats["total_events"],
            "total_spans": self.stats["total_spans"],
            "total_traces": self.stats["total_traces"],
            "total_metrics": self.stats["total_metrics"],
            "events_by_type": {
                event_type.value: count
                for event_type, count in self.stats["events_by_type"].items()
            },
            "errors_tracked": self.stats["errors_tracked"],
            "avg_span_duration_ms": round(self.stats["avg_span_duration_ms"], 2),
            "active_spans": len(self.active_spans)
        }


class TelemetryPattern:
    """
    Comprehensive telemetry pattern coordinator with LangChain/LangGraph.
    """
    
    def __init__(self, model_name: str = "gpt-4"):
        self.collector = TelemetryCollector(model_name=model_name)
    
    def track(
        self,
        event_type: EventType,
        agent_id: str,
        properties: Dict[str, Any]
    ) -> str:
        """Track an event."""
        event = self.collector.track_event(event_type, agent_id, properties)
        return event.event_id
    
    def trace_operation(
        self,
        operation: str,
        tags: Optional[Dict[str, Any]] = None
    ) -> Span:
        """Start tracing an operation."""
        return self.collector.start_span(operation, tags=tags)
    
    def complete_trace(self, span: Span, success: bool = True):
        """Complete a trace."""
        status = "success" if success else "error"
        self.collector.end_span(span.span_id, status)
    
    def measure(
        self,
        metric_name: str,
        value: float,
        dimensions: Optional[Dict[str, str]] = None
    ):
        """Measure a metric."""
        self.collector.record_metric(metric_name, value, dimensions)
    
    def analyze(self) -> Dict[str, Any]:
        """Analyze telemetry data."""
        return self.collector.analyze_telemetry()
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get telemetry statistics."""
        return self.collector.get_statistics()



if __name__ == "__main__":
    print("=== Telemetry Pattern with LangChain/LangGraph Demo ===\n")
    
    # Note: This demo shows the structure. In production, you'd need:
    # - export OPENAI_API_KEY="your-key-here"
    # - Actual LLM calls would be made
    
    print("Scenario 1: Event Tracking")
    print("-" * 50)
    telemetry = TelemetryPattern(model_name="gpt-4")
    
    # Track various events
    events = [
        (EventType.USER_ACTION, "user-123", {"action": "login", "success": True}),
        (EventType.SYSTEM_EVENT, "api-server", {"event": "startup", "version": "1.2.3"}),
        (EventType.PERFORMANCE, "database", {"query_time_ms": 45.2, "query": "SELECT"}),
        (EventType.ERROR, "payment-service", {"error": "timeout", "retry_count": 3}),
        (EventType.BUSINESS, "checkout", {"order_id": "ORD-123", "amount": 99.99}),
    ]
    
    print("Tracking events:")
    for event_type, agent_id, properties in events:
        event_id = telemetry.track(event_type, agent_id, properties)
        print(f"  ✓ Tracked {event_type.value} event {event_id[:8]}...")
    
    print("\nScenario 2: Distributed Tracing")
    print("-" * 50)
    
    # Trace a complex operation
    print("Tracing order processing:")
    order_span = telemetry.trace_operation("process_order", tags={"order_id": "ORD-456"})
    print(f"  Started trace: {order_span.span_id[:8]}...")
    
    # Simulate child operations
    validate_span = telemetry.collector.start_span(
        "validate_payment",
        parent_id=order_span.span_id,
        trace_id=order_span.span_id
    )
    telemetry.collector.end_span(validate_span.span_id, "success")
    print(f"  Completed validation: {validate_span.duration_ms():.2f}ms")
    
    inventory_span = telemetry.collector.start_span(
        "check_inventory",
        parent_id=order_span.span_id,
        trace_id=order_span.span_id
    )
    telemetry.collector.end_span(inventory_span.span_id, "success")
    print(f"  Completed inventory check: {inventory_span.duration_ms():.2f}ms")
    
    telemetry.complete_trace(order_span, success=True)
    print(f"  Total order processing: {order_span.duration_ms():.2f}ms")
    
    print("\nScenario 3: Custom Metrics")
    print("-" * 50)
    
    # Record various metrics
    metrics_data = [
        ("cpu_usage", 65.5, {"host": "server-1", "service": "api"}),
        ("memory_mb", 2048.0, {"host": "server-1", "service": "api"}),
        ("request_count", 1543.0, {"endpoint": "/api/users", "status": "200"}),
        ("response_time_ms", 125.3, {"endpoint": "/api/orders", "status": "200"}),
        ("error_rate", 2.5, {"service": "payment", "type": "timeout"}),
    ]
    
    print("Recording metrics:")
    for metric_name, value, dimensions in metrics_data:
        telemetry.measure(metric_name, value, dimensions)
        print(f"  ✓ {metric_name}: {value} {dimensions}")
    
    print("\nScenario 4: Telemetry Analysis")
    print("-" * 50)
    
    print("Analyzing telemetry data...")
    analysis = telemetry.analyze()
    
    print(f"\nInsights:")
    print(f"  {analysis['insights'][:200]}...")
    
    print(f"\nAnomalies Detected: {len(analysis['anomalies'])}")
    for anomaly in analysis['anomalies']:
        print(f"  ⚠️ {anomaly}")
    
    print(f"\nRecommendations:")
    for i, recommendation in enumerate(analysis['recommendations'][:3], 1):
        print(f"  {i}. {recommendation[:100]}...")
    
    print(f"\nEvents Analyzed: {analysis['events_analyzed']}")
    print(f"Traces Analyzed: {analysis['traces_analyzed']}")
    
    print("\nScenario 5: Aggregated Metrics")
    print("-" * 50)
    
    aggregated = telemetry.collector.get_aggregated_metrics(
        AggregationPeriod.HOUR,
        start_time=datetime.now() - timedelta(hours=1)
    )
    
    print(f"Period: {aggregated.period.value}")
    print(f"Error Rate: {aggregated.error_rate:.2f}%")
    
    print("\nEvent Counts:")
    for event_type, count in aggregated.event_counts.items():
        if count > 0:
            print(f"  {event_type.value}: {count}")
    
    print("\nMetric Aggregations:")
    for metric_name, stats in list(aggregated.metrics.items())[:3]:
        print(f"  {metric_name}:")
        print(f"    avg={stats['avg']:.2f}, min={stats['min']:.2f}, max={stats['max']:.2f}")
    
    print("\nScenario 6: Statistics")
    print("-" * 50)
    
    stats = telemetry.get_statistics()
    print(f"Total Events: {stats['total_events']}")
    print(f"Total Spans: {stats['total_spans']}")
    print(f"Total Traces: {stats['total_traces']}")
    print(f"Total Metrics: {stats['total_metrics']}")
    print(f"Errors Tracked: {stats['errors_tracked']}")
    print(f"Avg Span Duration: {stats['avg_span_duration_ms']:.2f}ms")
    
    print("\nEvents by Type:")
    for event_type, count in stats['events_by_type'].items():
        if count > 0:
            print(f"  {event_type}: {count}")
    
    print("\n" + "=" * 50)
    print("Telemetry pattern demonstration complete!")
    print("=" * 50)

    

Agentic Architecture Pattern: Telemetry

Pattern 173: 173 Telemetry – Line-by-Line Explanation

Overview

Pattern 173: Telemetry Pattern with LangChain/LangGraph

Purpose: Collect detailed operational data for analysis and optimization with LLM insights.

Code Structure Analysis

File Information

• Pattern Number: 173
• Pattern Name: 173 Telemetry
• File Path: patterns\quality_performance\173_telemetry.py
• Total Lines: 687

Detailed Code Explanation

Module Docstring

"""
Pattern 173: Telemetry Pattern with LangChain/LangGraph

Purpose: Collect detailed operational data for analysis and optimization with LLM insights.
"""

Explanation: This module docstring describes the purpose and key features of the 173 Telemetry pattern.

Imports Section

from typing import Dict, Any, List, Optional, TypedDict
from dataclasses import dataclass, field
from datetime import datetime, timedelta
from collections import defaultdict
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
import statistics
import uuid  # First 20 imports shown

Explanation: The module imports necessary dependencies: – from typing import Dict, Any, List, Optional, TypedDict: Type hints for better code documentation – from dataclasses import dataclass, field: Data class decorators for clean data structures – from datetime import datetime, timedelta: Date and time utilities – from enum import Enum: Enumeration support for fixed sets of values – from langchain_core.prompts import ChatPromptTemplate: LangChain framework integration – from langchain_openai import ChatOpenAI: LangChain framework integration – from langgraph.graph import StateGraph, END: LangGraph for workflow orchestration

Enumerations

EventType (Line 19)

class EventType(Enum):
    USER_ACTION
    SYSTEM_EVENT
    ERROR
    PERFORMANCE
    BUSINESS

Purpose: Telemetry event types.

Values: – USER_ACTION: Represents user action – SYSTEM_EVENT: Represents system event – ERROR: Error severity level – PERFORMANCE: Represents performance – BUSINESS: Represents business

TraceLevel (Line 28)

class TraceLevel(Enum):
    CRITICAL
    ERROR
    WARNING
    INFO
    DEBUG

Purpose: Trace detail levels.

Values: – CRITICAL: Represents critical – ERROR: Error severity level – WARNING: Warning severity level – INFO: Informational severity level – DEBUG: Represents debug

AggregationPeriod (Line 37)

class AggregationPeriod(Enum):
    MINUTE
    HOUR
    DAY
    WEEK

Purpose: Data aggregation periods.

Values: – MINUTE: Represents minute – HOUR: Represents hour – DAY: Represents day – WEEK: Represents week

Classes

EventType (Line 19)

Description: Telemetry event types.

TraceLevel (Line 28)

Description: Trace detail levels.

AggregationPeriod (Line 37)

Description: Data aggregation periods.

TelemetryEvent (Line 46)

Description: Telemetry event.

Span (Line 57)

Description: Distributed tracing span.

Methods:

duration_ms (Line 68)

def duration_ms(self):
    """Get span duration in milliseconds."""

Purpose: Get span duration in milliseconds.

Parameters: – self: Instance reference

MetricPoint (Line 76)

Description: Custom metric data point.

Trace (Line 86)

Description: Distributed trace containing multiple spans.

AggregatedMetrics (Line 97)

Description: Aggregated metrics summary.

TelemetryState (Line 107)

Description: State for telemetry analysis workflow.

TelemetryCollector (Line 117)

Description: Advanced telemetry collector with LangChain/LangGraph integration.

Methods:

__init__ (Line 122)

def __init__(self, model_name, temperature):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – model_name: LLM model to use (e.g., “gpt-4”) – temperature: Randomness in LLM responses (0.0-1.0)

_build_telemetry_workflow (Line 142)

def _build_telemetry_workflow(self):
    """Build LangGraph workflow for telemetry analysis."""

Purpose: Build LangGraph workflow for telemetry analysis.

Parameters: – self: Instance reference

_collect_data_node (Line 163)

def _collect_data_node(self, state):
    """Collect and organize telemetry data."""

Purpose: Collect and organize telemetry data.

Parameters: – self: Instance reference – state: Current workflow state

_analyze_patterns_node (Line 205)

def _analyze_patterns_node(self, state):
    """Analyze telemetry patterns using LLM."""

Purpose: Analyze telemetry patterns using LLM.

Parameters: – self: Instance reference – state: Current workflow state

_detect_anomalies_node (Line 230)

def _detect_anomalies_node(self, state):
    """Detect anomalies in telemetry data."""

Purpose: Detect anomalies in telemetry data.

Parameters: – self: Instance reference – state: Current workflow state

_generate_insights_node (Line 260)

def _generate_insights_node(self, state):
    """Generate insights from telemetry data."""

Purpose: Generate insights from telemetry data.

Parameters: – self: Instance reference – state: Current workflow state

_provide_recommendations_node (Line 284)

def _provide_recommendations_node(self, state):
    """Provide recommendations based on telemetry analysis."""

Purpose: Provide recommendations based on telemetry analysis.

Parameters: – self: Instance reference – state: Current workflow state

track_event (Line 316)

def track_event(self, event_type, agent_id, properties, severity):
    """Track a telemetry event."""

Purpose: Track a telemetry event.

Parameters: – self: Instance reference – event_type: Parameter for event_type – agent_id: Parameter for agent_id – properties: Parameter for properties – severity: Parameter for severity

start_span (Line 341)

def start_span(self, operation, parent_id, trace_id, tags):
    """Start a new tracing span."""

Purpose: Start a new tracing span.

Parameters: – self: Instance reference – operation: Parameter for operation – parent_id: Parameter for parent_id – trace_id: Parameter for trace_id – tags: Parameter for tags

end_span (Line 376)

def end_span(self, span_id, status, error):
    """End a tracing span."""

Purpose: End a tracing span.

Parameters: – self: Instance reference – span_id: Parameter for span_id – status: Parameter for status – error: Parameter for error

record_metric (Line 392)

def record_metric(self, metric_name, value, dimensions, unit):
    """Record a custom metric."""

Purpose: Record a custom metric.

Parameters: – self: Instance reference – metric_name: Parameter for metric_name – value: Parameter for value – dimensions: Parameter for dimensions – unit: Parameter for unit

get_aggregated_metrics (Line 410)

def get_aggregated_metrics(self, period, start_time, end_time):
    """Get aggregated metrics for a period."""

Purpose: Get aggregated metrics for a period.

Parameters: – self: Instance reference – period: Parameter for period – start_time: Parameter for start_time – end_time: Parameter for end_time

analyze_telemetry (Line 468)

def analyze_telemetry(self):
    """Analyze telemetry data using workflow."""

Purpose: Analyze telemetry data using workflow.

Parameters: – self: Instance reference

get_statistics (Line 490)

def get_statistics(self):
    """Get telemetry statistics."""

Purpose: Get telemetry statistics.

Parameters: – self: Instance reference

TelemetryPattern (Line 507)

Description: Comprehensive telemetry pattern coordinator with LangChain/LangGraph.

Methods:

__init__ (Line 512)

def __init__(self, model_name):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – model_name: LLM model to use (e.g., “gpt-4”)

track (Line 515)

def track(self, event_type, agent_id, properties):
    """Track an event."""

Purpose: Track an event.

Parameters: – self: Instance reference – event_type: Parameter for event_type – agent_id: Parameter for agent_id – properties: Parameter for properties

trace_operation (Line 525)

def trace_operation(self, operation, tags):
    """Start tracing an operation."""

Purpose: Start tracing an operation.

Parameters: – self: Instance reference – operation: Parameter for operation – tags: Parameter for tags

complete_trace (Line 533)

def complete_trace(self, span, success):
    """Complete a trace."""

Purpose: Complete a trace.

Parameters: – self: Instance reference – span: Parameter for span – success: Parameter for success

measure (Line 538)

def measure(self, metric_name, value, dimensions):
    """Measure a metric."""

Purpose: Measure a metric.

Parameters: – self: Instance reference – metric_name: Parameter for metric_name – value: Parameter for value – dimensions: Parameter for dimensions

analyze (Line 547)

def analyze(self):
    """Analyze telemetry data."""

Purpose: Analyze telemetry data.

Parameters: – self: Instance reference

get_statistics (Line 551)

def get_statistics(self):
    """Get telemetry statistics."""

Purpose: Get telemetry statistics.

Parameters: – self: Instance reference

Design Patterns Used

• Strategy Pattern: Different execution strategies for various scenarios
• State Pattern: Managing different states in the workflow
• Chain of Responsibility: Sequential processing through workflow nodes
• Observer Pattern: Tracking and monitoring pattern execution
• Factory Pattern: Creating instances with proper configuration

Usage Example

# Initialize the pattern
pattern = EventType(name="example", model_name="gpt-4")

# Configure and execute
result = pattern.execute(data={"key": "value"})

# Check results
print(result)

Key Concepts

• LLM Integration: Uses LangChain/OpenAI for intelligent decision-making
• Workflow Orchestration: LangGraph manages execution flow through nodes
• State Management: TypedDict defines workflow state structure
• Error Handling: Comprehensive error handling and fallback mechanisms
• Extensibility: Designed for easy customization and extension

Extension Points

This pattern can be extended by: 1. Adding custom workflow nodes 2. Implementing additional validation logic 3. Integrating with external services 4. Customizing LLM prompts and responses 5. Adding monitoring and logging capabilities

    """
Pattern 172: Logging

Purpose: Enhanced implementation with full LangChain/LangGraph integration
"""

from typing import Dict, Any, List, Optional
from dataclasses import dataclass, field
from datetime import datetime
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
from typing_extensions import TypedDict


@dataclass
class LoggingConfig:
    """Configuration for Logging."""
    model_name: str = "gpt-4"
    temperature: float = 0.7
    max_iterations: int = 10
    enable_logging: bool = True


@dataclass
class LoggingState:
    """State information for Logging."""
    current_step: int = 0
    data: Dict[str, Any] = field(default_factory=dict)
    results: List[Any] = field(default_factory=list)
    metadata: Dict[str, Any] = field(default_factory=dict)
    timestamp: datetime = field(default_factory=datetime.now)


class LoggingAgent:
    """Agent implementing Logging."""
    
    def __init__(self, name: str, config: LoggingConfig = None):
        self.name = name
        self.config = config or LoggingConfig()
        self.llm = ChatOpenAI(
            model=self.config.model_name,
            temperature=self.config.temperature
        )
        self.state = LoggingState()
        self.execution_history: List[Dict[str, Any]] = []
    
    def process(self, input_data: Any) -> Dict[str, Any]:
        """Process input according to Logging."""
        self.state.current_step += 1
        
        prompt = ChatPromptTemplate.from_messages([
            ("system", f"You are {self.name}, implementing Logging. "
                      f"Provide clear, structured responses."),
            ("human", "Process this input: {input_data}")
        ])
        
        chain = prompt | self.llm
        response = chain.invoke({"input_data": str(input_data)})
        
        result = {
            "agent": self.name,
            "step": self.state.current_step,
            "input": input_data,
            "output": str(response.content),
            "timestamp": datetime.now()
        }
        
        self.execution_history.append(result)
        self.state.results.append(result)
        
        return result
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get agent statistics."""
        return {
            "name": self.name,
            "total_executions": len(self.execution_history),
            "current_step": self.state.current_step,
            "results_count": len(self.state.results)
        }


class LoggingPattern:
    """
    Logging - Enhanced implementation
    
    This pattern provides comprehensive functionality for logging
    with full LangChain/LangGraph integration.
    
    Features:
    - Multi-agent coordination
    - State management
    - Execution tracking
    - LangGraph workflow integration
    """
    
    def __init__(self, config: LoggingConfig = None):
        self.config = config or LoggingConfig()
        self.agents: Dict[str, LoggingAgent] = {}
        self.workflow = self._build_workflow()
        self.execution_log: List[Dict[str, Any]] = []
    
    def _build_workflow(self) -> StateGraph:
        """Build LangGraph workflow for pattern execution."""
        
        class WorkflowState(TypedDict):
            input: str
            processed: str
            output: str
            metadata: Dict[str, Any]
        
        def process_input(state: WorkflowState) -> WorkflowState:
            """Process input node."""
            return {
                **state,
                "processed": f"Processed: {state['input']}",
                "metadata": {"step": "process_input"}
            }
        
        def generate_output(state: WorkflowState) -> WorkflowState:
            """Generate output node."""
            llm = ChatOpenAI(model=self.config.model_name, temperature=self.config.temperature)
            
            prompt = ChatPromptTemplate.from_messages([
                ("system", "You are implementing Logging. Generate comprehensive output."),
                ("human", "{input}")
            ])
            
            chain = prompt | llm
            response = chain.invoke({"input": state["processed"]})
            
            return {
                **state,
                "output": str(response.content),
                "metadata": {**state.get("metadata", {}), "step": "generate_output"}
            }
        
        # Build workflow graph
        workflow = StateGraph(WorkflowState)
        workflow.add_node("process", process_input)
        workflow.add_node("generate", generate_output)
        workflow.set_entry_point("process")
        workflow.add_edge("process", "generate")
        workflow.add_edge("generate", END)
        
        return workflow.compile()
    
    def register_agent(self, agent: LoggingAgent):
        """Register an agent with the pattern."""
        self.agents[agent.name] = agent
        if self.config.enable_logging:
            print(f"Registered agent: {agent.name}")
    
    def execute(self, input_data: Any, agent_name: str = None) -> Dict[str, Any]:
        """Execute the pattern with given input."""
        
