aishastr.live

Agentic Design Pattern: Plugin Extension

Written by

admin

in


    """
Pattern 158: Plugin/Extension Architecture

Description:
    Enables extensibility through a plugin system where new capabilities can be
    added dynamically without modifying core code. Plugins follow a standard
    interface and can be loaded, configured, and executed at runtime.

Components:
    - Plugin interface/protocol
    - Plugin registry
    - Plugin loader
    - Lifecycle management
    - Dependency injection
    - Configuration system

Use Cases:
    - Custom tool integration
    - Domain-specific capabilities
    - Third-party extensions
    - Modular agent systems
    - Dynamic feature addition

Benefits:
    - Extensibility without core changes
    - Modular architecture
    - Third-party contributions
    - Dynamic capability loading
    - Separation of concerns

Trade-offs:
    - Complexity overhead
    - Plugin management burden
    - Potential security risks
    - Version compatibility issues

LangChain Implementation:
    Uses protocol classes for interfaces, dynamic loading,
    registry pattern, and lifecycle hooks
"""

import os
import importlib
import inspect
from typing import Dict, Any, List, Optional, Callable, Protocol
from dataclasses import dataclass, field
from enum import Enum
from abc import ABC, abstractmethod
from dotenv import load_dotenv

from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser

load_dotenv()


class PluginStatus(Enum):
    """Plugin status"""
    UNLOADED = "unloaded"
    LOADED = "loaded"
    INITIALIZED = "initialized"
    ACTIVE = "active"
    DISABLED = "disabled"
    ERROR = "error"


class PluginType(Enum):
    """Plugin types"""
    TOOL = "tool"
    PROCESSOR = "processor"
    FILTER = "filter"
    ANALYZER = "analyzer"
    GENERATOR = "generator"


@dataclass
class PluginMetadata:
    """Plugin metadata"""
    name: str
    version: str
    author: str
    description: str
    plugin_type: PluginType
    dependencies: List[str] = field(default_factory=list)
    config_schema: Dict[str, Any] = field(default_factory=dict)


@dataclass
class PluginConfig:
    """Plugin configuration"""
    enabled: bool = True
    priority: int = 0
    settings: Dict[str, Any] = field(default_factory=dict)


class Plugin(Protocol):
    """Plugin interface using Protocol (structural typing)"""
    
    @property
    def metadata(self) -> PluginMetadata:
        """Get plugin metadata"""
        ...
    
    def initialize(self, config: PluginConfig) -> bool:
        """Initialize plugin"""
        ...
    
    def execute(self, input_data: Any, context: Dict[str, Any]) -> Any:
        """Execute plugin functionality"""
        ...
    
    def shutdown(self) -> bool:
        """Cleanup and shutdown"""
        ...


class BasePlugin(ABC):
    """Base class for plugins"""
    
    def __init__(self):
        """Initialize base plugin"""
        self.config: Optional[PluginConfig] = None
        self.status = PluginStatus.UNLOADED
        self._metadata: Optional[PluginMetadata] = None
    
    @property
    def metadata(self) -> PluginMetadata:
        """Get plugin metadata"""
        if self._metadata is None:
            self._metadata = self._create_metadata()
        return self._metadata
    
    @abstractmethod
    def _create_metadata(self) -> PluginMetadata:
        """Create plugin metadata"""
        pass
    
    def initialize(self, config: PluginConfig) -> bool:
        """Initialize plugin"""
        try:
            self.config = config
            self._on_initialize()
            self.status = PluginStatus.INITIALIZED
            if config.enabled:
                self.status = PluginStatus.ACTIVE
            return True
        except Exception as e:
            self.status = PluginStatus.ERROR
            print(f"Plugin initialization failed: {e}")
            return False
    
    def _on_initialize(self):
        """Override for custom initialization"""
        pass
    