        # Use workflow if no specific agent requested
        if agent_name is None:
            result = self.workflow.invoke({
                "input": str(input_data),
                "processed": "",
                "output": "",
                "metadata": {}
            })
            
            execution_record = {
                "pattern": "Logging",
                "input": input_data,
                "output": result.get("output", ""),
                "metadata": result.get("metadata", {}),
                "timestamp": datetime.now()
            }
            
            self.execution_log.append(execution_record)
            return execution_record
        
        # Execute with specific agent
        agent = self.agents.get(agent_name)
        if not agent:
            raise ValueError(f"Agent {agent_name} not found")
        
        result = agent.process(input_data)
        self.execution_log.append(result)
        return result
    
    def run(self, input_data: Any) -> Dict[str, Any]:
        """Main execution method."""
        return self.execute(input_data)
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get pattern statistics."""
        return {
            "pattern": "Logging",
            "total_executions": len(self.execution_log),
            "registered_agents": len(self.agents),
            "agents": [agent.get_statistics() for agent in self.agents.values()]
        }


# Example usage
if __name__ == "__main__":
    print("=== Logging Demo ===\n")
    
    # Create configuration
    config = LoggingConfig(
        model_name="gpt-4",
        temperature=0.7,
        enable_logging=True
    )
    
    # Create pattern
    pattern = LoggingPattern(config)
    
    # Create and register agents
    agent1 = LoggingAgent("Agent1", config)
    agent2 = LoggingAgent("Agent2", config)
    pattern.register_agent(agent1)
    pattern.register_agent(agent2)
    
    print("\nExecuting pattern workflow...")
    
    # Execute with workflow
    result1 = pattern.run("Demonstrate logging functionality")
    print(f"\nWorkflow Result:")
    print(f"  Output: {result1['output'][:150]}...")
    
    # Execute with specific agents
    result2 = pattern.execute("Specific task for Agent1", agent_name="Agent1")
    print(f"\nAgent1 Result:")
    print(f"  Output: {result2['output'][:150]}...")
    
    result3 = pattern.execute("Specific task for Agent2", agent_name="Agent2")
    print(f"\nAgent2 Result:")
    print(f"  Output: {result3['output'][:150]}...")
    
    # Show statistics
    print("\n=== Statistics ===")
    stats = pattern.get_statistics()
    print(f"Pattern: {stats['pattern']}")
    print(f"Total Executions: {stats['total_executions']}")
    print(f"Registered Agents: {stats['registered_agents']}")
    print(f"\nAgent Statistics:")
    for agent_stats in stats['agents']:
        print(f"  {agent_stats['name']}: {agent_stats['total_executions']} executions")

    

Agentic Architecture Pattern: Logging

Pattern 172: 172 Logging – Line-by-Line Explanation

Overview

Pattern 172: Logging

Purpose: Enhanced implementation with full LangChain/LangGraph integration

Code Structure Analysis

File Information

• Pattern Number: 172
• Pattern Name: 172 Logging
• File Path: patterns\quality_performance\172_logging.py
• Total Lines: 247

Detailed Code Explanation

Module Docstring

"""
Pattern 172: Logging

Purpose: Enhanced implementation with full LangChain/LangGraph integration
"""

Explanation: This module docstring describes the purpose and key features of the 172 Logging pattern.

Imports Section

from typing import Dict, Any, List, Optional
from dataclasses import dataclass, field
from datetime import datetime
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
from typing_extensions import TypedDict  # First 20 imports shown

Explanation: The module imports necessary dependencies: – from typing import Dict, Any, List, Optional: Type hints for better code documentation – from dataclasses import dataclass, field: Data class decorators for clean data structures – from datetime import datetime: Date and time utilities – from enum import Enum: Enumeration support for fixed sets of values – from langchain_core.prompts import ChatPromptTemplate: LangChain framework integration – from langchain_openai import ChatOpenAI: LangChain framework integration – from langgraph.graph import StateGraph, END: LangGraph for workflow orchestration – from typing_extensions import TypedDict: Type hints for better code documentation

Classes

LoggingConfig (Line 18)

Description: Configuration for Logging.

LoggingState (Line 27)

Description: State information for Logging.

LoggingAgent (Line 36)

Description: Agent implementing Logging.

Methods:

__init__ (Line 39)

def __init__(self, name, config):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – name: Identifier or label – config: Configuration settings

process (Line 49)

def process(self, input_data):
    """Process input according to Logging."""

Purpose: Process input according to Logging.

Parameters: – self: Instance reference – input_data: Parameter for input_data

get_statistics (Line 75)

def get_statistics(self):
    """Get agent statistics."""

Purpose: Get agent statistics.

Parameters: – self: Instance reference

LoggingPattern (Line 85)

Description: Logging – Enhanced implementation

This pattern provides comprehensive functionality for logging with full LangChain/LangGraph integration.

Features: – Multi-agent coordination – State management – Execution tracking – LangGraph workflow integration

Methods:

__init__ (Line 99)

def __init__(self, config):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – config: Configuration settings

_build_workflow (Line 105)

def _build_workflow(self):
    """Build LangGraph workflow for pattern execution."""

Purpose: Build LangGraph workflow for pattern execution.

Parameters: – self: Instance reference

register_agent (Line 150)

def register_agent(self, agent):
    """Register an agent with the pattern."""

Purpose: Register an agent with the pattern.

Parameters: – self: Instance reference – agent: Parameter for agent

execute (Line 156)

def execute(self, input_data, agent_name):
    """Execute the pattern with given input."""

Purpose: Execute the pattern with given input.

Parameters: – self: Instance reference – input_data: Parameter for input_data – agent_name: Parameter for agent_name

run (Line 188)

def run(self, input_data):
    """Main execution method."""

Purpose: Main execution method.

Parameters: – self: Instance reference – input_data: Parameter for input_data

get_statistics (Line 192)

def get_statistics(self):
    """Get pattern statistics."""

Purpose: Get pattern statistics.

Parameters: – self: Instance reference

WorkflowState (Line 108)

Description: No class docstring.

Design Patterns Used

• Strategy Pattern: Different execution strategies for various scenarios
• State Pattern: Managing different states in the workflow
• Chain of Responsibility: Sequential processing through workflow nodes
• Observer Pattern: Tracking and monitoring pattern execution
• Factory Pattern: Creating instances with proper configuration

Usage Example

# Initialize the pattern
pattern = LoggingConfig(name="example", model_name="gpt-4")

# Configure and execute
result = pattern.execute(data={"key": "value"})

# Check results
print(result)

Key Concepts

• LLM Integration: Uses LangChain/OpenAI for intelligent decision-making
• Workflow Orchestration: LangGraph manages execution flow through nodes
• State Management: TypedDict defines workflow state structure
• Error Handling: Comprehensive error handling and fallback mechanisms
• Extensibility: Designed for easy customization and extension

Extension Points

This pattern can be extended by: 1. Adding custom workflow nodes 2. Implementing additional validation logic 3. Integrating with external services 4. Customizing LLM prompts and responses 5. Adding monitoring and logging capabilities

    """
Pattern 171: Monitoring Pattern with LangChain/LangGraph

Purpose: Track agent performance, health, and metrics in real-time with LangGraph integration.
"""

from typing import Dict, Any, List, Optional, TypedDict
from dataclasses import dataclass, field
from datetime import datetime, timedelta
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
import statistics


class HealthStatus(Enum):
    """Health status levels."""
    HEALTHY = "healthy"
    DEGRADED = "degraded"
    UNHEALTHY = "unhealthy"


class MonitoringStrategy(Enum):
    """Monitoring strategies."""
    REAL_TIME = "real_time"
    PERIODIC = "periodic"
    THRESHOLD_BASED = "threshold_based"
    ANOMALY_DETECTION = "anomaly_detection"


class AlertLevel(Enum):
    """Alert severity levels."""
    INFO = "info"
    WARNING = "warning"
    ERROR = "error"
    CRITICAL = "critical"


class ThresholdType(Enum):
    """Threshold comparison types."""
    GREATER_THAN = "greater_than"
    LESS_THAN = "less_than"
    EQUALS = "equals"
    RANGE = "range"


@dataclass
class Metric:
    """Performance metric."""
    name: str
    value: float
    unit: str
    tags: Dict[str, str] = field(default_factory=dict)
    timestamp: datetime = field(default_factory=datetime.now)


@dataclass
class HealthCheck:
    """Health check result."""
    component: str
    status: HealthStatus
    message: str
    metrics: List[Metric] = field(default_factory=list)
    timestamp: datetime = field(default_factory=datetime.now)


@dataclass
class Alert:
    """Monitoring alert."""
    alert_id: str
    level: AlertLevel
    component: str
    message: str
    metric_name: str
    metric_value: float
    threshold: float
    triggered_at: datetime = field(default_factory=datetime.now)
    acknowledged: bool = False


@dataclass
class Threshold:
    """Metric threshold configuration."""
    metric_name: str
    threshold_type: ThresholdType
    value: float
    alert_level: AlertLevel
    max_value: Optional[float] = None  # For range thresholds


@dataclass
class TimeSeriesData:
    """Time-series data for a metric."""
    metric_name: str
    values: List[float]
    timestamps: List[datetime]
    window_size: int = 10


@dataclass
class Dashboard:
    """Monitoring dashboard data."""
    components: Dict[str, HealthStatus]
    total_metrics: int
    active_alerts: int
    alert_distribution: Dict[AlertLevel, int]
    recent_anomalies: List[str]
    system_health: str


class MonitoringState(TypedDict):
    """State for monitoring workflow."""
    component: str
    metrics: List[Dict[str, Any]]
    health_status: str
    alerts: List[str]
    analysis: str
    recommendations: List[str]


class MonitoringEngine:
    """
    Core monitoring engine with LangChain/LangGraph integration.
    """
    
    def __init__(self, model_name: str = "gpt-4", temperature: float = 0.3):
        self.llm = ChatOpenAI(model=model_name, temperature=temperature)
        self.metrics: Dict[str, List[Metric]] = {}
        self.health_checks: Dict[str, HealthCheck] = {}
        self.alerts: Dict[str, Alert] = {}
        self.thresholds: Dict[str, Threshold] = {}
        self.time_series: Dict[str, TimeSeriesData] = {}
        self.monitoring_strategies: Dict[str, MonitoringStrategy] = {}
        self.anomaly_history: List[Dict[str, Any]] = []
        self.workflow = self._build_monitoring_workflow()
        
        # Statistics
        self.stats = {
            "total_metrics_collected": 0,
            "alerts_triggered": 0,
            "anomalies_detected": 0,
            "health_checks_performed": 0,
            "alerts_by_level": {level: 0 for level in AlertLevel},
            "components_monitored": set()
        }
    
    def _build_monitoring_workflow(self) -> StateGraph:
        """Build LangGraph workflow for intelligent monitoring."""
        workflow = StateGraph(MonitoringState)
        
        # Add nodes
        workflow.add_node("collect_metrics", self._collect_metrics_node)
        workflow.add_node("analyze_health", self._analyze_health_node)
        workflow.add_node("detect_anomalies", self._detect_anomalies_node)
        workflow.add_node("generate_alerts", self._generate_alerts_node)
        workflow.add_node("provide_recommendations", self._provide_recommendations_node)
        
        # Define edges
        workflow.set_entry_point("collect_metrics")
        workflow.add_edge("collect_metrics", "analyze_health")
        workflow.add_edge("analyze_health", "detect_anomalies")
        workflow.add_edge("detect_anomalies", "generate_alerts")
        workflow.add_edge("generate_alerts", "provide_recommendations")
        workflow.add_edge("provide_recommendations", END)
        
        return workflow.compile()
    
    def _collect_metrics_node(self, state: MonitoringState) -> MonitoringState:
        """Collect and organize metrics."""
        component = state["component"]
        
        # Get recent metrics for component
        metrics_data = []
        if component in self.metrics:
            recent_metrics = self.metrics[component][-10:]
            metrics_data = [
                {
                    "name": m.name,
                    "value": m.value,
                    "unit": m.unit,
                    "timestamp": m.timestamp.isoformat()
                }
                for m in recent_metrics
            ]
        
        state["metrics"] = metrics_data
        return state
    
    def _analyze_health_node(self, state: MonitoringState) -> MonitoringState:
        """Analyze component health using LLM."""
        metrics = state["metrics"]
        
        if not metrics:
            state["health_status"] = HealthStatus.HEALTHY.value
            state["analysis"] = "No metrics available for analysis"
            return state
        
        # Use LLM to analyze metrics
        prompt = ChatPromptTemplate.from_template(
            "Analyze the following metrics and determine system health:\n"
            "Metrics: {metrics}\n\n"
            "Provide a health status (HEALTHY, DEGRADED, or UNHEALTHY) and brief analysis."
        )
        
        chain = prompt | self.llm
        response = chain.invoke({"metrics": str(metrics)})
        analysis = response.content
        
        # Determine health status
        analysis_lower = analysis.lower()
        if "unhealthy" in analysis_lower:
            health_status = HealthStatus.UNHEALTHY.value
        elif "degraded" in analysis_lower:
            health_status = HealthStatus.DEGRADED.value
        else:
            health_status = HealthStatus.HEALTHY.value
        
        state["health_status"] = health_status
        state["analysis"] = analysis
        
        return state
    
    def _detect_anomalies_node(self, state: MonitoringState) -> MonitoringState:
        """Detect anomalies in metrics using statistical analysis."""
        component = state["component"]
        alerts = []
        
        # Check thresholds
        if component in self.metrics:
            recent_metrics = self.metrics[component][-5:]
            for metric in recent_metrics:
                if metric.name in self.thresholds:
                    threshold = self.thresholds[metric.name]
                    if self._check_threshold(metric.value, threshold):
                        alerts.append(
                            f"Threshold exceeded: {metric.name} = {metric.value} "
                            f"(threshold: {threshold.value})"
                        )
        
        # Statistical anomaly detection
        if component in self.time_series:
            ts_data = self.time_series[component]
            if len(ts_data.values) >= 5:
                mean = statistics.mean(ts_data.values)
                stdev = statistics.stdev(ts_data.values)
                latest = ts_data.values[-1]
                
                # Check if latest value is an outlier (> 2 standard deviations)
                if abs(latest - mean) > 2 * stdev:
                    alerts.append(
                        f"Anomaly detected: {ts_data.metric_name} = {latest:.2f} "
                        f"(mean: {mean:.2f}, stdev: {stdev:.2f})"
                    )
        
        state["alerts"] = alerts
        return state
    
    def _generate_alerts_node(self, state: MonitoringState) -> MonitoringState:
        """Generate and categorize alerts."""
        alerts = state.get("alerts", [])
        
        if alerts:
            # Use LLM to categorize alert severity
            prompt = ChatPromptTemplate.from_template(
                "Categorize these monitoring alerts by severity (INFO, WARNING, ERROR, CRITICAL):\n"
                "Alerts: {alerts}\n\n"
                "Provide severity level for each alert."
            )
            
            chain = prompt | self.llm
            response = chain.invoke({"alerts": str(alerts)})
            
            # Update alerts with LLM categorization
            categorized_alerts = []
            for alert in alerts:
                severity = "WARNING"  # Default
                if "critical" in response.content.lower():
                    severity = "CRITICAL"
                elif "error" in response.content.lower():
                    severity = "ERROR"
                categorized_alerts.append(f"[{severity}] {alert}")
            
            state["alerts"] = categorized_alerts
        
        return state
    
    def _provide_recommendations_node(self, state: MonitoringState) -> MonitoringState:
        """Provide recommendations based on analysis."""
        health_status = state.get("health_status", "HEALTHY")
        alerts = state.get("alerts", [])
        analysis = state.get("analysis", "")
        
        recommendations = []
        
        if health_status != HealthStatus.HEALTHY.value or alerts:
            # Use LLM to generate recommendations
            prompt = ChatPromptTemplate.from_template(
                "Based on the following system analysis, provide 3-5 actionable recommendations:\n"
                "Health Status: {health_status}\n"
                "Analysis: {analysis}\n"
                "Alerts: {alerts}\n\n"
                "Provide specific, actionable recommendations."
            )
            
            chain = prompt | self.llm
            response = chain.invoke({
                "health_status": health_status,
                "analysis": analysis,
                "alerts": str(alerts)
            })
            
            # Parse recommendations
            recommendations = [
                line.strip() 
                for line in response.content.split('\n') 
                if line.strip() and len(line.strip()) > 10
            ][:5]
        
        state["recommendations"] = recommendations
        return state
    
    def _check_threshold(self, value: float, threshold: Threshold) -> bool:
        """Check if value breaches threshold."""
        if threshold.threshold_type == ThresholdType.GREATER_THAN:
            return value > threshold.value
        elif threshold.threshold_type == ThresholdType.LESS_THAN:
            return value < threshold.value
        elif threshold.threshold_type == ThresholdType.EQUALS:
            return abs(value - threshold.value) < 0.01
        elif threshold.threshold_type == ThresholdType.RANGE:
            return not (threshold.value <= value  ts_data.window_size:
            ts_data.values = ts_data.values[-ts_data.window_size:]
            ts_data.timestamps = ts_data.timestamps[-ts_data.window_size:]
        
        # Update statistics
        self.stats["total_metrics_collected"] += 1
        self.stats["components_monitored"].add(component)
    
    def set_threshold(
        self,
        metric_name: str,
        value: float,
        threshold_type: ThresholdType = ThresholdType.GREATER_THAN,
        alert_level: AlertLevel = AlertLevel.WARNING,
        max_value: Optional[float] = None
    ):
        """Set threshold for a metric."""
        self.thresholds[metric_name] = Threshold(
            metric_name=metric_name,
            threshold_type=threshold_type,
            value=value,
            alert_level=alert_level,
            max_value=max_value
        )
    
    def check_health(self, component: str) -> HealthCheck:
        """Perform comprehensive health check using workflow."""
        initial_state: MonitoringState = {
            "component": component,
            "metrics": [],
            "health_status": HealthStatus.HEALTHY.value,
            "alerts": [],
            "analysis": "",
            "recommendations": []
        }
        
        # Run workflow
        result = self.workflow.invoke(initial_state)
        
        # Create health check
        status = HealthStatus(result["health_status"])
        health_check = HealthCheck(
            component=component,
            status=status,
            message=result["analysis"],
            metrics=self.metrics.get(component, [])[-5:]
        )
        
        self.health_checks[component] = health_check
        self.stats["health_checks_performed"] += 1
        
        # Create alerts if needed
        for alert_msg in result["alerts"]:
            alert = Alert(
                alert_id=f"alert_{len(self.alerts)}",
                level=AlertLevel.WARNING,
                component=component,
                message=alert_msg,
                metric_name=component,
                metric_value=0.0,
                threshold=0.0
            )
            self.alerts[alert.alert_id] = alert
            self.stats["alerts_triggered"] += 1
            self.stats["alerts_by_level"][AlertLevel.WARNING] += 1
        
        return health_check
    
    def get_metric_stats(self, component: str, metric_name: str) -> Dict[str, Any]:
        """Get statistics for a metric."""
        if component not in self.metrics:
            return {}
        
        values = [
            m.value for m in self.metrics[component]
            if m.name == metric_name
        ]
        
        if not values:
            return {}
        
        return {
            "count": len(values),
            "min": min(values),
            "max": max(values),
            "avg": statistics.mean(values),
            "median": statistics.median(values),
            "stdev": statistics.stdev(values) if len(values) > 1 else 0,
            "latest": values[-1]
        }
    
    def get_dashboard(self) -> Dashboard:
        """Get comprehensive monitoring dashboard."""
        components = {
            component: check.status
            for component, check in self.health_checks.items()
        }
        
        total_metrics = sum(len(metrics) for metrics in self.metrics.values())
        active_alerts = len([a for a in self.alerts.values() if not a.acknowledged])
        
        alert_distribution = {
            level: sum(
                1 for a in self.alerts.values()
                if a.level == level and not a.acknowledged
            )
            for level in AlertLevel
        }
        
        recent_anomalies = [
            a["message"] for a in self.anomaly_history[-5:]
        ]
        
        # Determine overall system health
        unhealthy_count = sum(
            1 for status in components.values()
            if status == HealthStatus.UNHEALTHY
        )
        degraded_count = sum(
            1 for status in components.values()
            if status == HealthStatus.DEGRADED
        )
        
        if unhealthy_count > 0:
            system_health = "UNHEALTHY"
        elif degraded_count > 0:
            system_health = "DEGRADED"
        else:
            system_health = "HEALTHY"
        
        return Dashboard(
            components=components,
            total_metrics=total_metrics,
            active_alerts=active_alerts,
            alert_distribution=alert_distribution,
            recent_anomalies=recent_anomalies,
            system_health=system_health
        )
    
    def acknowledge_alert(self, alert_id: str):
        """Acknowledge an alert."""
        if alert_id in self.alerts:
            self.alerts[alert_id].acknowledged = True
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get monitoring statistics."""
        return {
            "total_metrics_collected": self.stats["total_metrics_collected"],
            "alerts_triggered": self.stats["alerts_triggered"],
            "anomalies_detected": self.stats["anomalies_detected"],
            "health_checks_performed": self.stats["health_checks_performed"],
            "alerts_by_level": {
                level.value: count
                for level, count in self.stats["alerts_by_level"].items()
            },
            "components_monitored": len(self.stats["components_monitored"]),
            "active_alerts": len([a for a in self.alerts.values() if not a.acknowledged]),
            "thresholds_configured": len(self.thresholds)
        }


class MonitoringPattern:
    """
    Comprehensive monitoring pattern coordinator with LangChain/LangGraph.
    """
    
    def __init__(self, model_name: str = "gpt-4"):
        self.engine = MonitoringEngine(model_name=model_name)
    
    def monitor(
        self,
        component: str,
        metric_name: str,
        value: float,
        unit: str = ""
    ) -> Dict[str, Any]:
        """Monitor a component with automated analysis."""
        # Record metric
        self.engine.record_metric(component, metric_name, value, unit)
        