    @abstractmethod
    def execute(self, input_data: Any, context: Dict[str, Any]) -> Any:
        """Execute plugin functionality"""
        pass
    
    def shutdown(self) -> bool:
        """Cleanup and shutdown"""
        try:
            self._on_shutdown()
            self.status = PluginStatus.UNLOADED
            return True
        except Exception as e:
            print(f"Plugin shutdown failed: {e}")
            return False
    
    def _on_shutdown(self):
        """Override for custom shutdown"""
        pass


# Example Plugin Implementations

class SentimentAnalysisPlugin(BasePlugin):
    """Plugin for sentiment analysis"""
    
    def _create_metadata(self) -> PluginMetadata:
        return PluginMetadata(
            name="sentiment_analyzer",
            version="1.0.0",
            author="Example Corp",
            description="Analyzes sentiment of text using LLM",
            plugin_type=PluginType.ANALYZER
        )
    
    def _on_initialize(self):
        """Initialize LLM"""
        self.llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0)
    
    def execute(self, input_data: Any, context: Dict[str, Any]) -> Any:
        """Analyze sentiment"""
        text = input_data if isinstance(input_data, str) else str(input_data)
        
        prompt = ChatPromptTemplate.from_messages([
            ("system", "Analyze the sentiment of the given text. Respond with: positive, negative, or neutral."),
            ("user", "{text}")
        ])
        
        chain = prompt | self.llm | StrOutputParser()
        sentiment = chain.invoke({"text": text})
        
        return {
            "sentiment": sentiment.strip().lower(),
            "text": text,
            "plugin": self.metadata.name
        }


class TextSummarizerPlugin(BasePlugin):
    """Plugin for text summarization"""
    
    def _create_metadata(self) -> PluginMetadata:
        return PluginMetadata(
            name="text_summarizer",
            version="1.0.0",
            author="Example Corp",
            description="Summarizes text content",
            plugin_type=PluginType.PROCESSOR
        )
    
    def _on_initialize(self):
        """Initialize LLM"""
        max_words = self.config.settings.get("max_words", 50) if self.config else 50
        self.max_words = max_words
        self.llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0.3)
    
    def execute(self, input_data: Any, context: Dict[str, Any]) -> Any:
        """Summarize text"""
        text = input_data if isinstance(input_data, str) else str(input_data)
        
        prompt = ChatPromptTemplate.from_messages([
            ("system", "Summarize the following text in {max_words} words or less."),
            ("user", "{text}")
        ])
        
        chain = prompt | self.llm | StrOutputParser()
        summary = chain.invoke({"text": text, "max_words": self.max_words})
        
        return {
            "summary": summary,
            "original_length": len(text.split()),
            "summary_length": len(summary.split()),
            "plugin": self.metadata.name
        }


class KeywordExtractorPlugin(BasePlugin):
    """Plugin for keyword extraction"""
    
    def _create_metadata(self) -> PluginMetadata:
        return PluginMetadata(
            name="keyword_extractor",
            version="1.0.0",
            author="Example Corp",
            description="Extracts keywords from text",
            plugin_type=PluginType.ANALYZER
        )
    
    def _on_initialize(self):
        """Initialize LLM"""
        self.llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0)
    
    def execute(self, input_data: Any, context: Dict[str, Any]) -> Any:
        """Extract keywords"""
        text = input_data if isinstance(input_data, str) else str(input_data)
        
        prompt = ChatPromptTemplate.from_messages([
            ("system", "Extract 5-7 key terms or phrases from the text. Return as comma-separated list."),
            ("user", "{text}")
        ])
        
        chain = prompt | self.llm | StrOutputParser()
        keywords_str = chain.invoke({"text": text})
        
        keywords = [k.strip() for k in keywords_str.split(",")]
        
        return {
            "keywords": keywords,
            "count": len(keywords),
            "plugin": self.metadata.name
        }


class ProfanityFilterPlugin(BasePlugin):
    """Plugin for filtering profanity"""
    
    def _create_metadata(self) -> PluginMetadata:
        return PluginMetadata(
            name="profanity_filter",
            version="1.0.0",
            author="Example Corp",
            description="Filters inappropriate content",
            plugin_type=PluginType.FILTER
        )
    
    def _on_initialize(self):
        """Initialize filter"""
        self.llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0)
    
    def execute(self, input_data: Any, context: Dict[str, Any]) -> Any:
        """Filter content"""
        text = input_data if isinstance(input_data, str) else str(input_data)
        
        prompt = ChatPromptTemplate.from_messages([
            ("system", "Analyze if the text contains profanity or inappropriate content. Respond with 'yes' or 'no'."),
            ("user", "{text}")
        ])
        
        chain = prompt | self.llm | StrOutputParser()
        contains_profanity = chain.invoke({"text": text})
        
        is_inappropriate = "yes" in contains_profanity.lower()
        
        return {
            "is_appropriate": not is_inappropriate,
            "filtered": is_inappropriate,
            "text": text if not is_inappropriate else "[CONTENT FILTERED]",
            "plugin": self.metadata.name
        }


class PluginRegistry:
    """Manages plugin registration and lifecycle"""
    
    def __init__(self):
        """Initialize plugin registry"""
        self.plugins: Dict[str, BasePlugin] = {}
        self.configs: Dict[str, PluginConfig] = {}
        self.execution_order: List[str] = []
    
    def register_plugin(self, plugin: BasePlugin, config: Optional[PluginConfig] = None):
        """Register a plugin"""
        if config is None:
            config = PluginConfig()
        
        plugin_name = plugin.metadata.name
        
        # Check dependencies
        for dep in plugin.metadata.dependencies:
            if dep not in self.plugins:
                print(f"Warning: Missing dependency '{dep}' for plugin '{plugin_name}'")
        
        # Register plugin
        self.plugins[plugin_name] = plugin
        self.configs[plugin_name] = config
        
        # Initialize plugin
        success = plugin.initialize(config)
        
        if success:
            # Add to execution order based on priority
            self._update_execution_order(plugin_name, config.priority)
            print(f"✓ Registered and initialized plugin: {plugin_name}")
        else:
            print(f"✗ Failed to initialize plugin: {plugin_name}")
        
        return success
    
    def _update_execution_order(self, plugin_name: str, priority: int):
        """Update execution order based on priority"""
        if plugin_name in self.execution_order:
            self.execution_order.remove(plugin_name)
        
        # Insert based on priority (higher priority first)
        inserted = False
        for i, name in enumerate(self.execution_order):
            if self.configs[name].priority  bool:
        """Unregister a plugin"""
        if plugin_name not in self.plugins:
            return False
        
        plugin = self.plugins[plugin_name]
        plugin.shutdown()
        
        del self.plugins[plugin_name]
        del self.configs[plugin_name]
        self.execution_order.remove(plugin_name)
        
        print(f"✓ Unregistered plugin: {plugin_name}")
        return True
    
    def get_plugin(self, plugin_name: str) -> Optional[BasePlugin]:
        """Get plugin by name"""
        return self.plugins.get(plugin_name)
    
    def execute_plugin(self, plugin_name: str, input_data: Any, 
                      context: Optional[Dict[str, Any]] = None) -> Any:
        """Execute specific plugin"""
        if context is None:
            context = {}
        
        plugin = self.plugins.get(plugin_name)
        if not plugin:
            raise ValueError(f"Plugin not found: {plugin_name}")
        
        if plugin.status != PluginStatus.ACTIVE:
            raise RuntimeError(f"Plugin not active: {plugin_name}")
        
        return plugin.execute(input_data, context)
    
    def execute_pipeline(self, input_data: Any, 
                        plugin_types: Optional[List[PluginType]] = None,
                        context: Optional[Dict[str, Any]] = None) -> Any:
        """Execute plugins in pipeline"""
        if context is None:
            context = {}
        
        result = input_data
        
        for plugin_name in self.execution_order:
            plugin = self.plugins[plugin_name]
            config = self.configs[plugin_name]
            
            # Skip if not enabled
            if not config.enabled or plugin.status != PluginStatus.ACTIVE:
                continue
            
            # Filter by type if specified
            if plugin_types and plugin.metadata.plugin_type not in plugin_types:
                continue
            
            # Execute plugin
            result = plugin.execute(result, context)
        
        return result
    
    def list_plugins(self) -> List[Dict[str, Any]]:
        """List all registered plugins"""
        return [
            {
                "name": plugin.metadata.name,
                "version": plugin.metadata.version,
                "type": plugin.metadata.plugin_type.value,
                "status": plugin.status.value,
                "enabled": self.configs[name].enabled,
                "priority": self.configs[name].priority
            }
            for name, plugin in self.plugins.items()
        ]
    
    def shutdown_all(self):
        """Shutdown all plugins"""
        for plugin_name in list(self.plugins.keys()):
            self.unregister_plugin(plugin_name)


def demonstrate_plugin_architecture():
    """Demonstrate plugin/extension architecture"""
    print("=" * 80)
    print("PLUGIN/EXTENSION ARCHITECTURE DEMONSTRATION")
    print("=" * 80)
    