        # Perform health check
        health_check = self.engine.check_health(component)
        
        return {
            "component": component,
            "health_status": health_check.status.value,
            "message": health_check.message,
            "alerts": [
                alert.message
                for alert in self.engine.alerts.values()
                if alert.component == component and not alert.acknowledged
            ]
        }
    
    def configure_monitoring(
        self,
        metric_name: str,
        threshold: float,
        threshold_type: ThresholdType = ThresholdType.GREATER_THAN
    ):
        """Configure monitoring thresholds."""
        self.engine.set_threshold(metric_name, threshold, threshold_type)
    
    def get_dashboard(self) -> Dict[str, Any]:
        """Get monitoring dashboard."""
        dashboard = self.engine.get_dashboard()
        return {
            "system_health": dashboard.system_health,
            "components": {
                comp: status.value
                for comp, status in dashboard.components.items()
            },
            "total_metrics": dashboard.total_metrics,
            "active_alerts": dashboard.active_alerts,
            "recent_anomalies": dashboard.recent_anomalies
        }
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get monitoring statistics."""
        return self.engine.get_statistics()


if __name__ == "__main__":
    print("=== Monitoring Pattern with LangChain/LangGraph Demo ===\n")
    
    # Note: This demo shows the structure. In production, you'd need:
    # - export OPENAI_API_KEY="your-key-here"
    # - Actual LLM calls would be made
    
    print("Scenario 1: Basic Monitoring")
    print("-" * 50)
    monitor = MonitoringPattern(model_name="gpt-4")
    
    # Configure thresholds
    monitor.configure_monitoring("cpu_usage", 80.0, ThresholdType.GREATER_THAN)
    monitor.configure_monitoring("memory_usage", 90.0, ThresholdType.GREATER_THAN)
    monitor.configure_monitoring("response_time", 1000.0, ThresholdType.GREATER_THAN)
    
    print("✓ Configured monitoring thresholds")
    
    # Monitor different components
    components = [
        ("api_server", "cpu_usage", 65.0, "%"),
        ("api_server", "memory_usage", 78.0, "%"),
        ("api_server", "response_time", 250.0, "ms"),
        ("database", "cpu_usage", 45.0, "%"),
        ("database", "query_time", 150.0, "ms"),
        ("cache", "hit_rate", 92.0, "%"),
    ]
    
    print("\nMonitoring components:")
    for component, metric, value, unit in components:
        result = monitor.monitor(component, metric, value, unit)
        print(f"  {component}.{metric}: {value}{unit} -> {result['health_status']}")
    
    print("\nScenario 2: Threshold Violations")
    print("-" * 50)
    
    # Simulate threshold violations
    violations = [
        ("api_server", "cpu_usage", 85.0, "%"),
        ("database", "cpu_usage", 92.0, "%"),
        ("api_server", "response_time", 1500.0, "ms"),
    ]
    
    print("Simulating threshold violations:")
    for component, metric, value, unit in violations:
        result = monitor.monitor(component, metric, value, unit)
        print(f"\n  {component}.{metric}: {value}{unit}")
        print(f"    Status: {result['health_status']}")
        print(f"    Message: {result['message'][:100]}...")
        if result['alerts']:
            print(f"    Alerts: {len(result['alerts'])} active")
    
    print("\nScenario 3: Monitoring Dashboard")
    print("-" * 50)
    
    dashboard = monitor.get_dashboard()
    print(f"System Health: {dashboard['system_health']}")
    print(f"Total Metrics: {dashboard['total_metrics']}")
    print(f"Active Alerts: {dashboard['active_alerts']}")
    
    print("\nComponent Health:")
    for component, status in dashboard['components'].items():
        print(f"  {component}: {status}")
    
    if dashboard['recent_anomalies']:
        print("\nRecent Anomalies:")
        for anomaly in dashboard['recent_anomalies']:
            print(f"  - {anomaly}")
    
    print("\nScenario 4: Statistics")
    print("-" * 50)
    
    stats = monitor.get_statistics()
    print(f"Total Metrics Collected: {stats['total_metrics_collected']}")
    print(f"Alerts Triggered: {stats['alerts_triggered']}")
    print(f"Health Checks Performed: {stats['health_checks_performed']}")
    print(f"Components Monitored: {stats['components_monitored']}")
    print(f"Thresholds Configured: {stats['thresholds_configured']}")
    
    print("\nAlerts by Level:")
    for level, count in stats['alerts_by_level'].items():
        print(f"  {level}: {count}")
    
    print("\n" + "=" * 50)
    print("Monitoring pattern demonstration complete!")
    print("=" * 50)

    

Agentic Architecture Pattern: Monitoring

Pattern 171: 171 Monitoring – Line-by-Line Explanation

Overview

Pattern 171: Monitoring Pattern with LangChain/LangGraph

Purpose: Track agent performance, health, and metrics in real-time with LangGraph integration.

Code Structure Analysis

File Information

• Pattern Number: 171
• Pattern Name: 171 Monitoring
• File Path: patterns\quality_performance\171_monitoring.py
• Total Lines: 677

Detailed Code Explanation

Module Docstring

"""
Pattern 171: Monitoring Pattern with LangChain/LangGraph

Purpose: Track agent performance, health, and metrics in real-time with LangGraph integration.
"""

Explanation: This module docstring describes the purpose and key features of the 171 Monitoring pattern.

Imports Section

from typing import Dict, Any, List, Optional, TypedDict
from dataclasses import dataclass, field
from datetime import datetime, timedelta
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
import statistics  # First 20 imports shown

Explanation: The module imports necessary dependencies: – from typing import Dict, Any, List, Optional, TypedDict: Type hints for better code documentation – from dataclasses import dataclass, field: Data class decorators for clean data structures – from datetime import datetime, timedelta: Date and time utilities – from enum import Enum: Enumeration support for fixed sets of values – from langchain_core.prompts import ChatPromptTemplate: LangChain framework integration – from langchain_openai import ChatOpenAI: LangChain framework integration – from langgraph.graph import StateGraph, END: LangGraph for workflow orchestration

Enumerations

HealthStatus (Line 17)

class HealthStatus(Enum):
    HEALTHY
    DEGRADED
    UNHEALTHY

Purpose: Health status levels.

Values: – HEALTHY: Represents healthy – DEGRADED: Represents degraded – UNHEALTHY: Represents unhealthy

MonitoringStrategy (Line 24)

class MonitoringStrategy(Enum):
    REAL_TIME
    PERIODIC
    THRESHOLD_BASED
    ANOMALY_DETECTION

Purpose: Monitoring strategies.

Values: – REAL_TIME: Represents real time – PERIODIC: Represents periodic – THRESHOLD_BASED: Represents threshold based – ANOMALY_DETECTION: Represents anomaly detection

AlertLevel (Line 32)

class AlertLevel(Enum):
    INFO
    WARNING
    ERROR
    CRITICAL

Purpose: Alert severity levels.

Values: – INFO: Informational severity level – WARNING: Warning severity level – ERROR: Error severity level – CRITICAL: Represents critical

ThresholdType (Line 40)

class ThresholdType(Enum):
    GREATER_THAN
    LESS_THAN
    EQUALS
    RANGE

Purpose: Threshold comparison types.

Values: – GREATER_THAN: Represents greater than – LESS_THAN: Represents less than – EQUALS: Represents equals – RANGE: Represents range

Classes

HealthStatus (Line 17)

Description: Health status levels.

MonitoringStrategy (Line 24)

Description: Monitoring strategies.

AlertLevel (Line 32)

Description: Alert severity levels.

ThresholdType (Line 40)

Description: Threshold comparison types.

Metric (Line 49)

Description: Performance metric.

HealthCheck (Line 59)

Description: Health check result.

Alert (Line 69)

Description: Monitoring alert.

Threshold (Line 83)

Description: Metric threshold configuration.

TimeSeriesData (Line 93)

Description: Time-series data for a metric.

Dashboard (Line 102)

Description: Monitoring dashboard data.

MonitoringState (Line 112)

Description: State for monitoring workflow.

MonitoringEngine (Line 122)

Description: Core monitoring engine with LangChain/LangGraph integration.

Methods:

__init__ (Line 127)

def __init__(self, model_name, temperature):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – model_name: LLM model to use (e.g., “gpt-4”) – temperature: Randomness in LLM responses (0.0-1.0)

_build_monitoring_workflow (Line 148)

def _build_monitoring_workflow(self):
    """Build LangGraph workflow for intelligent monitoring."""

Purpose: Build LangGraph workflow for intelligent monitoring.

Parameters: – self: Instance reference

_collect_metrics_node (Line 169)

def _collect_metrics_node(self, state):
    """Collect and organize metrics."""

Purpose: Collect and organize metrics.

Parameters: – self: Instance reference – state: Current workflow state

_analyze_health_node (Line 190)

def _analyze_health_node(self, state):
    """Analyze component health using LLM."""

Purpose: Analyze component health using LLM.

Parameters: – self: Instance reference – state: Current workflow state

_detect_anomalies_node (Line 224)

def _detect_anomalies_node(self, state):
    """Detect anomalies in metrics using statistical analysis."""

Purpose: Detect anomalies in metrics using statistical analysis.

Parameters: – self: Instance reference – state: Current workflow state

_generate_alerts_node (Line 259)

def _generate_alerts_node(self, state):
    """Generate and categorize alerts."""

Purpose: Generate and categorize alerts.

Parameters: – self: Instance reference – state: Current workflow state

_provide_recommendations_node (Line 288)

def _provide_recommendations_node(self, state):
    """Provide recommendations based on analysis."""

Purpose: Provide recommendations based on analysis.

Parameters: – self: Instance reference – state: Current workflow state

_check_threshold (Line 323)

def _check_threshold(self, value, threshold):
    """Check if value breaches threshold."""

Purpose: Check if value breaches threshold.

Parameters: – self: Instance reference – value: Parameter for value – threshold: Parameter for threshold

record_metric (Line 335)

def record_metric(self, component, name, value, unit, tags):
    """Record a metric value."""

Purpose: Record a metric value.

Parameters: – self: Instance reference – component: Parameter for component – name: Identifier or label – value: Parameter for value – unit: Parameter for unit – tags: Parameter for tags

set_threshold (Line 377)

def set_threshold(self, metric_name, value, threshold_type, alert_level, max_value):
    """Set threshold for a metric."""

Purpose: Set threshold for a metric.

Parameters: – self: Instance reference – metric_name: Parameter for metric_name – value: Parameter for value – threshold_type: Parameter for threshold_type – alert_level: Parameter for alert_level – max_value: Parameter for max_value

check_health (Line 394)

def check_health(self, component):
    """Perform comprehensive health check using workflow."""

Purpose: Perform comprehensive health check using workflow.

Parameters: – self: Instance reference – component: Parameter for component

get_metric_stats (Line 437)

def get_metric_stats(self, component, metric_name):
    """Get statistics for a metric."""

Purpose: Get statistics for a metric.

Parameters: – self: Instance reference – component: Parameter for component – metric_name: Parameter for metric_name

get_dashboard (Line 460)

def get_dashboard(self):
    """Get comprehensive monitoring dashboard."""

Purpose: Get comprehensive monitoring dashboard.

Parameters: – self: Instance reference

acknowledge_alert (Line 508)

def acknowledge_alert(self, alert_id):
    """Acknowledge an alert."""

Purpose: Acknowledge an alert.

Parameters: – self: Instance reference – alert_id: Parameter for alert_id

get_statistics (Line 513)

def get_statistics(self):
    """Get monitoring statistics."""

Purpose: Get monitoring statistics.

Parameters: – self: Instance reference

MonitoringPattern (Line 530)

Description: Comprehensive monitoring pattern coordinator with LangChain/LangGraph.

Methods:

__init__ (Line 535)

def __init__(self, model_name):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – model_name: LLM model to use (e.g., “gpt-4”)

monitor (Line 538)

def monitor(self, component, metric_name, value, unit):
    """Monitor a component with automated analysis."""

Purpose: Monitor a component with automated analysis.

Parameters: – self: Instance reference – component: Parameter for component – metric_name: Parameter for metric_name – value: Parameter for value – unit: Parameter for unit

configure_monitoring (Line 563)

def configure_monitoring(self, metric_name, threshold, threshold_type):
    """Configure monitoring thresholds."""

Purpose: Configure monitoring thresholds.

Parameters: – self: Instance reference – metric_name: Parameter for metric_name – threshold: Parameter for threshold – threshold_type: Parameter for threshold_type

get_dashboard (Line 572)

def get_dashboard(self):
    """Get monitoring dashboard."""

Purpose: Get monitoring dashboard.

Parameters: – self: Instance reference

get_statistics (Line 586)

def get_statistics(self):
    """Get monitoring statistics."""

Purpose: Get monitoring statistics.

Parameters: – self: Instance reference

Design Patterns Used

• Strategy Pattern: Different execution strategies for various scenarios
• State Pattern: Managing different states in the workflow
• Chain of Responsibility: Sequential processing through workflow nodes
• Observer Pattern: Tracking and monitoring pattern execution
• Factory Pattern: Creating instances with proper configuration

Usage Example

# Initialize the pattern
pattern = HealthStatus(name="example", model_name="gpt-4")

# Configure and execute
result = pattern.execute(data={"key": "value"})

# Check results
print(result)

Key Concepts

• LLM Integration: Uses LangChain/OpenAI for intelligent decision-making
• Workflow Orchestration: LangGraph manages execution flow through nodes
• State Management: TypedDict defines workflow state structure
• Error Handling: Comprehensive error handling and fallback mechanisms
• Extensibility: Designed for easy customization and extension

Extension Points

This pattern can be extended by: 1. Adding custom workflow nodes 2. Implementing additional validation logic 3. Integrating with external services 4. Customizing LLM prompts and responses 5. Adding monitoring and logging capabilities

    """
Pattern 170: Partitioning Pattern with LangChain/LangGraph

Purpose: Divide tasks/data into logical partitions for parallel processing with intelligent strategies.
Different from sharding - partitions can be dynamic and application-specific.
"""

from typing import Dict, Any, List, Optional, Callable, TypedDict
from dataclasses import dataclass, field
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
import statistics
import hashlib


class PartitionStrategy(Enum):
    """Partitioning strategies."""
    BY_SIZE = "by_size"
    BY_KEY = "by_key"
    BY_TYPE = "by_type"
    BY_PRIORITY = "by_priority"
    BY_HASH = "by_hash"
    BY_RANGE = "by_range"
    ADAPTIVE = "adaptive"


class LoadBalanceStrategy(Enum):
    """Load balancing strategies."""
    ROUND_ROBIN = "round_robin"
    LEAST_LOADED = "least_loaded"
    WEIGHTED = "weighted"
    RANDOM = "random"


class PartitionHealth(Enum):
    """Partition health status."""
    BALANCED = "balanced"
    SKEWED = "skewed"
    OVERLOADED = "overloaded"
    UNDERUTILIZED = "underutilized"


@dataclass
class Partition:
    """Logical partition of data/tasks with metadata."""
    id: str
    strategy: PartitionStrategy
    criteria: Any
    items: List[Any] = field(default_factory=list)
    metadata: Dict[str, Any] = field(default_factory=dict)
    weight: float = 1.0
    
    def size(self) -> int:
        """Get partition size."""
        return len(self.items)


@dataclass
class PartitionAssignment:
    """Assignment of partition to agent."""
    partition_id: str
    agent_id: str
    assigned_at: float
    weight: float = 1.0


@dataclass
class PartitionMetrics:
    """Metrics for a partition."""
    partition_id: str
    item_count: int
    avg_item_size: float
    load_factor: float
    health: PartitionHealth


@dataclass
class RebalanceResult:
    """Result of rebalancing operation."""
    partitions_moved: int
    items_redistributed: int
    balance_improvement: float
    new_partition_count: int


class PartitioningState(TypedDict):
    """State for partitioning workflow."""
    items: List[Any]
    strategy: str
    partition_count: int
    partitions: List[Dict[str, Any]]
    metrics: Dict[str, Any]
    recommendations: List[str]
    balance_score: float


class IntelligentPartitioner:
    """
    Intelligent partitioner with LangChain/LangGraph integration.
    """
    
    def __init__(self, model_name: str = "gpt-4", temperature: float = 0.3):
        self.llm = ChatOpenAI(model=model_name, temperature=temperature)
        self.partitions: Dict[str, Partition] = {}
        self.assignments: Dict[str, PartitionAssignment] = {}
        self.workflow = self._build_partitioning_workflow()
        
        # Statistics
        self.stats = {
            "total_partitions_created": 0,
            "total_items_partitioned": 0,
            "rebalance_operations": 0,
            "assignments_made": 0,
            "partitions_by_strategy": {strategy: 0 for strategy in PartitionStrategy}
        }
    
    def _build_partitioning_workflow(self) -> StateGraph:
        """Build LangGraph workflow for intelligent partitioning."""
        workflow = StateGraph(PartitioningState)
        
        # Add nodes
        workflow.add_node("analyze_data", self._analyze_data_node)
        workflow.add_node("select_strategy", self._select_strategy_node)
        workflow.add_node("create_partitions", self._create_partitions_node)
        workflow.add_node("evaluate_balance", self._evaluate_balance_node)
        workflow.add_node("provide_recommendations", self._provide_recommendations_node)
        
        # Define edges
        workflow.set_entry_point("analyze_data")
        workflow.add_edge("analyze_data", "select_strategy")
        workflow.add_edge("select_strategy", "create_partitions")
        workflow.add_edge("create_partitions", "evaluate_balance")
        workflow.add_edge("evaluate_balance", "provide_recommendations")
        workflow.add_edge("provide_recommendations", END)
        
        return workflow.compile()
    
    def _analyze_data_node(self, state: PartitioningState) -> PartitioningState:
        """Analyze data characteristics."""
        items = state["items"]
        
        metrics = {
            "total_items": len(items),
            "has_keys": all(isinstance(item, dict) for item in items),
            "item_types": len(set(type(item).__name__ for item in items)),
        }
        
        state["metrics"] = metrics
        return state
    
    def _select_strategy_node(self, state: PartitioningState) -> PartitioningState:
        """Select optimal partitioning strategy using LLM."""
        metrics = state["metrics"]
        requested_strategy = state.get("strategy", "")
        
        if requested_strategy:
            state["strategy"] = requested_strategy
        else:
            # Use LLM to recommend strategy
            prompt = ChatPromptTemplate.from_template(
                "Based on the data characteristics, recommend a partitioning strategy:\n"
                "Total Items: {total_items}\n"
                "Has Dictionary Keys: {has_keys}\n"
                "Item Type Diversity: {item_types}\n\n"
                "Choose from: by_size, by_key, by_type, by_priority, by_hash, by_range, adaptive\n"
                "Provide just the strategy name."
            )
            
            chain = prompt | self.llm
            response = chain.invoke(metrics)
            
            # Extract strategy from response
            strategy_str = str(response.content).lower().strip()
            if "size" in strategy_str:
                state["strategy"] = PartitionStrategy.BY_SIZE.value
            elif "key" in strategy_str:
                state["strategy"] = PartitionStrategy.BY_KEY.value
            elif "hash" in strategy_str:
                state["strategy"] = PartitionStrategy.BY_HASH.value
            else:
                state["strategy"] = PartitionStrategy.ADAPTIVE.value
        
        return state
    
    def _create_partitions_node(self, state: PartitioningState) -> PartitioningState:
        """Create partitions based on selected strategy."""
        items = state["items"]
        strategy = state["strategy"]
        partition_count = state.get("partition_count", 4)
        
        partitions_data = []
        
        if strategy == PartitionStrategy.BY_SIZE.value:
            size = max(1, len(items) // partition_count)
            for i in range(0, len(items), size):
                partitions_data.append({
                    "id": f"size_partition_{len(partitions_data)}",
                    "strategy": strategy,
                    "item_count": min(size, len(items) - i)
                })
        else:
            # Default partitioning
            for i in range(partition_count):
                partitions_data.append({
                    "id": f"partition_{i}",
                    "strategy": strategy,
                    "item_count": 0
                })
        
        state["partitions"] = partitions_data
        return state
    
    def _evaluate_balance_node(self, state: PartitioningState) -> PartitioningState:
        """Evaluate partition balance."""
        partitions = state["partitions"]
        
        if not partitions:
            state["balance_score"] = 0.0
            return state
        