    # Initialize registry
    registry = PluginRegistry()
    
    # Example 1: Register plugins
    print("\n" + "="*80)
    print("EXAMPLE 1: Registering Plugins")
    print("="*80)
    
    # Register sentiment analyzer
    sentiment_plugin = SentimentAnalysisPlugin()
    registry.register_plugin(sentiment_plugin, PluginConfig(
        enabled=True,
        priority=10
    ))
    
    # Register summarizer
    summarizer_plugin = TextSummarizerPlugin()
    registry.register_plugin(summarizer_plugin, PluginConfig(
        enabled=True,
        priority=5,
        settings={"max_words": 30}
    ))
    
    # Register keyword extractor
    keyword_plugin = KeywordExtractorPlugin()
    registry.register_plugin(keyword_plugin, PluginConfig(
        enabled=True,
        priority=3
    ))
    
    # Register profanity filter
    filter_plugin = ProfanityFilterPlugin()
    registry.register_plugin(filter_plugin, PluginConfig(
        enabled=True,
        priority=20  # High priority - run first
    ))
    
    # Example 2: List plugins
    print("\n" + "="*80)
    print("EXAMPLE 2: List Registered Plugins")
    print("="*80)
    plugins = registry.list_plugins()
    for plugin in plugins:
        print(f"- {plugin['name']} v{plugin['version']}")
        print(f"  Type: {plugin['type']}, Status: {plugin['status']}")
        print(f"  Priority: {plugin['priority']}, Enabled: {plugin['enabled']}")
    
    # Example 3: Execute individual plugin
    print("\n" + "="*80)
    print("EXAMPLE 3: Execute Individual Plugin")
    print("="*80)
    text = "This product is amazing! I love it so much!"
    result = registry.execute_plugin("sentiment_analyzer", text)
    print(f"Input: {text}")
    print(f"Result: {result}")
    
    # Example 4: Execute multiple plugins
    print("\n" + "="*80)
    print("EXAMPLE 4: Execute Multiple Plugins")
    print("="*80)
    text = "Artificial intelligence is transforming how we work and live. Machine learning models can now process vast amounts of data and make predictions with high accuracy."
    
    print(f"Input: {text[:100]}...\n")
    
    # Sentiment
    sentiment_result = registry.execute_plugin("sentiment_analyzer", text)
    print(f"Sentiment: {sentiment_result['sentiment']}")
    
    # Summary
    summary_result = registry.execute_plugin("text_summarizer", text)
    print(f"Summary: {summary_result['summary']}")
    
    # Keywords
    keyword_result = registry.execute_plugin("keyword_extractor", text)
    print(f"Keywords: {', '.join(keyword_result['keywords'])}")
    
    # Example 5: Filter by plugin type
    print("\n" + "="*80)
    print("EXAMPLE 5: Execute by Plugin Type")
    print("="*80)
    text = "The company announced record profits this quarter."
    
    # Execute only analyzers
    print("Executing ANALYZER plugins:")
    result = registry.execute_pipeline(text, plugin_types=[PluginType.ANALYZER])
    print(f"Result: {result}")
    
    # Example 6: Content filtering
    print("\n" + "="*80)
    print("EXAMPLE 6: Content Filtering")
    print("="*80)
    clean_text = "This is a nice and appropriate message."
    result = registry.execute_plugin("profanity_filter", clean_text)
    print(f"Text: {clean_text}")
    print(f"Is Appropriate: {result['is_appropriate']}")
    print(f"Output: {result['text']}")
    
    # Example 7: Execution order (priority)
    print("\n" + "="*80)
    print("EXAMPLE 7: Plugin Execution Order")
    print("="*80)
    print("Execution order (by priority):")
    for i, plugin_name in enumerate(registry.execution_order, 1):
        plugin = registry.get_plugin(plugin_name)
        priority = registry.configs[plugin_name].priority
        print(f"{i}. {plugin_name} (priority: {priority})")
    
    # Cleanup
    print("\n" + "="*80)
    print("CLEANUP: Shutting Down Plugins")
    print("="*80)
    registry.shutdown_all()
    
    # Summary
    print("\n" + "="*80)
    print("SUMMARY: Plugin Architecture Best Practices")
    print("="*80)
    print("""
1. PLUGIN INTERFACE:
   - Define clear plugin interface/protocol
   - Standardize lifecycle methods (initialize, execute, shutdown)
   - Use metadata for plugin description
   - Support configuration schemas

2. REGISTRY MANAGEMENT:
   - Centralized plugin registration
   - Dependency checking
   - Priority-based execution
   - Enable/disable plugins dynamically

3. LIFECYCLE MANAGEMENT:
   - Proper initialization
   - Resource cleanup on shutdown
   - Error handling and status tracking
   - Graceful failure handling

4. EXECUTION PATTERNS:
   - Individual plugin execution
   - Pipeline execution
   - Type-based filtering
   - Context passing between plugins

5. CONFIGURATION:
   - Plugin-specific settings
   - Priority control
   - Enable/disable flags
   - Runtime configuration updates

6. EXTENSIBILITY:
   - Easy to add new plugins
   - No core code modification
   - Third-party plugin support
   - Hot-loading capabilities (advanced)

Benefits:
✓ Modular and extensible architecture
✓ Separation of concerns
✓ Easy feature addition
✓ Third-party contributions
✓ Dynamic capability enhancement
✓ Clean core codebase

Challenges:
- Plugin management complexity
- Version compatibility
- Security considerations for third-party plugins
- Performance overhead
- Dependency management
- Testing across plugin combinations
    """)


if __name__ == "__main__":
    demonstrate_plugin_architecture()

Agentic Design Pattern: Plugin Extension

Pattern 158: Plugin Extension – Line-by-Line Explanation

Overview

This document provides a detailed line-by-line explanation of the Plugin Extension pattern implementation.

Pattern Description

Pattern 158: Plugin/Extension Architecture

Description: Enables extensibility through a plugin system where new capabilities can be added dynamically without modifying core code. Plugins follow a standard interface and can be loaded, configured, and executed at runtime.

Components: – Plugin interface/protocol – Plugin registry – Plugin loader – Lifecycle management – Dependency injection – Configuration system

Use Cases: – Custom tool integration – Domain-specific capabilities – Third-party extensions – Modular agent systems – Dynamic feature addition

Benefits: – Extensibility without core changes – Modular architecture – Third-party contributions – Dynamic capability loading – Separation of concerns

Trade-offs: – Complexity overhead – Plugin management burden – Potential security risks – Version compatibility issues

LangChain Implementation: Uses protocol classes for interfaces, dynamic loading, registry pattern, and lifecycle hooks

File Statistics

  • Total Lines: 640
  • Classes: 11
  • Functions: 1
  • Constants: 0

Module Documentation

"""
Pattern 158: Plugin/Extension Architecture

Description:
    Enables extensibility through a plugin system where new capabilities can be
    added dynamically without modifying core code. Plugins follow a standard
    interface and can be loaded, configured, and executed at runtime.

Components:
    - Plugin interface/protocol
    - Plugin registry
    - Plugin loader
    - Lifecycle management
    - Dependency injection
    - Configuration system

Use Cases:
    - Custom tool integration
    - Domain-specific capabilities
    - Third-party extensions
    - Modular agent systems
    - Dynamic feature addition

Benefits:
    - Extensibility without core changes
    - Modular architecture
    - Third-party contributions
    - Dynamic capability loading
    - Separation of concerns

Trade-offs:
    - Complexity overhead
    - Plugin management burden
    - Potential security risks
    - Version compatibility issues

LangChain Implementation:
    Uses protocol classes for interfaces, dynamic loading,
    registry pattern, and lifecycle hooks
"""

Purpose: This module implements the Plugin Extension pattern, providing a structured approach to agent design and implementation.

Import Statements

Standard Library Imports

import dataclasses.dataclass
import dataclasses.field
import enum.Enum
import langchain_core.output_parsers.StrOutputParser
import langchain_core.prompts.ChatPromptTemplate
import os
import typing.Any
import typing.Callable
import typing.Dict
import typing.List
import typing.Optional
import typing.Protocol

Purpose: Core Python functionality: – dataclasses.dataclass: Simplified class definitions with @dataclass – dataclasses.field: Simplified class definitions with @dataclass – enum.Enum: Enumeration support – langchain_core.output_parsers.StrOutputParser: Regular expression operations – langchain_core.prompts.ChatPromptTemplate: Regular expression operations – os: Environment variable access and system operations – typing.Any: Type hints for better code documentation (List, Dict, Any, Optional) – typing.Callable: Type hints for better code documentation (List, Dict, Any, Optional) – typing.Dict: Type hints for better code documentation (List, Dict, Any, Optional) – typing.List: Type hints for better code documentation (List, Dict, Any, Optional) – typing.Optional: Type hints for better code documentation (List, Dict, Any, Optional) – typing.Protocol: Type hints for better code documentation (List, Dict, Any, Optional)

LangChain Imports

from langchain_openai.ChatOpenAI

Purpose: LangChain framework components: – LLM Interface: Wrapper for OpenAI models – Chains: Composable building blocks for complex workflows

Other Imports

from abc.ABC
from abc.abstractmethod
from dotenv.load_dotenv
from importlib
from inspect

Purpose: Additional dependencies for specialized functionality.