        # Calculate balance score
        sizes = [p["item_count"] for p in partitions]
        if sizes and max(sizes) > 0:
            avg_size = statistics.mean(sizes)
            variance = statistics.variance(sizes) if len(sizes) > 1 else 0
            balance_score = 1.0 / (1.0 + variance / (avg_size ** 2) if avg_size > 0 else 1.0)
        else:
            balance_score = 1.0
        
        state["balance_score"] = balance_score
        return state
    
    def _provide_recommendations_node(self, state: PartitioningState) -> PartitioningState:
        """Provide recommendations for partitioning."""
        balance_score = state.get("balance_score", 0.0)
        partitions = state["partitions"]
        
        recommendations = []
        
        if balance_score < 0.7:
            recommendations.append("Partitions are unbalanced - consider rebalancing")
        
        if len(partitions)  state["metrics"]["total_items"]:
            recommendations.append("Too many partitions - some will be empty")
        
        state["recommendations"] = recommendations
        return state
    
    def partition_by_size(
        self,
        items: List[Any],
        partition_size: int,
        prefix: str = "size"
    ) -> List[Partition]:
        """Partition by fixed size."""
        partitions = []
        
        for i in range(0, len(items), partition_size):
            partition_id = f"{prefix}_partition_{len(partitions)}"
            partition = Partition(
                id=partition_id,
                strategy=PartitionStrategy.BY_SIZE,
                criteria=partition_size,
                items=items[i:i+partition_size]
            )
            partitions.append(partition)
            self.partitions[partition_id] = partition
            self.stats["total_partitions_created"] += 1
        
        self.stats["total_items_partitioned"] += len(items)
        self.stats["partitions_by_strategy"][PartitionStrategy.BY_SIZE] += len(partitions)
        
        return partitions
    
    def partition_by_key(
        self,
        items: List[Dict],
        key: str,
        prefix: str = "key"
    ) -> List[Partition]:
        """Partition by key value."""
        groups: Dict[Any, List[Dict]] = {}
        
        for item in items:
            key_value = item.get(key)
            if key_value not in groups:
                groups[key_value] = []
            groups[key_value].append(item)
        
        partitions = []
        for key_value, group_items in groups.items():
            partition_id = f"{prefix}_{key}_{key_value}"
            partition = Partition(
                id=partition_id,
                strategy=PartitionStrategy.BY_KEY,
                criteria=key_value,
                items=group_items,
                metadata={"key": key, "value": key_value}
            )
            partitions.append(partition)
            self.partitions[partition_id] = partition
            self.stats["total_partitions_created"] += 1
        
        self.stats["total_items_partitioned"] += len(items)
        self.stats["partitions_by_strategy"][PartitionStrategy.BY_KEY] += len(partitions)
        
        return partitions
    
    def partition_by_hash(
        self,
        items: List[Any],
        num_partitions: int,
        hash_func: Optional[Callable[[Any], str]] = None
    ) -> List[Partition]:
        """Partition using consistent hashing."""
        if not hash_func:
            hash_func = lambda x: hashlib.md5(str(x).encode()).hexdigest()
        
        # Initialize partitions
        partition_map: Dict[int, List[Any]] = {i: [] for i in range(num_partitions)}
        
        for item in items:
            hash_value = hash_func(item)
            partition_idx = int(hash_value, 16) % num_partitions
            partition_map[partition_idx].append(item)
        
        partitions = []
        for idx, partition_items in partition_map.items():
            partition_id = f"hash_partition_{idx}"
            partition = Partition(
                id=partition_id,
                strategy=PartitionStrategy.BY_HASH,
                criteria=idx,
                items=partition_items
            )
            partitions.append(partition)
            self.partitions[partition_id] = partition
            self.stats["total_partitions_created"] += 1
        
        self.stats["total_items_partitioned"] += len(items)
        self.stats["partitions_by_strategy"][PartitionStrategy.BY_HASH] += len(partitions)
        
        return partitions
    
    def partition_by_predicate(
        self,
        items: List[Any],
        predicates: Dict[str, Callable[[Any], bool]]
    ) -> List[Partition]:
        """Partition using custom predicates."""
        partitions = []
        
        for name, predicate in predicates.items():
            matching_items = [item for item in items if predicate(item)]
            partition_id = f"predicate_{name}"
            partition = Partition(
                id=partition_id,
                strategy=PartitionStrategy.BY_TYPE,
                criteria=name,
                items=matching_items,
                metadata={"predicate": name}
            )
            partitions.append(partition)
            self.partitions[partition_id] = partition
            self.stats["total_partitions_created"] += 1
        
        self.stats["total_items_partitioned"] += sum(len(p.items) for p in partitions)
        self.stats["partitions_by_strategy"][PartitionStrategy.BY_TYPE] += len(partitions)
        
        return partitions
    
    def intelligent_partition(
        self,
        items: List[Any],
        target_partitions: int = 4
    ) -> Dict[str, Any]:
        """Use workflow to intelligently partition data."""
        initial_state: PartitioningState = {
            "items": items,
            "strategy": "",
            "partition_count": target_partitions,
            "partitions": [],
            "metrics": {},
            "recommendations": [],
            "balance_score": 0.0
        }
        
        # Run workflow
        result = self.workflow.invoke(initial_state)
        
        return {
            "strategy": result["strategy"],
            "partitions": result["partitions"],
            "balance_score": result["balance_score"],
            "recommendations": result["recommendations"]
        }
    
    def assign_partition(
        self,
        partition_id: str,
        agent_id: str,
        weight: float = 1.0
    ) -> PartitionAssignment:
        """Assign partition to agent."""
        import time
        
        assignment = PartitionAssignment(
            partition_id=partition_id,
            agent_id=agent_id,
            assigned_at=time.time(),
            weight=weight
        )
        
        self.assignments[partition_id] = assignment
        self.stats["assignments_made"] += 1
        
        return assignment
    
    def get_agent_workload(self, agent_id: str) -> List[Partition]:
        """Get all partitions assigned to agent."""
        assigned_partition_ids = [
            pid for pid, assignment in self.assignments.items()
            if assignment.agent_id == agent_id
        ]
        
        return [
            self.partitions[pid]
            for pid in assigned_partition_ids
            if pid in self.partitions
        ]
    
    def get_partition_metrics(self, partition_id: str) -> Optional[PartitionMetrics]:
        """Get metrics for a partition."""
        partition = self.partitions.get(partition_id)
        if not partition:
            return None
        
        item_count = len(partition.items)
        avg_item_size = statistics.mean(
            [len(str(item)) for item in partition.items]
        ) if partition.items else 0.0
        
        # Calculate load factor relative to other partitions
        all_sizes = [len(p.items) for p in self.partitions.values()]
        avg_size = statistics.mean(all_sizes) if all_sizes else 0
        load_factor = item_count / avg_size if avg_size > 0 else 1.0
        
        # Determine health
        if load_factor > 1.5:
            health = PartitionHealth.OVERLOADED
        elif load_factor  0.2:
            health = PartitionHealth.SKEWED
        else:
            health = PartitionHealth.BALANCED
        
        return PartitionMetrics(
            partition_id=partition_id,
            item_count=item_count,
            avg_item_size=avg_item_size,
            load_factor=load_factor,
            health=health
        )
    
    def rebalance(
        self,
        target_partition_count: Optional[int] = None
    ) -> RebalanceResult:
        """Rebalance partitions by redistributing items."""
        # Collect all items
        all_items = []
        old_partition_count = len(self.partitions)
        
        for partition in self.partitions.values():
            all_items.extend(partition.items)
        
        # Clear partitions and assignments
        self.partitions.clear()
        self.assignments.clear()
        
        # Determine new partition count
        if target_partition_count is None:
            target_partition_count = max(1, old_partition_count)
        
        # Create new balanced partitions
        items_per_partition = max(1, len(all_items) // target_partition_count)
        new_partitions = self.partition_by_size(all_items, items_per_partition, "rebalanced")
        
        self.stats["rebalance_operations"] += 1
        
        return RebalanceResult(
            partitions_moved=old_partition_count,
            items_redistributed=len(all_items),
            balance_improvement=0.5,  # Simplified
            new_partition_count=len(new_partitions)
        )
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get comprehensive partitioning statistics."""
        partition_sizes = [len(p.items) for p in self.partitions.values()]
        
        return {
            "total_partitions_created": self.stats["total_partitions_created"],
            "active_partitions": len(self.partitions),
            "total_items_partitioned": self.stats["total_items_partitioned"],
            "items_in_partitions": sum(partition_sizes),
            "rebalance_operations": self.stats["rebalance_operations"],
            "assignments_made": self.stats["assignments_made"],
            "active_assignments": len(self.assignments),
            "avg_partition_size": statistics.mean(partition_sizes) if partition_sizes else 0.0,
            "min_partition_size": min(partition_sizes) if partition_sizes else 0,
            "max_partition_size": max(partition_sizes) if partition_sizes else 0,
            "partitions_by_strategy": {
                strategy.value: count
                for strategy, count in self.stats["partitions_by_strategy"].items()
                if count > 0
            }
        }


class PartitioningPattern:
    """
    Comprehensive partitioning pattern coordinator with LangChain/LangGraph.
    """
    
    def __init__(self, model_name: str = "gpt-4"):
        self.partitioner = IntelligentPartitioner(model_name=model_name)
    
    def partition(
        self,
        items: List[Any],
        strategy: PartitionStrategy = PartitionStrategy.BY_SIZE,
        **kwargs
    ) -> List[Partition]:
        """Partition items using specified strategy."""
        if strategy == PartitionStrategy.BY_SIZE:
            size = kwargs.get("size", 10)
            return self.partitioner.partition_by_size(items, size)
        elif strategy == PartitionStrategy.BY_KEY:
            key = kwargs.get("key", "id")
            return self.partitioner.partition_by_key(items, key)
        elif strategy == PartitionStrategy.BY_HASH:
            num_partitions = kwargs.get("num_partitions", 4)
            return self.partitioner.partition_by_hash(items, num_partitions)
        else:
            return self.partitioner.partition_by_size(items, 10)
    
    def smart_partition(
        self,
        items: List[Any],
        target_partitions: int = 4
    ) -> Dict[str, Any]:
        """Use AI to determine optimal partitioning."""
        return self.partitioner.intelligent_partition(items, target_partitions)
    
    def assign(self, partition_id: str, agent_id: str) -> PartitionAssignment:
        """Assign partition to agent."""
        return self.partitioner.assign_partition(partition_id, agent_id)
    
    def get_workload(self, agent_id: str) -> List[Partition]:
        """Get agent's workload."""
        return self.partitioner.get_agent_workload(agent_id)
    
    def rebalance(self, target_count: Optional[int] = None) -> RebalanceResult:
        """Rebalance partitions."""
        return self.partitioner.rebalance(target_count)
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get statistics."""
        return self.partitioner.get_statistics()



if __name__ == "__main__":
    print("=== Partitioning Pattern with LangChain/LangGraph Demo ===\n")
    
    # Note: This demo shows the structure. In production, you'd need:
    # - export OPENAI_API_KEY="your-key-here"
    # - Actual LLM calls would be made
    
    print("Scenario 1: Basic Partitioning Strategies")
    print("-" * 50)
    partitioner = PartitioningPattern(model_name="gpt-4")
    
    # Sample dataset
    items = [
        {"id": i, "type": "A" if i % 2 == 0 else "B", "priority": i % 3, "size": i * 10}
        for i in range(50)
    ]
    
    print(f"Dataset: {len(items)} items")
    
    # Partition by size
    print("\n1. Partitioning by size (10 items per partition):")
    size_partitions = partitioner.partition(items, PartitionStrategy.BY_SIZE, size=10)
    for p in size_partitions[:3]:
        print(f"  {p.id}: {p.size()} items")
    print(f"  ... ({len(size_partitions)} total partitions)")
    
    # Partition by key
    print("\n2. Partitioning by key (type):")
    key_partitions = partitioner.partition(items, PartitionStrategy.BY_KEY, key="type")
    for p in key_partitions:
        print(f"  {p.id}: {p.size()} items (criteria: {p.criteria})")
    
    # Partition by hash
    print("\n3. Partitioning by hash (4 partitions):")
    hash_partitions = partitioner.partition(items, PartitionStrategy.BY_HASH, num_partitions=4)
    for p in hash_partitions:
        print(f"  {p.id}: {p.size()} items")
    
    print("\nScenario 2: Intelligent Partitioning")
    print("-" * 50)
    
    # Use AI to determine optimal strategy
    print("Using LLM to determine optimal partitioning strategy...")
    smart_result = partitioner.smart_partition(items, target_partitions=5)
    
    print(f"\nRecommended Strategy: {smart_result['strategy']}")
    print(f"Balance Score: {smart_result['balance_score']:.2f}")
    print(f"Partitions Created: {len(smart_result['partitions'])}")
    
    if smart_result['recommendations']:
        print("\nRecommendations:")
        for rec in smart_result['recommendations']:
            print(f"  - {rec}")
    
    print("\nScenario 3: Partition Assignment")
    print("-" * 50)
    
    # Assign partitions to agents
    agents = ["agent-001", "agent-002", "agent-003", "agent-004"]
    
    print("Assigning partitions to agents:")
    for i, partition in enumerate(hash_partitions):
        agent_id = agents[i % len(agents)]
        partitioner.assign(partition.id, agent_id)
        print(f"  {partition.id} → {agent_id}")
    
    # Get workload for each agent
    print("\nAgent workloads:")
    for agent_id in agents:
        workload = partitioner.get_workload(agent_id)
        total_items = sum(p.size() for p in workload)
        print(f"  {agent_id}: {len(workload)} partitions, {total_items} items")
    
    print("\nScenario 4: Partition Metrics")
    print("-" * 50)
    
    # Get metrics for partitions
    print("Partition health metrics:")
    for partition in hash_partitions[:3]:
        metrics = partitioner.partitioner.get_partition_metrics(partition.id)
        if metrics:
            print(f"\n  {metrics.partition_id}:")
            print(f"    Item Count: {metrics.item_count}")
            print(f"    Avg Item Size: {metrics.avg_item_size:.1f} chars")
            print(f"    Load Factor: {metrics.load_factor:.2f}")
            print(f"    Health: {metrics.health.value}")
    
    print("\nScenario 5: Dynamic Rebalancing")
    print("-" * 50)
    
    # Create imbalanced partitions
    imbalanced_items = items[:30]
    partitioner2 = PartitioningPattern(model_name="gpt-4")
    
    # Create partitions with different sizes
    partitioner2.partitioner.partition_by_size(imbalanced_items[:5], 5)
    partitioner2.partitioner.partition_by_size(imbalanced_items[5:25], 20)
    
    print(f"Before rebalancing: {len(partitioner2.partitioner.partitions)} partitions")
    
    # Show imbalance
    sizes_before = [p.size() for p in partitioner2.partitioner.partitions.values()]
    print(f"  Partition sizes: {sizes_before}")
    print(f"  Min: {min(sizes_before)}, Max: {max(sizes_before)}, Avg: {statistics.mean(sizes_before):.1f}")
    
    # Rebalance
    print("\nRebalancing partitions...")
    result = partitioner2.rebalance(target_count=4)
    
    print(f"\nRebalance Results:")
    print(f"  Partitions Moved: {result.partitions_moved}")
    print(f"  Items Redistributed: {result.items_redistributed}")
    print(f"  New Partition Count: {result.new_partition_count}")
    
    sizes_after = [p.size() for p in partitioner2.partitioner.partitions.values()]
    print(f"\nAfter rebalancing:")
    print(f"  Partition sizes: {sizes_after}")
    print(f"  Min: {min(sizes_after)}, Max: {max(sizes_after)}, Avg: {statistics.mean(sizes_after):.1f}")
    
    print("\nScenario 6: Custom Predicate Partitioning")
    print("-" * 50)
    
    # Partition by custom predicates
    predicates = {
        "high_priority": lambda x: x.get("priority") == 0,
        "medium_priority": lambda x: x.get("priority") == 1,
        "low_priority": lambda x: x.get("priority") == 2,
        "large_items": lambda x: x.get("size", 0) > 250
    }
    
    print("Partitioning by custom predicates:")
    predicate_partitions = partitioner.partitioner.partition_by_predicate(items, predicates)
    
    for p in predicate_partitions:
        print(f"  {p.metadata.get('predicate', 'unknown')}: {p.size()} items")
    
    print("\nScenario 7: Statistics")
    print("-" * 50)
    
    stats = partitioner.get_statistics()
    print(f"Total Partitions Created: {stats['total_partitions_created']}")
    print(f"Active Partitions: {stats['active_partitions']}")
    print(f"Total Items Partitioned: {stats['total_items_partitioned']}")
    print(f"Items in Partitions: {stats['items_in_partitions']}")
    print(f"Rebalance Operations: {stats['rebalance_operations']}")
    print(f"Assignments Made: {stats['assignments_made']}")
    print(f"Active Assignments: {stats['active_assignments']}")
    
    print(f"\nPartition Size Statistics:")
    print(f"  Average: {stats['avg_partition_size']:.1f} items")
    print(f"  Min: {stats['min_partition_size']} items")
    print(f"  Max: {stats['max_partition_size']} items")
    
    if stats['partitions_by_strategy']:
        print(f"\nPartitions by Strategy:")
        for strategy, count in stats['partitions_by_strategy'].items():
            print(f"  {strategy}: {count}")
    
    print("\n" + "=" * 50)
    print("Partitioning pattern demonstration complete!")
    print("=" * 50)

    

Agentic Architecture Pattern: Partitioning

Pattern 170: 170 Partitioning – Line-by-Line Explanation

Overview

Pattern 170: Partitioning Pattern with LangChain/LangGraph

Purpose: Divide tasks/data into logical partitions for parallel processing with intelligent strategies. Different from sharding – partitions can be dynamic and application-specific.

Code Structure Analysis

File Information

• Pattern Number: 170
• Pattern Name: 170 Partitioning
• File Path: patterns\scaling\170_partitioning.py
• Total Lines: 741

Detailed Code Explanation

Module Docstring

"""
Pattern 170: Partitioning Pattern with LangChain/LangGraph

Purpose: Divide tasks/data into logical partitions for parallel processing with intelligent strategies.
Different from sharding - partitions can be dynamic and application-specific.
"""

Explanation: This module docstring describes the purpose and key features of the 170 Partitioning pattern.

Imports Section

from typing import Dict, Any, List, Optional, Callable, TypedDict
from dataclasses import dataclass, field
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
import statistics
import hashlib
        import time  # First 20 imports shown

Explanation: The module imports necessary dependencies: – from typing import Dict, Any, List, Optional, Callable, TypedDict: Type hints for better code documentation – from dataclasses import dataclass, field: Data class decorators for clean data structures – from enum import Enum: Enumeration support for fixed sets of values – from langchain_core.prompts import ChatPromptTemplate: LangChain framework integration – from langchain_openai import ChatOpenAI: LangChain framework integration – from langgraph.graph import StateGraph, END: LangGraph for workflow orchestration

Enumerations

PartitionStrategy (Line 18)

class PartitionStrategy(Enum):
    BY_SIZE
    BY_KEY
    BY_TYPE
    BY_PRIORITY
    BY_HASH
    BY_RANGE
    ADAPTIVE

Purpose: Partitioning strategies.

Values: – BY_SIZE: Represents by size – BY_KEY: Represents by key – BY_TYPE: Represents by type – BY_PRIORITY: Represents by priority – BY_HASH: Represents by hash – BY_RANGE: Represents by range – ADAPTIVE: Represents adaptive

LoadBalanceStrategy (Line 29)

class LoadBalanceStrategy(Enum):
    ROUND_ROBIN
    LEAST_LOADED
    WEIGHTED
    RANDOM

Purpose: Load balancing strategies.

Values: – ROUND_ROBIN: Represents round robin – LEAST_LOADED: Represents least loaded – WEIGHTED: Represents weighted – RANDOM: Represents random

PartitionHealth (Line 37)

class PartitionHealth(Enum):
    BALANCED
    SKEWED
    OVERLOADED
    UNDERUTILIZED

Purpose: Partition health status.

Values: – BALANCED: Represents balanced – SKEWED: Represents skewed – OVERLOADED: Represents overloaded – UNDERUTILIZED: Represents underutilized

Classes

PartitionStrategy (Line 18)

Description: Partitioning strategies.

LoadBalanceStrategy (Line 29)

Description: Load balancing strategies.

PartitionHealth (Line 37)

Description: Partition health status.

Partition (Line 46)

Description: Logical partition of data/tasks with metadata.

Methods:

size (Line 55)

def size(self):
    """Get partition size."""

Purpose: Get partition size.

Parameters: – self: Instance reference

PartitionAssignment (Line 61)

Description: Assignment of partition to agent.

PartitionMetrics (Line 70)

Description: Metrics for a partition.

RebalanceResult (Line 80)

Description: Result of rebalancing operation.

PartitioningState (Line 88)

Description: State for partitioning workflow.

IntelligentPartitioner (Line 99)

Description: Intelligent partitioner with LangChain/LangGraph integration.

Methods:

__init__ (Line 104)

def __init__(self, model_name, temperature):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – model_name: LLM model to use (e.g., “gpt-4”) – temperature: Randomness in LLM responses (0.0-1.0)

_build_partitioning_workflow (Line 119)

def _build_partitioning_workflow(self):
    """Build LangGraph workflow for intelligent partitioning."""

Purpose: Build LangGraph workflow for intelligent partitioning.

Parameters: – self: Instance reference

_analyze_data_node (Line 140)

def _analyze_data_node(self, state):
    """Analyze data characteristics."""

Purpose: Analyze data characteristics.

Parameters: – self: Instance reference – state: Current workflow state

_select_strategy_node (Line 153)

def _select_strategy_node(self, state):
    """Select optimal partitioning strategy using LLM."""

Purpose: Select optimal partitioning strategy using LLM.

Parameters: – self: Instance reference – state: Current workflow state

_create_partitions_node (Line 187)

def _create_partitions_node(self, state):
    """Create partitions based on selected strategy."""

Purpose: Create partitions based on selected strategy.

Parameters: – self: Instance reference – state: Current workflow state

_evaluate_balance_node (Line 215)

def _evaluate_balance_node(self, state):
    """Evaluate partition balance."""

Purpose: Evaluate partition balance.

Parameters: – self: Instance reference – state: Current workflow state

_provide_recommendations_node (Line 235)

def _provide_recommendations_node(self, state):
    """Provide recommendations for partitioning."""

Purpose: Provide recommendations for partitioning.

Parameters: – self: Instance reference – state: Current workflow state

partition_by_size (Line 254)

def partition_by_size(self, items, partition_size, prefix):
    """Partition by fixed size."""

Purpose: Partition by fixed size.

Parameters: – self: Instance reference – items: Parameter for items – partition_size: Parameter for partition_size – prefix: Parameter for prefix

partition_by_key (Line 280)

def partition_by_key(self, items, key, prefix):
    """Partition by key value."""

Purpose: Partition by key value.

Parameters: – self: Instance reference – items: Parameter for items – key: Parameter for key – prefix: Parameter for prefix

partition_by_hash (Line 314)

def partition_by_hash(self, items, num_partitions, hash_func):
    """Partition using consistent hashing."""

Purpose: Partition using consistent hashing.

Parameters: – self: Instance reference – items: Parameter for items – num_partitions: Parameter for num_partitions – hash_func: Parameter for hash_func

partition_by_predicate (Line 350)

def partition_by_predicate(self, items, predicates):
    """Partition using custom predicates."""

Purpose: Partition using custom predicates.

Parameters: – self: Instance reference – items: Parameter for items – predicates: Parameter for predicates

intelligent_partition (Line 377)

def intelligent_partition(self, items, target_partitions):
    """Use workflow to intelligently partition data."""

Purpose: Use workflow to intelligently partition data.

Parameters: – self: Instance reference – items: Parameter for items – target_partitions: Parameter for target_partitions

assign_partition (Line 403)

def assign_partition(self, partition_id, agent_id, weight):
    """Assign partition to agent."""

Purpose: Assign partition to agent.

Parameters: – self: Instance reference – partition_id: Parameter for partition_id – agent_id: Parameter for agent_id – weight: Parameter for weight

get_agent_workload (Line 424)

def get_agent_workload(self, agent_id):
    """Get all partitions assigned to agent."""

Purpose: Get all partitions assigned to agent.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id

get_partition_metrics (Line 437)

def get_partition_metrics(self, partition_id):
    """Get metrics for a partition."""

Purpose: Get metrics for a partition.

Parameters: – self: Instance reference – partition_id: Parameter for partition_id

…and 2 more methods

PartitioningPattern (Line 527)

Description: Comprehensive partitioning pattern coordinator with LangChain/LangGraph.

Methods:

__init__ (Line 532)

def __init__(self, model_name):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – model_name: LLM model to use (e.g., “gpt-4”)

partition (Line 535)

def partition(self, items, strategy):
    """Partition items using specified strategy."""

Purpose: Partition items using specified strategy.

Parameters: – self: Instance reference – items: Parameter for items – strategy: Parameter for strategy

smart_partition (Line 554)

def smart_partition(self, items, target_partitions):
    """Use AI to determine optimal partitioning."""

Purpose: Use AI to determine optimal partitioning.

Parameters: – self: Instance reference – items: Parameter for items – target_partitions: Parameter for target_partitions

assign (Line 562)

def assign(self, partition_id, agent_id):
    """Assign partition to agent."""

Purpose: Assign partition to agent.

Parameters: – self: Instance reference – partition_id: Parameter for partition_id – agent_id: Parameter for agent_id

get_workload (Line 566)

def get_workload(self, agent_id):
    """Get agent's workload."""

Purpose: Get agent’s workload.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id

rebalance (Line 570)

def rebalance(self, target_count):
    """Rebalance partitions."""

Purpose: Rebalance partitions.

Parameters: – self: Instance reference – target_count: Parameter for target_count

get_statistics (Line 574)

def get_statistics(self):
    """Get statistics."""

Purpose: Get statistics.

Parameters: – self: Instance reference

Design Patterns Used

• Strategy Pattern: Different execution strategies for various scenarios
• State Pattern: Managing different states in the workflow
• Chain of Responsibility: Sequential processing through workflow nodes
• Observer Pattern: Tracking and monitoring pattern execution
• Factory Pattern: Creating instances with proper configuration

Usage Example

# Initialize the pattern
pattern = PartitionStrategy(name="example", model_name="gpt-4")

# Configure and execute
result = pattern.execute(data={"key": "value"})

# Check results
print(result)

Key Concepts

• LLM Integration: Uses LangChain/OpenAI for intelligent decision-making
• Workflow Orchestration: LangGraph manages execution flow through nodes
• State Management: TypedDict defines workflow state structure
• Error Handling: Comprehensive error handling and fallback mechanisms
• Extensibility: Designed for easy customization and extension

Extension Points

This pattern can be extended by: 1. Adding custom workflow nodes 2. Implementing additional validation logic 3. Integrating with external services 4. Customizing LLM prompts and responses 5. Adding monitoring and logging capabilities

    """
Pattern 169: Sharding

Purpose: Enhanced implementation with full LangChain/LangGraph integration
"""

from typing import Dict, Any, List, Optional
from dataclasses import dataclass, field
from datetime import datetime
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
from typing_extensions import TypedDict


@dataclass
class ShardingConfig:
    """Configuration for Sharding."""
    model_name: str = "gpt-4"
    temperature: float = 0.7
    max_iterations: int = 10
    enable_logging: bool = True


@dataclass
class ShardingState:
    """State information for Sharding."""
    current_step: int = 0
    data: Dict[str, Any] = field(default_factory=dict)
    results: List[Any] = field(default_factory=list)
    metadata: Dict[str, Any] = field(default_factory=dict)
    timestamp: datetime = field(default_factory=datetime.now)


class ShardingAgent:
    """Agent implementing Sharding."""
    
    def __init__(self, name: str, config: ShardingConfig = None):
        self.name = name
        self.config = config or ShardingConfig()
        self.llm = ChatOpenAI(
            model=self.config.model_name,
            temperature=self.config.temperature
        )
        self.state = ShardingState()
        self.execution_history: List[Dict[str, Any]] = []
    
    def process(self, input_data: Any) -> Dict[str, Any]:
        """Process input according to Sharding."""
        self.state.current_step += 1
        
        prompt = ChatPromptTemplate.from_messages([
            ("system", f"You are {self.name}, implementing Sharding. "
                      f"Provide clear, structured responses."),
            ("human", "Process this input: {input_data}")
        ])
        
        chain = prompt | self.llm
        response = chain.invoke({"input_data": str(input_data)})
        
        result = {
            "agent": self.name,
            "step": self.state.current_step,
            "input": input_data,
            "output": str(response.content),
            "timestamp": datetime.now()
        }
        
        self.execution_history.append(result)
        self.state.results.append(result)
        
        return result
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get agent statistics."""
        return {
            "name": self.name,
            "total_executions": len(self.execution_history),
            "current_step": self.state.current_step,
            "results_count": len(self.state.results)
        }


class ShardingPattern:
    """
    Sharding - Enhanced implementation
    
    This pattern provides comprehensive functionality for sharding
    with full LangChain/LangGraph integration.
    
    Features:
    - Multi-agent coordination
    - State management
    - Execution tracking
    - LangGraph workflow integration
    """
    
    def __init__(self, config: ShardingConfig = None):
        self.config = config or ShardingConfig()
        self.agents: Dict[str, ShardingAgent] = {}
        self.workflow = self._build_workflow()
        self.execution_log: List[Dict[str, Any]] = []
    
    def _build_workflow(self) -> StateGraph:
        """Build LangGraph workflow for pattern execution."""
        
        class WorkflowState(TypedDict):
            input: str
            processed: str
            output: str
            metadata: Dict[str, Any]
        
        def process_input(state: WorkflowState) -> WorkflowState:
            """Process input node."""
            return {
                **state,
                "processed": f"Processed: {state['input']}",
                "metadata": {"step": "process_input"}
            }
        
        def generate_output(state: WorkflowState) -> WorkflowState:
            """Generate output node."""
            llm = ChatOpenAI(model=self.config.model_name, temperature=self.config.temperature)
            
            prompt = ChatPromptTemplate.from_messages([
                ("system", "You are implementing Sharding. Generate comprehensive output."),
                ("human", "{input}")
            ])
            
            chain = prompt | llm
            response = chain.invoke({"input": state["processed"]})
            
            return {
                **state,
                "output": str(response.content),
                "metadata": {**state.get("metadata", {}), "step": "generate_output"}
            }
        
        # Build workflow graph
        workflow = StateGraph(WorkflowState)
        workflow.add_node("process", process_input)
        workflow.add_node("generate", generate_output)
        workflow.set_entry_point("process")
        workflow.add_edge("process", "generate")
        workflow.add_edge("generate", END)
        
        return workflow.compile()
    
    def register_agent(self, agent: ShardingAgent):
        """Register an agent with the pattern."""
        self.agents[agent.name] = agent
        if self.config.enable_logging:
            print(f"Registered agent: {agent.name}")
    
    def execute(self, input_data: Any, agent_name: str = None) -> Dict[str, Any]:
        """Execute the pattern with given input."""
        
        # Use workflow if no specific agent requested
        if agent_name is None:
            result = self.workflow.invoke({
                "input": str(input_data),
                "processed": "",
                "output": "",
                "metadata": {}
            })
            
            execution_record = {
                "pattern": "Sharding",
                "input": input_data,
                "output": result.get("output", ""),
                "metadata": result.get("metadata", {}),
                "timestamp": datetime.now()
            }
            
            self.execution_log.append(execution_record)
            return execution_record
        
        # Execute with specific agent
        agent = self.agents.get(agent_name)
        if not agent:
            raise ValueError(f"Agent {agent_name} not found")
        
        result = agent.process(input_data)
        self.execution_log.append(result)
        return result
    
    def run(self, input_data: Any) -> Dict[str, Any]:
        """Main execution method."""
        return self.execute(input_data)
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get pattern statistics."""
        return {
            "pattern": "Sharding",
            "total_executions": len(self.execution_log),
            "registered_agents": len(self.agents),
            "agents": [agent.get_statistics() for agent in self.agents.values()]
        }


# Example usage
if __name__ == "__main__":
    print("=== Sharding Demo ===\n")
    
    # Create configuration
    config = ShardingConfig(
        model_name="gpt-4",
        temperature=0.7,
        enable_logging=True
    )
    
    # Create pattern
    pattern = ShardingPattern(config)
    
    # Create and register agents
    agent1 = ShardingAgent("Agent1", config)
    agent2 = ShardingAgent("Agent2", config)
    pattern.register_agent(agent1)
    pattern.register_agent(agent2)
    
    print("\nExecuting pattern workflow...")
    
    # Execute with workflow
    result1 = pattern.run("Demonstrate sharding functionality")
    print(f"\nWorkflow Result:")
    print(f"  Output: {result1['output'][:150]}...")
    
    # Execute with specific agents
    result2 = pattern.execute("Specific task for Agent1", agent_name="Agent1")
    print(f"\nAgent1 Result:")
    print(f"  Output: {result2['output'][:150]}...")
    
    result3 = pattern.execute("Specific task for Agent2", agent_name="Agent2")
    print(f"\nAgent2 Result:")
    print(f"  Output: {result3['output'][:150]}...")
    
    # Show statistics
    print("\n=== Statistics ===")
    stats = pattern.get_statistics()
    print(f"Pattern: {stats['pattern']}")
    print(f"Total Executions: {stats['total_executions']}")
    print(f"Registered Agents: {stats['registered_agents']}")
    print(f"\nAgent Statistics:")
    for agent_stats in stats['agents']:
        print(f"  {agent_stats['name']}: {agent_stats['total_executions']} executions")

    

Agentic Architecture Pattern: Sharding

Pattern 169: 169 Sharding – Line-by-Line Explanation

Overview

Pattern 169: Sharding

Purpose: Enhanced implementation with full LangChain/LangGraph integration

Code Structure Analysis

File Information

• Pattern Number: 169
• Pattern Name: 169 Sharding
• File Path: patterns\scaling\169_sharding.py
• Total Lines: 247

Detailed Code Explanation

Module Docstring

"""
Pattern 169: Sharding

Purpose: Enhanced implementation with full LangChain/LangGraph integration
"""

Explanation: This module docstring describes the purpose and key features of the 169 Sharding pattern.

Imports Section

from typing import Dict, Any, List, Optional
from dataclasses import dataclass, field
from datetime import datetime
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
from typing_extensions import TypedDict  # First 20 imports shown

Explanation: The module imports necessary dependencies: – from typing import Dict, Any, List, Optional: Type hints for better code documentation – from dataclasses import dataclass, field: Data class decorators for clean data structures – from datetime import datetime: Date and time utilities – from enum import Enum: Enumeration support for fixed sets of values – from langchain_core.prompts import ChatPromptTemplate: LangChain framework integration – from langchain_openai import ChatOpenAI: LangChain framework integration – from langgraph.graph import StateGraph, END: LangGraph for workflow orchestration – from typing_extensions import TypedDict: Type hints for better code documentation

Classes

ShardingConfig (Line 18)

Description: Configuration for Sharding.

ShardingState (Line 27)

Description: State information for Sharding.

ShardingAgent (Line 36)

Description: Agent implementing Sharding.

Methods:

__init__ (Line 39)

def __init__(self, name, config):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – name: Identifier or label – config: Configuration settings

process (Line 49)

def process(self, input_data):
    """Process input according to Sharding."""

Purpose: Process input according to Sharding.

Parameters: – self: Instance reference – input_data: Parameter for input_data

get_statistics (Line 75)

def get_statistics(self):
    """Get agent statistics."""

Purpose: Get agent statistics.

Parameters: – self: Instance reference

ShardingPattern (Line 85)

Description: Sharding – Enhanced implementation

This pattern provides comprehensive functionality for sharding with full LangChain/LangGraph integration.

Features: – Multi-agent coordination – State management – Execution tracking – LangGraph workflow integration

Methods:

__init__ (Line 99)

def __init__(self, config):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – config: Configuration settings

_build_workflow (Line 105)

def _build_workflow(self):
    """Build LangGraph workflow for pattern execution."""

Purpose: Build LangGraph workflow for pattern execution.

Parameters: – self: Instance reference

register_agent (Line 150)

def register_agent(self, agent):
    """Register an agent with the pattern."""

Purpose: Register an agent with the pattern.

Parameters: – self: Instance reference – agent: Parameter for agent

execute (Line 156)

def execute(self, input_data, agent_name):
    """Execute the pattern with given input."""

Purpose: Execute the pattern with given input.

Parameters: – self: Instance reference – input_data: Parameter for input_data – agent_name: Parameter for agent_name

run (Line 188)

def run(self, input_data):
    """Main execution method."""

Purpose: Main execution method.

Parameters: – self: Instance reference – input_data: Parameter for input_data

get_statistics (Line 192)

def get_statistics(self):
    """Get pattern statistics."""

Purpose: Get pattern statistics.

Parameters: – self: Instance reference

WorkflowState (Line 108)

Description: No class docstring.

Design Patterns Used

• Strategy Pattern: Different execution strategies for various scenarios
• State Pattern: Managing different states in the workflow
• Chain of Responsibility: Sequential processing through workflow nodes
• Observer Pattern: Tracking and monitoring pattern execution
• Factory Pattern: Creating instances with proper configuration

Usage Example

# Initialize the pattern
pattern = ShardingConfig(name="example", model_name="gpt-4")

# Configure and execute
result = pattern.execute(data={"key": "value"})

# Check results
print(result)

Key Concepts

• LLM Integration: Uses LangChain/OpenAI for intelligent decision-making
• Workflow Orchestration: LangGraph manages execution flow through nodes
• State Management: TypedDict defines workflow state structure
• Error Handling: Comprehensive error handling and fallback mechanisms
• Extensibility: Designed for easy customization and extension

Extension Points

This pattern can be extended by: 1. Adding custom workflow nodes 2. Implementing additional validation logic 3. Integrating with external services 4. Customizing LLM prompts and responses 5. Adding monitoring and logging capabilities

    """
Pattern 168: Load Distribution

Purpose: Enhanced implementation with full LangChain/LangGraph integration
"""

from typing import Dict, Any, List, Optional
from dataclasses import dataclass, field
from datetime import datetime
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
from typing_extensions import TypedDict


@dataclass
class LoadDistributionConfig:
    """Configuration for Load Distribution."""
    model_name: str = "gpt-4"
    temperature: float = 0.7
    max_iterations: int = 10
    enable_logging: bool = True


@dataclass
class LoadDistributionState:
    """State information for Load Distribution."""
    current_step: int = 0
    data: Dict[str, Any] = field(default_factory=dict)
    results: List[Any] = field(default_factory=list)
    metadata: Dict[str, Any] = field(default_factory=dict)
    timestamp: datetime = field(default_factory=datetime.now)


class LoadDistributionAgent:
    """Agent implementing Load Distribution."""
    
    def __init__(self, name: str, config: LoadDistributionConfig = None):
        self.name = name
        self.config = config or LoadDistributionConfig()
        self.llm = ChatOpenAI(
            model=self.config.model_name,
            temperature=self.config.temperature
        )
        self.state = LoadDistributionState()
        self.execution_history: List[Dict[str, Any]] = []
    
    def process(self, input_data: Any) -> Dict[str, Any]:
        """Process input according to Load Distribution."""
        self.state.current_step += 1
        
        prompt = ChatPromptTemplate.from_messages([
            ("system", f"You are {self.name}, implementing Load Distribution. "
                      f"Provide clear, structured responses."),
            ("human", "Process this input: {input_data}")
        ])
        
        chain = prompt | self.llm
        response = chain.invoke({"input_data": str(input_data)})
        
        result = {
            "agent": self.name,
            "step": self.state.current_step,
            "input": input_data,
            "output": str(response.content),
            "timestamp": datetime.now()
        }
        
        self.execution_history.append(result)
        self.state.results.append(result)
        
        return result
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get agent statistics."""
        return {
            "name": self.name,
            "total_executions": len(self.execution_history),
            "current_step": self.state.current_step,
            "results_count": len(self.state.results)
        }


class LoadDistributionPattern:
    """
    Load Distribution - Enhanced implementation
    
    This pattern provides comprehensive functionality for load distribution
    with full LangChain/LangGraph integration.
    
    Features:
    - Multi-agent coordination
    - State management
    - Execution tracking
    - LangGraph workflow integration
    """
    
    def __init__(self, config: LoadDistributionConfig = None):
        self.config = config or LoadDistributionConfig()
        self.agents: Dict[str, LoadDistributionAgent] = {}
        self.workflow = self._build_workflow()
        self.execution_log: List[Dict[str, Any]] = []
    
    def _build_workflow(self) -> StateGraph:
        """Build LangGraph workflow for pattern execution."""
        
        class WorkflowState(TypedDict):
            input: str
            processed: str
            output: str
            metadata: Dict[str, Any]
        
        def process_input(state: WorkflowState) -> WorkflowState:
            """Process input node."""
            return {
                **state,
                "processed": f"Processed: {state['input']}",
                "metadata": {"step": "process_input"}
            }
        
        def generate_output(state: WorkflowState) -> WorkflowState:
            """Generate output node."""
            llm = ChatOpenAI(model=self.config.model_name, temperature=self.config.temperature)
            
            prompt = ChatPromptTemplate.from_messages([
                ("system", "You are implementing Load Distribution. Generate comprehensive output."),
                ("human", "{input}")
            ])
            
            chain = prompt | llm
            response = chain.invoke({"input": state["processed"]})
            
            return {
                **state,
                "output": str(response.content),
                "metadata": {**state.get("metadata", {}), "step": "generate_output"}
            }
        
        # Build workflow graph
        workflow = StateGraph(WorkflowState)
        workflow.add_node("process", process_input)
        workflow.add_node("generate", generate_output)
        workflow.set_entry_point("process")
        workflow.add_edge("process", "generate")
        workflow.add_edge("generate", END)
        
        return workflow.compile()
    
    def register_agent(self, agent: LoadDistributionAgent):
        """Register an agent with the pattern."""
        self.agents[agent.name] = agent
        if self.config.enable_logging:
            print(f"Registered agent: {agent.name}")
    
    def execute(self, input_data: Any, agent_name: str = None) -> Dict[str, Any]:
        """Execute the pattern with given input."""
        
        # Use workflow if no specific agent requested
        if agent_name is None:
            result = self.workflow.invoke({
                "input": str(input_data),
                "processed": "",
                "output": "",
                "metadata": {}
            })
            
            execution_record = {
                "pattern": "Load Distribution",
                "input": input_data,
                "output": result.get("output", ""),
                "metadata": result.get("metadata", {}),
                "timestamp": datetime.now()
            }
            
            self.execution_log.append(execution_record)
            return execution_record
        
        # Execute with specific agent
        agent = self.agents.get(agent_name)
        if not agent:
            raise ValueError(f"Agent {agent_name} not found")
        
        result = agent.process(input_data)
        self.execution_log.append(result)
        return result
    
    def run(self, input_data: Any) -> Dict[str, Any]:
        """Main execution method."""
        return self.execute(input_data)
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get pattern statistics."""
        return {
            "pattern": "Load Distribution",
            "total_executions": len(self.execution_log),
            "registered_agents": len(self.agents),
            "agents": [agent.get_statistics() for agent in self.agents.values()]
        }


# Example usage
if __name__ == "__main__":
    print("=== Load Distribution Demo ===\n")
    
    # Create configuration
    config = LoadDistributionConfig(
        model_name="gpt-4",
        temperature=0.7,
        enable_logging=True
    )
    
    # Create pattern
    pattern = LoadDistributionPattern(config)
    
    # Create and register agents
    agent1 = LoadDistributionAgent("Agent1", config)
    agent2 = LoadDistributionAgent("Agent2", config)
    pattern.register_agent(agent1)
    pattern.register_agent(agent2)
    
    print("\nExecuting pattern workflow...")
    
    # Execute with workflow
    result1 = pattern.run("Demonstrate load distribution functionality")
    print(f"\nWorkflow Result:")
    print(f"  Output: {result1['output'][:150]}...")
    
    # Execute with specific agents
    result2 = pattern.execute("Specific task for Agent1", agent_name="Agent1")
    print(f"\nAgent1 Result:")
    print(f"  Output: {result2['output'][:150]}...")
    
    result3 = pattern.execute("Specific task for Agent2", agent_name="Agent2")
    print(f"\nAgent2 Result:")
    print(f"  Output: {result3['output'][:150]}...")
    
    # Show statistics
    print("\n=== Statistics ===")
    stats = pattern.get_statistics()
    print(f"Pattern: {stats['pattern']}")
    print(f"Total Executions: {stats['total_executions']}")
    print(f"Registered Agents: {stats['registered_agents']}")
    print(f"\nAgent Statistics:")
    for agent_stats in stats['agents']:
        print(f"  {agent_stats['name']}: {agent_stats['total_executions']} executions")

    

Agentic Architecture Pattern: Load Distribution

Pattern 168: 168 Load Distribution – Line-by-Line Explanation

Overview

Pattern 168: Load Distribution

Purpose: Enhanced implementation with full LangChain/LangGraph integration

Code Structure Analysis

File Information

• Pattern Number: 168
• Pattern Name: 168 Load Distribution
• File Path: patterns\scaling\168_load_distribution.py
• Total Lines: 247

Detailed Code Explanation

Module Docstring

"""
Pattern 168: Load Distribution

Purpose: Enhanced implementation with full LangChain/LangGraph integration
"""

Explanation: This module docstring describes the purpose and key features of the 168 Load Distribution pattern.

Imports Section

from typing import Dict, Any, List, Optional
from dataclasses import dataclass, field
from datetime import datetime
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
from typing_extensions import TypedDict  # First 20 imports shown

Explanation: The module imports necessary dependencies: – from typing import Dict, Any, List, Optional: Type hints for better code documentation – from dataclasses import dataclass, field: Data class decorators for clean data structures – from datetime import datetime: Date and time utilities – from enum import Enum: Enumeration support for fixed sets of values – from langchain_core.prompts import ChatPromptTemplate: LangChain framework integration – from langchain_openai import ChatOpenAI: LangChain framework integration – from langgraph.graph import StateGraph, END: LangGraph for workflow orchestration – from typing_extensions import TypedDict: Type hints for better code documentation

Classes

LoadDistributionConfig (Line 18)

Description: Configuration for Load Distribution.

LoadDistributionState (Line 27)

Description: State information for Load Distribution.

LoadDistributionAgent (Line 36)

Description: Agent implementing Load Distribution.

Methods:

__init__ (Line 39)

def __init__(self, name, config):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – name: Identifier or label – config: Configuration settings

process (Line 49)

def process(self, input_data):
    """Process input according to Load Distribution."""

Purpose: Process input according to Load Distribution.

Parameters: – self: Instance reference – input_data: Parameter for input_data

get_statistics (Line 75)

def get_statistics(self):
    """Get agent statistics."""

Purpose: Get agent statistics.

Parameters: – self: Instance reference

LoadDistributionPattern (Line 85)

Description: Load Distribution – Enhanced implementation

This pattern provides comprehensive functionality for load distribution with full LangChain/LangGraph integration.

Features: – Multi-agent coordination – State management – Execution tracking – LangGraph workflow integration

Methods:

__init__ (Line 99)

def __init__(self, config):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – config: Configuration settings

_build_workflow (Line 105)

def _build_workflow(self):
    """Build LangGraph workflow for pattern execution."""

Purpose: Build LangGraph workflow for pattern execution.

Parameters: – self: Instance reference

register_agent (Line 150)

def register_agent(self, agent):
    """Register an agent with the pattern."""

Purpose: Register an agent with the pattern.

Parameters: – self: Instance reference – agent: Parameter for agent

execute (Line 156)

def execute(self, input_data, agent_name):
    """Execute the pattern with given input."""

Purpose: Execute the pattern with given input.

Parameters: – self: Instance reference – input_data: Parameter for input_data – agent_name: Parameter for agent_name

run (Line 188)

def run(self, input_data):
    """Main execution method."""

Purpose: Main execution method.

Parameters: – self: Instance reference – input_data: Parameter for input_data

get_statistics (Line 192)

def get_statistics(self):
    """Get pattern statistics."""

Purpose: Get pattern statistics.

Parameters: – self: Instance reference

WorkflowState (Line 108)

Description: No class docstring.

Design Patterns Used

• Strategy Pattern: Different execution strategies for various scenarios
• State Pattern: Managing different states in the workflow
• Chain of Responsibility: Sequential processing through workflow nodes
• Observer Pattern: Tracking and monitoring pattern execution
• Factory Pattern: Creating instances with proper configuration

Usage Example

# Initialize the pattern
pattern = LoadDistributionConfig(name="example", model_name="gpt-4")

# Configure and execute
result = pattern.execute(data={"key": "value"})

# Check results
print(result)

Key Concepts

• LLM Integration: Uses LangChain/OpenAI for intelligent decision-making
• Workflow Orchestration: LangGraph manages execution flow through nodes
• State Management: TypedDict defines workflow state structure
• Error Handling: Comprehensive error handling and fallback mechanisms
• Extensibility: Designed for easy customization and extension

Extension Points

This pattern can be extended by: 1. Adding custom workflow nodes 2. Implementing additional validation logic 3. Integrating with external services 4. Customizing LLM prompts and responses 5. Adding monitoring and logging capabilities

    """
Pattern 167: Resource Allocation

Purpose: Enhanced implementation with full LangChain/LangGraph integration
"""

from typing import Dict, Any, List, Optional
from dataclasses import dataclass, field
from datetime import datetime
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
from typing_extensions import TypedDict


@dataclass
class ResourceAllocationConfig:
    """Configuration for Resource Allocation."""
    model_name: str = "gpt-4"
    temperature: float = 0.7
    max_iterations: int = 10
    enable_logging: bool = True


@dataclass
class ResourceAllocationState:
    """State information for Resource Allocation."""
    current_step: int = 0
    data: Dict[str, Any] = field(default_factory=dict)
    results: List[Any] = field(default_factory=list)
    metadata: Dict[str, Any] = field(default_factory=dict)
    timestamp: datetime = field(default_factory=datetime.now)


class ResourceAllocationAgent:
    """Agent implementing Resource Allocation."""
    
    def __init__(self, name: str, config: ResourceAllocationConfig = None):
        self.name = name
        self.config = config or ResourceAllocationConfig()
        self.llm = ChatOpenAI(
            model=self.config.model_name,
            temperature=self.config.temperature
        )
        self.state = ResourceAllocationState()
        self.execution_history: List[Dict[str, Any]] = []
    
    def process(self, input_data: Any) -> Dict[str, Any]:
        """Process input according to Resource Allocation."""
        self.state.current_step += 1
        
        prompt = ChatPromptTemplate.from_messages([
            ("system", f"You are {self.name}, implementing Resource Allocation. "
                      f"Provide clear, structured responses."),
            ("human", "Process this input: {input_data}")
        ])
        
        chain = prompt | self.llm
        response = chain.invoke({"input_data": str(input_data)})
        
        result = {
            "agent": self.name,
            "step": self.state.current_step,
            "input": input_data,
            "output": str(response.content),
            "timestamp": datetime.now()
        }
        
        self.execution_history.append(result)
        self.state.results.append(result)
        
        return result
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get agent statistics."""
        return {
            "name": self.name,
            "total_executions": len(self.execution_history),
            "current_step": self.state.current_step,
            "results_count": len(self.state.results)
        }


class ResourceAllocationPattern:
    """
    Resource Allocation - Enhanced implementation
    
    This pattern provides comprehensive functionality for resource allocation
    with full LangChain/LangGraph integration.
    
    Features:
    - Multi-agent coordination
    - State management
    - Execution tracking
    - LangGraph workflow integration
    """
    
    def __init__(self, config: ResourceAllocationConfig = None):
        self.config = config or ResourceAllocationConfig()
        self.agents: Dict[str, ResourceAllocationAgent] = {}
        self.workflow = self._build_workflow()
        self.execution_log: List[Dict[str, Any]] = []
    
    def _build_workflow(self) -> StateGraph:
        """Build LangGraph workflow for pattern execution."""
        
        class WorkflowState(TypedDict):
            input: str
            processed: str
            output: str
            metadata: Dict[str, Any]
        
        def process_input(state: WorkflowState) -> WorkflowState:
            """Process input node."""
            return {
                **state,
                "processed": f"Processed: {state['input']}",
                "metadata": {"step": "process_input"}
            }
        
        def generate_output(state: WorkflowState) -> WorkflowState:
            """Generate output node."""
            llm = ChatOpenAI(model=self.config.model_name, temperature=self.config.temperature)
            
            prompt = ChatPromptTemplate.from_messages([
                ("system", "You are implementing Resource Allocation. Generate comprehensive output."),
                ("human", "{input}")
            ])
            
            chain = prompt | llm
            response = chain.invoke({"input": state["processed"]})
            
            return {
                **state,
                "output": str(response.content),
                "metadata": {**state.get("metadata", {}), "step": "generate_output"}
            }
        
        # Build workflow graph
        workflow = StateGraph(WorkflowState)
        workflow.add_node("process", process_input)
        workflow.add_node("generate", generate_output)
        workflow.set_entry_point("process")
        workflow.add_edge("process", "generate")
        workflow.add_edge("generate", END)
        
        return workflow.compile()
    
    def register_agent(self, agent: ResourceAllocationAgent):
        """Register an agent with the pattern."""
        self.agents[agent.name] = agent
        if self.config.enable_logging:
            print(f"Registered agent: {agent.name}")
    
    def execute(self, input_data: Any, agent_name: str = None) -> Dict[str, Any]:
        """Execute the pattern with given input."""
        
        # Use workflow if no specific agent requested
        if agent_name is None:
            result = self.workflow.invoke({
                "input": str(input_data),
                "processed": "",
                "output": "",
                "metadata": {}
            })
            
            execution_record = {
                "pattern": "Resource Allocation",
                "input": input_data,
                "output": result.get("output", ""),
                "metadata": result.get("metadata", {}),
                "timestamp": datetime.now()
            }
            
            self.execution_log.append(execution_record)
            return execution_record
        
        # Execute with specific agent
        agent = self.agents.get(agent_name)
        if not agent:
            raise ValueError(f"Agent {agent_name} not found")
        
        result = agent.process(input_data)
        self.execution_log.append(result)
        return result
    
    def run(self, input_data: Any) -> Dict[str, Any]:
        """Main execution method."""
        return self.execute(input_data)
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get pattern statistics."""
        return {
            "pattern": "Resource Allocation",
            "total_executions": len(self.execution_log),
            "registered_agents": len(self.agents),
            "agents": [agent.get_statistics() for agent in self.agents.values()]
        }


# Example usage
if __name__ == "__main__":
    print("=== Resource Allocation Demo ===\n")
    
    # Create configuration
    config = ResourceAllocationConfig(
        model_name="gpt-4",
        temperature=0.7,
        enable_logging=True
    )
    
    # Create pattern
    pattern = ResourceAllocationPattern(config)
    
    # Create and register agents
    agent1 = ResourceAllocationAgent("Agent1", config)
    agent2 = ResourceAllocationAgent("Agent2", config)
    pattern.register_agent(agent1)
    pattern.register_agent(agent2)
    
    print("\nExecuting pattern workflow...")
    
    # Execute with workflow
    result1 = pattern.run("Demonstrate resource allocation functionality")
    print(f"\nWorkflow Result:")
    print(f"  Output: {result1['output'][:150]}...")
    
    # Execute with specific agents
    result2 = pattern.execute("Specific task for Agent1", agent_name="Agent1")
    print(f"\nAgent1 Result:")
    print(f"  Output: {result2['output'][:150]}...")
    
    result3 = pattern.execute("Specific task for Agent2", agent_name="Agent2")
    print(f"\nAgent2 Result:")
    print(f"  Output: {result3['output'][:150]}...")
    
    # Show statistics
    print("\n=== Statistics ===")
    stats = pattern.get_statistics()
    print(f"Pattern: {stats['pattern']}")
    print(f"Total Executions: {stats['total_executions']}")
    print(f"Registered Agents: {stats['registered_agents']}")
    print(f"\nAgent Statistics:")
    for agent_stats in stats['agents']:
        print(f"  {agent_stats['name']}: {agent_stats['total_executions']} executions")

    

Agentic Architecture Pattern: Resource Allocation

Pattern 167: 167 Resource Allocation – Line-by-Line Explanation

Overview

Pattern 167: Resource Allocation

Purpose: Enhanced implementation with full LangChain/LangGraph integration

Code Structure Analysis

File Information

• Pattern Number: 167
• Pattern Name: 167 Resource Allocation
• File Path: patterns\scaling\167_resource_allocation.py
• Total Lines: 247

Detailed Code Explanation

Module Docstring

"""
Pattern 167: Resource Allocation

Purpose: Enhanced implementation with full LangChain/LangGraph integration
"""

Explanation: This module docstring describes the purpose and key features of the 167 Resource Allocation pattern.

Imports Section

from typing import Dict, Any, List, Optional
from dataclasses import dataclass, field
from datetime import datetime
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
from typing_extensions import TypedDict  # First 20 imports shown

Explanation: The module imports necessary dependencies: – from typing import Dict, Any, List, Optional: Type hints for better code documentation – from dataclasses import dataclass, field: Data class decorators for clean data structures – from datetime import datetime: Date and time utilities – from enum import Enum: Enumeration support for fixed sets of values – from langchain_core.prompts import ChatPromptTemplate: LangChain framework integration – from langchain_openai import ChatOpenAI: LangChain framework integration – from langgraph.graph import StateGraph, END: LangGraph for workflow orchestration – from typing_extensions import TypedDict: Type hints for better code documentation

Classes

ResourceAllocationConfig (Line 18)

Description: Configuration for Resource Allocation.

ResourceAllocationState (Line 27)

Description: State information for Resource Allocation.

ResourceAllocationAgent (Line 36)

Description: Agent implementing Resource Allocation.

Methods:

__init__ (Line 39)

def __init__(self, name, config):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – name: Identifier or label – config: Configuration settings

process (Line 49)

def process(self, input_data):
    """Process input according to Resource Allocation."""

Purpose: Process input according to Resource Allocation.

Parameters: – self: Instance reference – input_data: Parameter for input_data

get_statistics (Line 75)

def get_statistics(self):
    """Get agent statistics."""

Purpose: Get agent statistics.

Parameters: – self: Instance reference

ResourceAllocationPattern (Line 85)

Description: Resource Allocation – Enhanced implementation

This pattern provides comprehensive functionality for resource allocation with full LangChain/LangGraph integration.

Features: – Multi-agent coordination – State management – Execution tracking – LangGraph workflow integration

Methods:

__init__ (Line 99)

def __init__(self, config):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – config: Configuration settings

_build_workflow (Line 105)

def _build_workflow(self):
    """Build LangGraph workflow for pattern execution."""

Purpose: Build LangGraph workflow for pattern execution.

Parameters: – self: Instance reference

register_agent (Line 150)

def register_agent(self, agent):
    """Register an agent with the pattern."""

Purpose: Register an agent with the pattern.

Parameters: – self: Instance reference – agent: Parameter for agent

execute (Line 156)

def execute(self, input_data, agent_name):
    """Execute the pattern with given input."""

Purpose: Execute the pattern with given input.

Parameters: – self: Instance reference – input_data: Parameter for input_data – agent_name: Parameter for agent_name

run (Line 188)

def run(self, input_data):
    """Main execution method."""

Purpose: Main execution method.

Parameters: – self: Instance reference – input_data: Parameter for input_data

get_statistics (Line 192)

def get_statistics(self):
    """Get pattern statistics."""

Purpose: Get pattern statistics.

Parameters: – self: Instance reference

WorkflowState (Line 108)

Description: No class docstring.

Design Patterns Used

• Strategy Pattern: Different execution strategies for various scenarios
• State Pattern: Managing different states in the workflow
• Chain of Responsibility: Sequential processing through workflow nodes
• Observer Pattern: Tracking and monitoring pattern execution
• Factory Pattern: Creating instances with proper configuration

Usage Example

# Initialize the pattern
pattern = ResourceAllocationConfig(name="example", model_name="gpt-4")

# Configure and execute
result = pattern.execute(data={"key": "value"})

# Check results
print(result)

Key Concepts

• LLM Integration: Uses LangChain/OpenAI for intelligent decision-making
• Workflow Orchestration: LangGraph manages execution flow through nodes
• State Management: TypedDict defines workflow state structure
• Error Handling: Comprehensive error handling and fallback mechanisms
• Extensibility: Designed for easy customization and extension

Extension Points

This pattern can be extended by: 1. Adding custom workflow nodes 2. Implementing additional validation logic 3. Integrating with external services 4. Customizing LLM prompts and responses 5. Adding monitoring and logging capabilities

    """
Pattern 166: Agent Lifecycle Management Pattern with LangChain/LangGraph

Purpose: Manage complete lifecycle of agents from creation to termination with LLM insights.
States: Created → Initialized → Active → Suspended → Terminated
"""

from typing import Dict, Any, List, Optional, TypedDict
from dataclasses import dataclass, field
from datetime import datetime, timedelta
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
import statistics


class LifecycleState(Enum):
    """Agent lifecycle states."""
    CREATED = "created"
    INITIALIZED = "initialized"
    ACTIVE = "active"
    SUSPENDED = "suspended"
    TERMINATED = "terminated"


class TransitionTrigger(Enum):
    """Lifecycle transition triggers."""
    MANUAL = "manual"
    AUTOMATIC = "automatic"
    SCHEDULED = "scheduled"
    ERROR = "error"
    RESOURCE_LIMIT = "resource_limit"


class HealthStatus(Enum):
    """Agent health status."""
    HEALTHY = "healthy"
    DEGRADED = "degraded"
    UNHEALTHY = "unhealthy"
    UNKNOWN = "unknown"


class ResourceType(Enum):
    """Resource types for agents."""
    CPU = "cpu"
    MEMORY = "memory"
    NETWORK = "network"
    STORAGE = "storage"


@dataclass
class LifecycleEvent:
    """Event in agent lifecycle."""
    event_id: str
    from_state: LifecycleState
    to_state: LifecycleState
    trigger: TransitionTrigger
    reason: str
    metadata: Dict[str, Any] = field(default_factory=dict)
    timestamp: datetime = field(default_factory=datetime.now)


@dataclass
class ResourceAllocation:
    """Resource allocation for an agent."""
    cpu_cores: float
    memory_mb: int
    network_bandwidth_mbps: int
    storage_gb: int
    allocated_at: datetime = field(default_factory=datetime.now)


@dataclass
class AgentHealth:
    """Agent health metrics."""
    status: HealthStatus
    cpu_usage: float
    memory_usage: float
    error_count: int
    last_active: datetime
    uptime_seconds: float


@dataclass
class ManagedAgent:
    """Agent with comprehensive lifecycle management."""
    id: str
    name: str
    agent_type: str
    state: LifecycleState = LifecycleState.CREATED
    created_at: datetime = field(default_factory=datetime.now)
    events: List[LifecycleEvent] = field(default_factory=list)
    resources: Optional[ResourceAllocation] = None
    health: Optional[AgentHealth] = None
    configuration: Dict[str, Any] = field(default_factory=dict)
    tags: Dict[str, str] = field(default_factory=dict)


@dataclass
class LifecyclePolicy:
    """Policy for lifecycle management."""
    policy_id: str
    auto_suspend_after_idle_seconds: Optional[int] = None
    auto_terminate_after_suspended_seconds: Optional[int] = None
    max_error_count: Optional[int] = None
    health_check_interval_seconds: int = 60


class LifecycleManagementState(TypedDict):
    """State for lifecycle management workflow."""
    agent_id: str
    current_state: str
    target_state: str
    validation_result: bool
    health_check: Dict[str, Any]
    recommendations: List[str]
    transition_approved: bool


class LifecycleManager:
    """
    Advanced lifecycle manager with LangChain/LangGraph integration.
    """
    
    def __init__(self, model_name: str = "gpt-4", temperature: float = 0.3):
        self.llm = ChatOpenAI(model=model_name, temperature=temperature)
        self.agents: Dict[str, ManagedAgent] = {}
        self.agent_llms: Dict[str, ChatOpenAI] = {}
        self.policies: Dict[str, LifecyclePolicy] = {}
        self.workflow = self._build_lifecycle_workflow()
        
        # Statistics
        self.stats = {
            "total_agents_created": 0,
            "total_state_transitions": 0,
            "agents_by_state": {state: 0 for state in LifecycleState},
            "transitions_by_trigger": {trigger: 0 for trigger in TransitionTrigger},
            "avg_agent_lifetime_seconds": 0.0,
            "health_checks_performed": 0
        }
    
    def _build_lifecycle_workflow(self) -> StateGraph:
        """Build LangGraph workflow for lifecycle management."""
        workflow = StateGraph(LifecycleManagementState)
        
        # Add nodes
        workflow.add_node("validate_transition", self._validate_transition_node)
        workflow.add_node("check_health", self._check_health_node)
        workflow.add_node("analyze_impact", self._analyze_impact_node)
        workflow.add_node("approve_transition", self._approve_transition_node)
        workflow.add_node("provide_recommendations", self._provide_recommendations_node)
        
        # Define edges
        workflow.set_entry_point("validate_transition")
        workflow.add_edge("validate_transition", "check_health")
        workflow.add_edge("check_health", "analyze_impact")
        workflow.add_edge("analyze_impact", "approve_transition")
        workflow.add_edge("approve_transition", "provide_recommendations")
        workflow.add_edge("provide_recommendations", END)
        
        return workflow.compile()
    
    def _validate_transition_node(self, state: LifecycleManagementState) -> LifecycleManagementState:
        """Validate if state transition is allowed."""
        agent_id = state["agent_id"]
        current = state["current_state"]
        target = state["target_state"]
        
        # Valid transitions
        valid_transitions = {
            LifecycleState.CREATED.value: [LifecycleState.INITIALIZED.value, LifecycleState.TERMINATED.value],
            LifecycleState.INITIALIZED.value: [LifecycleState.ACTIVE.value, LifecycleState.TERMINATED.value],
            LifecycleState.ACTIVE.value: [LifecycleState.SUSPENDED.value, LifecycleState.TERMINATED.value],
            LifecycleState.SUSPENDED.value: [LifecycleState.ACTIVE.value, LifecycleState.TERMINATED.value],
            LifecycleState.TERMINATED.value: []
        }
        
        is_valid = target in valid_transitions.get(current, [])
        state["validation_result"] = is_valid
        
        return state
    
    def _check_health_node(self, state: LifecycleManagementState) -> LifecycleManagementState:
        """Check agent health."""
        agent_id = state["agent_id"]
        agent = self.agents.get(agent_id)
        
        health_data = {}
        if agent and agent.health:
            health_data = {
                "status": agent.health.status.value,
                "cpu_usage": agent.health.cpu_usage,
                "memory_usage": agent.health.memory_usage,
                "error_count": agent.health.error_count,
                "uptime_seconds": agent.health.uptime_seconds
            }
        else:
            health_data = {
                "status": HealthStatus.UNKNOWN.value,
                "message": "Health data not available"
            }
        
        state["health_check"] = health_data
        self.stats["health_checks_performed"] += 1
        
        return state
    
    def _analyze_impact_node(self, state: LifecycleManagementState) -> LifecycleManagementState:
        """Analyze impact of state transition using LLM."""
        current = state["current_state"]
        target = state["target_state"]
        health = state["health_check"]
        
        # Use LLM to analyze transition impact
        prompt = ChatPromptTemplate.from_template(
            "Analyze the impact of this agent lifecycle transition:\n"
            "Current State: {current_state}\n"
            "Target State: {target_state}\n"
            "Agent Health: {health_data}\n\n"
            "Provide a brief analysis of potential impacts and risks."
        )
        
        chain = prompt | self.llm
        response = chain.invoke({
            "current_state": current,
            "target_state": target,
            "health_data": str(health)
        })
        
        # Store analysis in recommendations for now
        state["recommendations"] = [str(response.content)[:200]]
        
        return state
    
    def _approve_transition_node(self, state: LifecycleManagementState) -> LifecycleManagementState:
        """Approve or reject transition based on validation and health."""
        validation = state["validation_result"]
        health = state["health_check"]
        
        # Approve if validation passes and agent is not critically unhealthy
        approved = validation
        if health.get("status") == HealthStatus.UNHEALTHY.value:
            # Still allow termination even if unhealthy
            if state["target_state"] != LifecycleState.TERMINATED.value:
                approved = False
        
        state["transition_approved"] = approved
        
        return state
    
    def _provide_recommendations_node(self, state: LifecycleManagementState) -> LifecycleManagementState:
        """Provide recommendations for lifecycle management."""
        approved = state["transition_approved"]
        health = state["health_check"]
        
        recommendations = []
        
        if not approved:
            recommendations.append("Transition not approved - check validation rules and agent health")
        
        if health.get("error_count", 0) > 5:
            recommendations.append("High error count detected - consider investigating before proceeding")
        
        if health.get("cpu_usage", 0) > 80:
            recommendations.append("High CPU usage - may need resource scaling")
        
        if health.get("memory_usage", 0) > 80:
            recommendations.append("High memory usage - check for memory leaks")
        
        if recommendations:
            state["recommendations"] = recommendations
        
        return state
    
    def create_agent(
        self,
        agent_id: str,
        name: str,
        agent_type: str = "general",
        configuration: Optional[Dict[str, Any]] = None,
        tags: Optional[Dict[str, str]] = None
    ) -> ManagedAgent:
        """Create a new managed agent."""
        if agent_id in self.agents:
            raise ValueError(f"Agent {agent_id} already exists")
        
        agent = ManagedAgent(
            id=agent_id,
            name=name,
            agent_type=agent_type,
            configuration=configuration or {},
            tags=tags or {}
        )
        
        self.agents[agent_id] = agent
        self.stats["total_agents_created"] += 1
        self.stats["agents_by_state"][LifecycleState.CREATED] += 1
        
        # Record event
        self._record_event(
            agent_id,
            LifecycleState.CREATED,
            LifecycleState.CREATED,
            TransitionTrigger.MANUAL,
            "Agent created"
        )
        
        return agent
    
    def initialize_agent(
        self,
        agent_id: str,
        model: str = "gpt-4",
        resources: Optional[ResourceAllocation] = None
    ) -> bool:
        """Initialize agent with resources."""
        result = self._execute_transition(
            agent_id,
            LifecycleState.INITIALIZED,
            TransitionTrigger.MANUAL,
            "Initialization requested"
        )
        
        if result:
            # Allocate resources
            if resources:
                self.agents[agent_id].resources = resources
            else:
                self.agents[agent_id].resources = ResourceAllocation(
                    cpu_cores=1.0,
                    memory_mb=512,
                    network_bandwidth_mbps=100,
                    storage_gb=10
                )
            
            # Initialize LLM
            self.agent_llms[agent_id] = ChatOpenAI(model=model, temperature=0.7)
            
            # Initialize health
            self.agents[agent_id].health = AgentHealth(
                status=HealthStatus.HEALTHY,
                cpu_usage=0.0,
                memory_usage=0.0,
                error_count=0,
                last_active=datetime.now(),
                uptime_seconds=0.0
            )
        
        return result
    
    def activate_agent(self, agent_id: str) -> bool:
        """Activate agent for work."""
        return self._execute_transition(
            agent_id,
            LifecycleState.ACTIVE,
            TransitionTrigger.MANUAL,
            "Activation requested"
        )
    
    def suspend_agent(
        self,
        agent_id: str,
        reason: str = "Manual suspension",
        trigger: TransitionTrigger = TransitionTrigger.MANUAL
    ) -> bool:
        """Suspend active agent."""
        return self._execute_transition(
            agent_id,
            LifecycleState.SUSPENDED,
            trigger,
            reason
        )
    
    def resume_agent(self, agent_id: str) -> bool:
        """Resume suspended agent."""
        return self._execute_transition(
            agent_id,
            LifecycleState.ACTIVE,
            TransitionTrigger.MANUAL,
            "Resume requested"
        )
    
    def terminate_agent(
        self,
        agent_id: str,
        reason: str = "Normal termination",
        trigger: TransitionTrigger = TransitionTrigger.MANUAL
    ) -> bool:
        """Terminate agent and cleanup resources."""
        result = self._execute_transition(
            agent_id,
            LifecycleState.TERMINATED,
            trigger,
            reason
        )
        
        if result:
            # Cleanup resources
            if agent_id in self.agent_llms:
                del self.agent_llms[agent_id]
            
            # Clear resource allocation
            if self.agents[agent_id].resources:
                self.agents[agent_id].resources = None
        
        return result
    
    def _execute_transition(
        self,
        agent_id: str,
        target_state: LifecycleState,
        trigger: TransitionTrigger,
        reason: str
    ) -> bool:
        """Execute state transition with workflow validation."""
        agent = self.agents.get(agent_id)
        if not agent:
            return False
        
        # Prepare workflow state
        initial_state: LifecycleManagementState = {
            "agent_id": agent_id,
            "current_state": agent.state.value,
            "target_state": target_state.value,
            "validation_result": False,
            "health_check": {},
            "recommendations": [],
            "transition_approved": False
        }
        
        # Run workflow
        result = self.workflow.invoke(initial_state)
        
        # Execute transition if approved
        if result["transition_approved"]:
            self._record_event(
                agent_id,
                agent.state,
                target_state,
                trigger,
                reason
            )
            
            # Update state counts
            self.stats["agents_by_state"][agent.state] -= 1
            agent.state = target_state
            self.stats["agents_by_state"][target_state] += 1
            self.stats["total_state_transitions"] += 1
            self.stats["transitions_by_trigger"][trigger] += 1
            
            return True
        
        return False
    
    def _record_event(
        self,
        agent_id: str,
        from_state: LifecycleState,
        to_state: LifecycleState,
        trigger: TransitionTrigger,
        reason: str,
        metadata: Optional[Dict[str, Any]] = None
    ):
        """Record a lifecycle event."""
        agent = self.agents[agent_id]
        event = LifecycleEvent(
            event_id=f"{agent_id}_event_{len(agent.events)}",
            from_state=from_state,
            to_state=to_state,
            trigger=trigger,
            reason=reason,
            metadata=metadata or {}
        )
        agent.events.append(event)
    
    def update_agent_health(
        self,
        agent_id: str,
        cpu_usage: float,
        memory_usage: float,
        error_count: int = 0
    ):
        """Update agent health metrics."""
        agent = self.agents.get(agent_id)
        if not agent or not agent.health:
            return
        
        agent.health.cpu_usage = cpu_usage
        agent.health.memory_usage = memory_usage
        agent.health.error_count = error_count
        agent.health.last_active = datetime.now()
        
        # Calculate uptime
        agent.health.uptime_seconds = (datetime.now() - agent.created_at).total_seconds()
        
        # Determine health status
        if error_count > 10 or cpu_usage > 95 or memory_usage > 95:
            agent.health.status = HealthStatus.UNHEALTHY
        elif error_count > 5 or cpu_usage > 80 or memory_usage > 80:
            agent.health.status = HealthStatus.DEGRADED
        else:
            agent.health.status = HealthStatus.HEALTHY
    
    def apply_policy(self, agent_id: str, policy: LifecyclePolicy):
        """Apply lifecycle policy to an agent."""
        self.policies[agent_id] = policy
    
    def check_policies(self):
        """Check and enforce lifecycle policies."""
        now = datetime.now()
        
        for agent_id, policy in self.policies.items():
            agent = self.agents.get(agent_id)
            if not agent or not agent.health:
                continue
            
            # Auto-suspend after idle
            if policy.auto_suspend_after_idle_seconds:
                idle_time = (now - agent.health.last_active).total_seconds()
                if (idle_time > policy.auto_suspend_after_idle_seconds and 
                    agent.state == LifecycleState.ACTIVE):
                    self.suspend_agent(
                        agent_id,
                        f"Auto-suspended after {idle_time:.0f}s idle",
                        TransitionTrigger.AUTOMATIC
                    )
            
            # Auto-terminate after suspended
            if policy.auto_terminate_after_suspended_seconds:
                if agent.state == LifecycleState.SUSPENDED:
                    suspended_time = (now - agent.events[-1].timestamp).total_seconds()
                    if suspended_time > policy.auto_terminate_after_suspended_seconds:
                        self.terminate_agent(
                            agent_id,
                            f"Auto-terminated after {suspended_time:.0f}s suspended",
                            TransitionTrigger.AUTOMATIC
                        )
            
            # Terminate on error threshold
            if policy.max_error_count:
                if agent.health.error_count > policy.max_error_count:
                    self.terminate_agent(
                        agent_id,
                        f"Error count exceeded: {agent.health.error_count}",
                        TransitionTrigger.ERROR
                    )
    
    def get_agent_history(self, agent_id: str) -> List[Dict[str, Any]]:
        """Get lifecycle history of agent."""
        agent = self.agents.get(agent_id)
        if not agent:
            return []
        
        return [
            {
                "event_id": e.event_id,
                "from_state": e.from_state.value,
                "to_state": e.to_state.value,
                "trigger": e.trigger.value,
                "reason": e.reason,
                "timestamp": e.timestamp.isoformat()
            }
            for e in agent.events
        ]
    
    def get_agents_by_state(self, state: LifecycleState) -> List[ManagedAgent]:
        """Get all agents in a specific state."""
        return [agent for agent in self.agents.values() if agent.state == state]
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get comprehensive lifecycle statistics."""
        # Calculate average lifetime for terminated agents
        terminated_agents = self.get_agents_by_state(LifecycleState.TERMINATED)
        if terminated_agents:
            lifetimes = [
                (agent.events[-1].timestamp - agent.created_at).total_seconds()
                for agent in terminated_agents
                if agent.events
            ]
            self.stats["avg_agent_lifetime_seconds"] = statistics.mean(lifetimes) if lifetimes else 0.0
        
        return {
            "total_agents_created": self.stats["total_agents_created"],
            "total_active_agents": len(self.agents),
            "total_state_transitions": self.stats["total_state_transitions"],
            "agents_by_state": {
                state.value: count
                for state, count in self.stats["agents_by_state"].items()
            },
            "transitions_by_trigger": {
                trigger.value: count
                for trigger, count in self.stats["transitions_by_trigger"].items()
            },
            "avg_agent_lifetime_seconds": round(self.stats["avg_agent_lifetime_seconds"], 2),
            "health_checks_performed": self.stats["health_checks_performed"],
            "policies_applied": len(self.policies)
        }


class AgentLifecycleManagementPattern:
    """
    Comprehensive agent lifecycle management pattern coordinator with LangChain/LangGraph.
    """
    
    def __init__(self, model_name: str = "gpt-4"):
        self.manager = LifecycleManager(model_name=model_name)
    
    def create(
        self,
        agent_id: str,
        name: str,
        agent_type: str = "general"
    ) -> ManagedAgent:
        """Create a new agent."""
        return self.manager.create_agent(agent_id, name, agent_type)
    
    def initialize(self, agent_id: str, model: str = "gpt-4") -> bool:
        """Initialize agent."""
        return self.manager.initialize_agent(agent_id, model)
    
    def activate(self, agent_id: str) -> bool:
        """Activate agent."""
        return self.manager.activate_agent(agent_id)
    
    def suspend(self, agent_id: str, reason: str = "Manual suspension") -> bool:
        """Suspend agent."""
        return self.manager.suspend_agent(agent_id, reason)
    
    def resume(self, agent_id: str) -> bool:
        """Resume agent."""
        return self.manager.resume_agent(agent_id)
    
    def terminate(self, agent_id: str, reason: str = "Normal termination") -> bool:
        """Terminate agent."""
        return self.manager.terminate_agent(agent_id, reason)
    
    def update_health(
        self,
        agent_id: str,
        cpu_usage: float,
        memory_usage: float,
        error_count: int = 0
    ):
        """Update agent health."""
        self.manager.update_agent_health(agent_id, cpu_usage, memory_usage, error_count)
    
    def get_history(self, agent_id: str) -> List[Dict[str, Any]]:
        """Get agent history."""
        return self.manager.get_agent_history(agent_id)
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get statistics."""
        return self.manager.get_statistics()



if __name__ == "__main__":
    print("=== Agent Lifecycle Management Pattern with LangChain/LangGraph Demo ===\n")
    
    # Note: This demo shows the structure. In production, you'd need:
    # - export OPENAI_API_KEY="your-key-here"
    # - Actual LLM calls would be made
    
    print("Scenario 1: Complete Lifecycle")
    print("-" * 50)
    manager = AgentLifecycleManagementPattern(model_name="gpt-4")
    
    # Create agents
    agent1 = manager.create("agent-001", "DataProcessor", "data_processing")
    print(f"✓ Created {agent1.name} ({agent1.id})")
    
    agent2 = manager.create("agent-002", "APIWorker", "api_integration")
    print(f"✓ Created {agent2.name} ({agent2.id})")
    
    agent3 = manager.create("agent-003", "MonitorAgent", "monitoring")
    print(f"✓ Created {agent3.name} ({agent3.id})")
    
    # Initialize agents
    print("\nInitializing agents...")
    manager.initialize("agent-001", "gpt-4")
    manager.initialize("agent-002", "gpt-3.5-turbo")
    manager.initialize("agent-003", "gpt-4")
    print("✓ All agents initialized")
    
    # Activate agents
    print("\nActivating agents...")
    manager.activate("agent-001")
    manager.activate("agent-002")
    manager.activate("agent-003")
    print("✓ All agents activated")
    
    print("\nScenario 2: Health Monitoring")
    print("-" * 50)
    
    # Simulate health updates
    health_updates = [
        ("agent-001", 45.5, 60.2, 0),
        ("agent-002", 78.3, 85.1, 3),
        ("agent-003", 92.5, 88.0, 8),
    ]
    
    print("Updating agent health metrics:")
    for agent_id, cpu, mem, errors in health_updates:
        manager.update_health(agent_id, cpu, mem, errors)
        agent = manager.manager.agents[agent_id]
        print(f"  {agent_id}: CPU={cpu}%, MEM={mem}%, Errors={errors}")
        if agent.health:
            print(f"    Health Status: {agent.health.status.value}")
    
    print("\nScenario 3: Suspension and Resumption")
    print("-" * 50)
    
    # Suspend unhealthy agent
    print("Suspending agent with high error count...")
    success = manager.suspend("agent-003", "High error rate detected")
    if success:
        print("✓ agent-003 suspended successfully")
    
    # Resume agent after fix
    print("\nResuming agent after fixes...")
    success = manager.resume("agent-003")
    if success:
        print("✓ agent-003 resumed successfully")
    
    print("\nScenario 4: Lifecycle Policies")
    print("-" * 50)
    
    # Apply auto-management policy
    policy = LifecyclePolicy(
        policy_id="aggressive_cleanup",
        auto_suspend_after_idle_seconds=300,
        auto_terminate_after_suspended_seconds=600,
        max_error_count=10
    )
    
    manager.manager.apply_policy("agent-002", policy)
    print(f"✓ Applied policy '{policy.policy_id}' to agent-002")
    print(f"  - Auto-suspend after {policy.auto_suspend_after_idle_seconds}s idle")
    print(f"  - Auto-terminate after {policy.auto_terminate_after_suspended_seconds}s suspended")
    print(f"  - Max errors: {policy.max_error_count}")
    
    # Check policies (would normally run in background)
    print("\nChecking policy enforcement...")
    manager.manager.check_policies()
    print("✓ Policy check complete")
    
    print("\nScenario 5: Lifecycle History")
    print("-" * 50)
    
    # Get detailed history for agent
    history = manager.get_history("agent-001")
    print(f"Lifecycle history for agent-001:")
    for event in history:
        print(f"  {event['from_state']} → {event['to_state']}")
        print(f"    Trigger: {event['trigger']}")
        print(f"    Reason: {event['reason']}")
        print(f"    Time: {event['timestamp']}")
    
    print("\nScenario 6: Agent Termination")
    print("-" * 50)
    
    # Terminate agents
    terminations = [
        ("agent-001", "Task completed successfully"),
        ("agent-002", "Resource reallocation"),
    ]
    
    print("Terminating agents:")
    for agent_id, reason in terminations:
        success = manager.terminate(agent_id, reason)
        if success:
            print(f"✓ {agent_id} terminated: {reason}")
    
    print("\nScenario 7: State Distribution")
    print("-" * 50)
    
    # Get agents by state
    active_agents = manager.manager.get_agents_by_state(LifecycleState.ACTIVE)
    terminated_agents = manager.manager.get_agents_by_state(LifecycleState.TERMINATED)
    
    print(f"Active agents: {len(active_agents)}")
    for agent in active_agents:
        print(f"  - {agent.id} ({agent.name})")
    
    print(f"\nTerminated agents: {len(terminated_agents)}")
    for agent in terminated_agents:
        print(f"  - {agent.id} ({agent.name})")
    
    print("\nScenario 8: Statistics")
    print("-" * 50)
    
    stats = manager.get_statistics()
    print(f"Total Agents Created: {stats['total_agents_created']}")
    print(f"Total Active Agents: {stats['total_active_agents']}")
    print(f"Total State Transitions: {stats['total_state_transitions']}")
    print(f"Health Checks Performed: {stats['health_checks_performed']}")
    print(f"Policies Applied: {stats['policies_applied']}")
    
    print("\nAgents by State:")
    for state, count in stats['agents_by_state'].items():
        if count > 0:
            print(f"  {state}: {count}")
    
    print("\nTransitions by Trigger:")
    for trigger, count in stats['transitions_by_trigger'].items():
        if count > 0:
            print(f"  {trigger}: {count}")
    
    if stats['avg_agent_lifetime_seconds'] > 0:
        print(f"\nAverage Agent Lifetime: {stats['avg_agent_lifetime_seconds']:.2f}s")
    
    print("\n" + "=" * 50)
    print("Agent lifecycle management pattern demonstration complete!")
    print("=" * 50)

    

Agentic Architecture Pattern: Agent Lifecycle Management

Pattern 166: 166 Agent Lifecycle Management – Line-by-Line Explanation

Overview

Pattern 166: Agent Lifecycle Management Pattern with LangChain/LangGraph

Purpose: Manage complete lifecycle of agents from creation to termination with LLM insights. States: Created → Initialized → Active → Suspended → Terminated

Code Structure Analysis

File Information

• Pattern Number: 166
• Pattern Name: 166 Agent Lifecycle Management
• File Path: patterns\scaling\166_agent_lifecycle_management.py
• Total Lines: 816

Detailed Code Explanation

Module Docstring

"""
Pattern 166: Agent Lifecycle Management Pattern with LangChain/LangGraph

Purpose: Manage complete lifecycle of agents from creation to termination with LLM insights.
States: Created → Initialized → Active → Suspended → Terminated
"""

Explanation: This module docstring describes the purpose and key features of the 166 Agent Lifecycle Management pattern.

Imports Section

from typing import Dict, Any, List, Optional, TypedDict
from dataclasses import dataclass, field
from datetime import datetime, timedelta
from enum import Enum
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
from langgraph.graph import StateGraph, END
import statistics  # First 20 imports shown

Explanation: The module imports necessary dependencies: – from typing import Dict, Any, List, Optional, TypedDict: Type hints for better code documentation – from dataclasses import dataclass, field: Data class decorators for clean data structures – from datetime import datetime, timedelta: Date and time utilities – from enum import Enum: Enumeration support for fixed sets of values – from langchain_core.prompts import ChatPromptTemplate: LangChain framework integration – from langchain_openai import ChatOpenAI: LangChain framework integration – from langgraph.graph import StateGraph, END: LangGraph for workflow orchestration

Enumerations

LifecycleState (Line 18)

class LifecycleState(Enum):
    CREATED
    INITIALIZED
    ACTIVE
    SUSPENDED
    TERMINATED

Purpose: Agent lifecycle states.

Values: – CREATED: Represents created – INITIALIZED: Represents initialized – ACTIVE: Currently operational – SUSPENDED: Represents suspended – TERMINATED: Represents terminated

TransitionTrigger (Line 27)

class TransitionTrigger(Enum):
    MANUAL
    AUTOMATIC
    SCHEDULED
    ERROR
    RESOURCE_LIMIT

Purpose: Lifecycle transition triggers.

Values: – MANUAL: Represents manual – AUTOMATIC: Represents automatic – SCHEDULED: Represents scheduled – ERROR: Error severity level – RESOURCE_LIMIT: Represents resource limit

HealthStatus (Line 36)

class HealthStatus(Enum):
    HEALTHY
    DEGRADED
    UNHEALTHY
    UNKNOWN

Purpose: Agent health status.

Values: – HEALTHY: Represents healthy – DEGRADED: Represents degraded – UNHEALTHY: Represents unhealthy – UNKNOWN: Represents unknown

ResourceType (Line 44)

class ResourceType(Enum):
    CPU
    MEMORY
    NETWORK
    STORAGE

Purpose: Resource types for agents.

Values: – CPU: Represents cpu – MEMORY: Represents memory – NETWORK: Represents network – STORAGE: Represents storage

Classes

LifecycleState (Line 18)

Description: Agent lifecycle states.

TransitionTrigger (Line 27)

Description: Lifecycle transition triggers.

HealthStatus (Line 36)

Description: Agent health status.

ResourceType (Line 44)

Description: Resource types for agents.

LifecycleEvent (Line 53)

Description: Event in agent lifecycle.

ResourceAllocation (Line 65)

Description: Resource allocation for an agent.

AgentHealth (Line 75)

Description: Agent health metrics.

ManagedAgent (Line 86)

Description: Agent with comprehensive lifecycle management.

LifecyclePolicy (Line 101)

Description: Policy for lifecycle management.

LifecycleManagementState (Line 110)

Description: State for lifecycle management workflow.

LifecycleManager (Line 121)

Description: Advanced lifecycle manager with LangChain/LangGraph integration.

Methods:

__init__ (Line 126)

def __init__(self, model_name, temperature):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – model_name: LLM model to use (e.g., “gpt-4”) – temperature: Randomness in LLM responses (0.0-1.0)

_build_lifecycle_workflow (Line 143)

def _build_lifecycle_workflow(self):
    """Build LangGraph workflow for lifecycle management."""

Purpose: Build LangGraph workflow for lifecycle management.

Parameters: – self: Instance reference

_validate_transition_node (Line 164)

def _validate_transition_node(self, state):
    """Validate if state transition is allowed."""

Purpose: Validate if state transition is allowed.

Parameters: – self: Instance reference – state: Current workflow state

_check_health_node (Line 184)

def _check_health_node(self, state):
    """Check agent health."""

Purpose: Check agent health.

Parameters: – self: Instance reference – state: Current workflow state

_analyze_impact_node (Line 209)

def _analyze_impact_node(self, state):
    """Analyze impact of state transition using LLM."""

Purpose: Analyze impact of state transition using LLM.

Parameters: – self: Instance reference – state: Current workflow state

_approve_transition_node (Line 236)

def _approve_transition_node(self, state):
    """Approve or reject transition based on validation and health."""

Purpose: Approve or reject transition based on validation and health.

Parameters: – self: Instance reference – state: Current workflow state

_provide_recommendations_node (Line 252)

def _provide_recommendations_node(self, state):
    """Provide recommendations for lifecycle management."""

Purpose: Provide recommendations for lifecycle management.

Parameters: – self: Instance reference – state: Current workflow state

create_agent (Line 276)

def create_agent(self, agent_id, name, agent_type, configuration, tags):
    """Create a new managed agent."""

Purpose: Create a new managed agent.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id – name: Identifier or label – agent_type: Parameter for agent_type – configuration: Parameter for configuration – tags: Parameter for tags

initialize_agent (Line 311)

def initialize_agent(self, agent_id, model, resources):
    """Initialize agent with resources."""

Purpose: Initialize agent with resources.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id – model: Parameter for model – resources: Parameter for resources

activate_agent (Line 352)

def activate_agent(self, agent_id):
    """Activate agent for work."""

Purpose: Activate agent for work.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id

suspend_agent (Line 361)

def suspend_agent(self, agent_id, reason, trigger):
    """Suspend active agent."""

Purpose: Suspend active agent.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id – reason: Parameter for reason – trigger: Parameter for trigger

resume_agent (Line 375)

def resume_agent(self, agent_id):
    """Resume suspended agent."""

Purpose: Resume suspended agent.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id

terminate_agent (Line 384)

def terminate_agent(self, agent_id, reason, trigger):
    """Terminate agent and cleanup resources."""

Purpose: Terminate agent and cleanup resources.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id – reason: Parameter for reason – trigger: Parameter for trigger

_execute_transition (Line 409)

def _execute_transition(self, agent_id, target_state, trigger, reason):
    """Execute state transition with workflow validation."""

Purpose: Execute state transition with workflow validation.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id – target_state: Parameter for target_state – trigger: Parameter for trigger – reason: Parameter for reason

_record_event (Line 456)

def _record_event(self, agent_id, from_state, to_state, trigger, reason, metadata):
    """Record a lifecycle event."""

Purpose: Record a lifecycle event.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id – from_state: Parameter for from_state – to_state: Parameter for to_state – trigger: Parameter for trigger – reason: Parameter for reason – metadata: Parameter for metadata

…and 6 more methods

AgentLifecycleManagementPattern (Line 601)

Description: Comprehensive agent lifecycle management pattern coordinator with LangChain/LangGraph.

Methods:

__init__ (Line 606)

def __init__(self, model_name):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – model_name: LLM model to use (e.g., “gpt-4”)

create (Line 609)

def create(self, agent_id, name, agent_type):
    """Create a new agent."""

Purpose: Create a new agent.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id – name: Identifier or label – agent_type: Parameter for agent_type

initialize (Line 618)

def initialize(self, agent_id, model):
    """Initialize agent."""

Purpose: Initialize agent.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id – model: Parameter for model

activate (Line 622)

def activate(self, agent_id):
    """Activate agent."""

Purpose: Activate agent.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id

suspend (Line 626)

def suspend(self, agent_id, reason):
    """Suspend agent."""

Purpose: Suspend agent.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id – reason: Parameter for reason

resume (Line 630)

def resume(self, agent_id):
    """Resume agent."""

Purpose: Resume agent.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id

terminate (Line 634)

def terminate(self, agent_id, reason):
    """Terminate agent."""

Purpose: Terminate agent.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id – reason: Parameter for reason

update_health (Line 638)

def update_health(self, agent_id, cpu_usage, memory_usage, error_count):
    """Update agent health."""

Purpose: Update agent health.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id – cpu_usage: Parameter for cpu_usage – memory_usage: Parameter for memory_usage – error_count: Parameter for error_count

get_history (Line 648)

def get_history(self, agent_id):
    """Get agent history."""

Purpose: Get agent history.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id

get_statistics (Line 652)

def get_statistics(self):
    """Get statistics."""

Purpose: Get statistics.

Parameters: – self: Instance reference

Design Patterns Used

• Strategy Pattern: Different execution strategies for various scenarios
• State Pattern: Managing different states in the workflow
• Chain of Responsibility: Sequential processing through workflow nodes
• Observer Pattern: Tracking and monitoring pattern execution
• Factory Pattern: Creating instances with proper configuration

Usage Example

# Initialize the pattern
pattern = LifecycleState(name="example", model_name="gpt-4")

# Configure and execute
result = pattern.execute(data={"key": "value"})

# Check results
print(result)

Key Concepts

• LLM Integration: Uses LangChain/OpenAI for intelligent decision-making
• Workflow Orchestration: LangGraph manages execution flow through nodes
• State Management: TypedDict defines workflow state structure
• Error Handling: Comprehensive error handling and fallback mechanisms
• Extensibility: Designed for easy customization and extension

Extension Points

This pattern can be extended by: 1. Adding custom workflow nodes 2. Implementing additional validation logic 3. Integrating with external services 4. Customizing LLM prompts and responses 5. Adding monitoring and logging capabilities

    """
Pattern 165: Agent Pool Pattern

Purpose: Maintain a pool of reusable agents to avoid creation/destruction overhead.
Similar to connection pooling in databases.
"""

from typing import Dict, Any, List, Optional
from dataclasses import dataclass, field
from datetime import datetime
from enum import Enum
from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate


class PooledAgentStatus(Enum):
    """Status of pooled agent."""
    AVAILABLE = "available"
    IN_USE = "in_use"
    RETIRED = "retired"


@dataclass
class PooledAgent:
    """Agent in the pool."""
    id: str
    role: str
    status: PooledAgentStatus = PooledAgentStatus.AVAILABLE
    created_at: datetime = field(default_factory=datetime.now)
    last_used: datetime = field(default_factory=datetime.now)
    tasks_completed: int = 0


class AgentPoolPattern:
    """
    Agent Pool Pattern - Reusable agent instances.
    
    Features:
    - Pre-warmed agent pool
    - Checkout/checkin mechanism
    - Pool size management (min/max)
    - Idle agent cleanup
    - Usage statistics
    """
    
    def __init__(self, min_size: int = 2, max_size: int = 10, model: str = "gpt-4"):
        self.min_size = min_size
        self.max_size = max_size
        self.model = model
        self.pool: Dict[str, PooledAgent] = {}
        self.agent_llms: Dict[str, ChatOpenAI] = {}
        self.checkouts = 0
        self.checkins = 0
        
        # Initialize pool with minimum size
        for i in range(min_size):
            self._create_agent(f"pooled-agent-{i}")
    
    def _create_agent(self, agent_id: str):
        """Internal: Create a new pooled agent."""
        agent = PooledAgent(id=agent_id, role="general")
        self.pool[agent_id] = agent
        self.agent_llms[agent_id] = ChatOpenAI(model=self.model, temperature=0.7)
    
    def checkout(self) -> Optional[str]:
        """Checkout an available agent from pool."""
        # Find available agent
        for agent_id, agent in self.pool.items():
            if agent.status == PooledAgentStatus.AVAILABLE:
                agent.status = PooledAgentStatus.IN_USE
                agent.last_used = datetime.now()
                self.checkouts += 1
                return agent_id
        
        # Try to create new agent if under max size
        if len(self.pool)  bool:
        """Return agent to pool."""
        agent = self.pool.get(agent_id)
        if agent and agent.status == PooledAgentStatus.IN_USE:
            agent.status = PooledAgentStatus.AVAILABLE
            self.checkins += 1
            return True
        return False
    
    def execute_task(self, agent_id: str, task: str) -> str:
        """Execute task with checked-out agent."""
        agent = self.pool.get(agent_id)
        llm = self.agent_llms.get(agent_id)
        
        if not agent or not llm or agent.status != PooledAgentStatus.IN_USE:
            return "Agent not available"
        
        prompt = ChatPromptTemplate.from_messages([
            ("system", f"You are {agent.id} from the agent pool"),
            ("human", "{task}")
        ])
        
        chain = prompt | llm
        response = chain.invoke({"task": task})
        agent.tasks_completed += 1
        
        return str(response.content)
    
    def cleanup_idle(self, max_idle_seconds: float = 600.0):
        """Remove agents that have been idle too long."""
        current_time = datetime.now()
        to_remove = []
        
        for agent_id, agent in self.pool.items():
            if agent.status == PooledAgentStatus.AVAILABLE:
                idle_time = (current_time - agent.last_used).total_seconds()
                if idle_time > max_idle_seconds and len(self.pool) > self.min_size:
                    to_remove.append(agent_id)
        
        for agent_id in to_remove:
            del self.pool[agent_id]
            del self.agent_llms[agent_id]
    
    def get_statistics(self) -> Dict[str, Any]:
        """Get pool statistics."""
        available = sum(1 for a in self.pool.values() if a.status == PooledAgentStatus.AVAILABLE)
        in_use = sum(1 for a in self.pool.values() if a.status == PooledAgentStatus.IN_USE)
        
        return {
            "pool_size": len(self.pool),
            "available": available,
            "in_use": in_use,
            "total_checkouts": self.checkouts,
            "total_checkins": self.checkins,
            "utilization": in_use / len(self.pool) if self.pool else 0
        }


if __name__ == "__main__":
    print("=== Agent Pool Pattern Demo ===\n")
    
    pool = AgentPoolPattern(min_size=2, max_size=5)
    
    print(f"Pool initialized with {len(pool.pool)} agents\n")
    
    # Checkout agents and use them
    print("Checking out agents...")
    agent1 = pool.checkout()
    agent2 = pool.checkout()
    
    if agent1:
        print(f"Checked out: {agent1}")
        result = pool.execute_task(agent1, "Process data")
        print(f"Result: {result[:50]}...")
        pool.checkin(agent1)
        print(f"Checked in: {agent1}")
    
    if agent2:
        print(f"Checked out: {agent2}")
        result = pool.execute_task(agent2, "Generate report")
        print(f"Result: {result[:50]}...")
        pool.checkin(agent2)
        print(f"Checked in: {agent2}")
    
    # Statistics
    print("\n=== Statistics ===")
    stats = pool.get_statistics()
    for key, value in stats.items():
        print(f"{key}: {value}")

    

Agentic Architecture Pattern: Agent Pool

Pattern 165: 165 Agent Pool – Line-by-Line Explanation

Overview

Pattern 165: Agent Pool Pattern

Purpose: Maintain a pool of reusable agents to avoid creation/destruction overhead. Similar to connection pooling in databases.

Code Structure Analysis

File Information

• Pattern Number: 165
• Pattern Name: 165 Agent Pool
• File Path: patterns\scaling\165_agent_pool.py
• Total Lines: 172

Detailed Code Explanation

Module Docstring

"""
Pattern 165: Agent Pool Pattern

Purpose: Maintain a pool of reusable agents to avoid creation/destruction overhead.
Similar to connection pooling in databases.
"""

Explanation: This module docstring describes the purpose and key features of the 165 Agent Pool pattern.

Imports Section

from typing import Dict, Any, List, Optional
from dataclasses import dataclass, field
from datetime import datetime
from enum import Enum
from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate  # First 20 imports shown

Explanation: The module imports necessary dependencies: – from typing import Dict, Any, List, Optional: Type hints for better code documentation – from dataclasses import dataclass, field: Data class decorators for clean data structures – from datetime import datetime: Date and time utilities – from enum import Enum: Enumeration support for fixed sets of values – from langchain_openai import ChatOpenAI: LangChain framework integration – from langchain_core.prompts import ChatPromptTemplate: LangChain framework integration

Enumerations

PooledAgentStatus (Line 16)

class PooledAgentStatus(Enum):
    AVAILABLE
    IN_USE
    RETIRED

Purpose: Status of pooled agent.

Values: – AVAILABLE: Represents available – IN_USE: Represents in use – RETIRED: Represents retired

Classes

PooledAgentStatus (Line 16)

Description: Status of pooled agent.

PooledAgent (Line 24)

Description: Agent in the pool.

AgentPoolPattern (Line 34)

Description: Agent Pool Pattern – Reusable agent instances.

Features: – Pre-warmed agent pool – Checkout/checkin mechanism – Pool size management (min/max) – Idle agent cleanup – Usage statistics

Methods:

__init__ (Line 46)

def __init__(self, min_size, max_size, model):
    """No method docstring."""

Purpose: No method docstring.

Parameters: – self: Instance reference – min_size: Parameter for min_size – max_size: Parameter for max_size – model: Parameter for model

_create_agent (Line 59)

def _create_agent(self, agent_id):
    """Internal: Create a new pooled agent."""

Purpose: Internal: Create a new pooled agent.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id

checkout (Line 65)

def checkout(self):
    """Checkout an available agent from pool."""

Purpose: Checkout an available agent from pool.

Parameters: – self: Instance reference

checkin (Line 83)

def checkin(self, agent_id):
    """Return agent to pool."""

Purpose: Return agent to pool.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id

execute_task (Line 92)

def execute_task(self, agent_id, task):
    """Execute task with checked-out agent."""

Purpose: Execute task with checked-out agent.

Parameters: – self: Instance reference – agent_id: Parameter for agent_id – task: Parameter for task

cleanup_idle (Line 111)

def cleanup_idle(self, max_idle_seconds):
    """Remove agents that have been idle too long."""

Purpose: Remove agents that have been idle too long.

Parameters: – self: Instance reference – max_idle_seconds: Parameter for max_idle_seconds

get_statistics (Line 126)

def get_statistics(self):
    """Get pool statistics."""

Purpose: Get pool statistics.

Parameters: – self: Instance reference

Design Patterns Used

• Strategy Pattern: Different execution strategies for various scenarios
• State Pattern: Managing different states in the workflow
• Chain of Responsibility: Sequential processing through workflow nodes
• Observer Pattern: Tracking and monitoring pattern execution
• Factory Pattern: Creating instances with proper configuration

Usage Example

# Initialize the pattern
pattern = PooledAgentStatus(name="example", model_name="gpt-4")

# Configure and execute
result = pattern.execute(data={"key": "value"})

# Check results
print(result)

Key Concepts

• LLM Integration: Uses LangChain/OpenAI for intelligent decision-making
• Workflow Orchestration: LangGraph manages execution flow through nodes
• State Management: TypedDict defines workflow state structure
• Error Handling: Comprehensive error handling and fallback mechanisms
• Extensibility: Designed for easy customization and extension

Extension Points

This pattern can be extended by: 1. Adding custom workflow nodes 2. Implementing additional validation logic 3. Integrating with external services 4. Customizing LLM prompts and responses 5. Adding monitoring and logging capabilities