Classes

Class: PluginStatus

Lines 58-65

Inherits from: Enum

Purpose:

Plugin status

Class: PluginType

Lines 68-74

Inherits from: Enum

Purpose:

Plugin types

Class: PluginMetadata

Lines 78-86

Purpose:

Plugin metadata

Class: PluginConfig

Lines 90-94

Purpose:

Plugin configuration

Class: Plugin

Lines 97-115

Inherits from: Protocol

Purpose:

Plugin interface using Protocol (structural typing)

Methods:

metadata(self)

Lines 101-103

Decorators: property

Purpose:

Get plugin metadata

Parameters:

Returns: PluginMetadata

initialize(self, config)

Lines 105-107

Purpose:

Initialize plugin

Parameters: – config: Config

Returns: bool

execute(self, input_data, context)

Lines 109-111

Purpose:

Execute plugin functionality

Parameters: – input_data: Input data – context: Context

Returns: Any

shutdown(self)

Lines 113-115

Purpose:

Cleanup and shutdown

Parameters:

Returns: bool

Class: BasePlugin

Lines 118-174

Inherits from: ABC

Purpose:

Base class for plugins

Methods:

__init__(self)

Lines 121-125

Purpose:

Initialize base plugin

Parameters:

metadata(self)

Lines 128-132

Decorators: property

Purpose:

Get plugin metadata

Parameters:

Returns: PluginMetadata

_create_metadata(self)

Lines 135-137

Decorators: abstractmethod

Purpose:

Create plugin metadata

Parameters:

Returns: PluginMetadata

initialize(self, config)

Lines 139-151

Purpose:

Initialize plugin

Parameters: – config: Config

Returns: bool

_on_initialize(self)

Lines 153-155

Purpose:

Override for custom initialization

Parameters:

execute(self, input_data, context)

Lines 158-160

Decorators: abstractmethod

Purpose:

Execute plugin functionality

Parameters: – input_data: Input data – context: Context

Returns: Any

shutdown(self)

Lines 162-170

Purpose:

Cleanup and shutdown

Parameters:

Returns: bool

_on_shutdown(self)

Lines 172-174

Purpose:

Override for custom shutdown

Parameters:

Class: SentimentAnalysisPlugin

Lines 179-211

Inherits from: BasePlugin

Purpose:

Plugin for sentiment analysis

Methods:

_create_metadata(self)

Lines 182-189

Purpose: Create and return a new instance or configuration.

Parameters:

Returns: PluginMetadata

_on_initialize(self)

Lines 191-193

Purpose:

Initialize LLM

Parameters:

execute(self, input_data, context)

Lines 195-211

Purpose:

Analyze sentiment

Parameters: – input_data: Input data – context: Context

Returns: Any

Class: TextSummarizerPlugin

Lines 214-249

Inherits from: BasePlugin

Purpose:

Plugin for text summarization

Methods:

_create_metadata(self)

Lines 217-224

Purpose: Create and return a new instance or configuration.

Parameters:

Returns: PluginMetadata

_on_initialize(self)

Lines 226-230

Purpose:

Initialize LLM

Parameters:

execute(self, input_data, context)

Lines 232-249

Purpose:

Summarize text

Parameters: – input_data: Input data – context: Context

Returns: Any

Class: KeywordExtractorPlugin

Lines 252-286

Inherits from: BasePlugin

Purpose:

Plugin for keyword extraction

Methods:

_create_metadata(self)

Lines 255-262

Purpose: Create and return a new instance or configuration.

Parameters:

Returns: PluginMetadata

_on_initialize(self)

Lines 264-266

Purpose:

Initialize LLM

Parameters:

execute(self, input_data, context)

Lines 268-286

Purpose:

Extract keywords

Parameters: – input_data: Input data – context: Context

Returns: Any

Class: ProfanityFilterPlugin

Lines 289-324

Inherits from: BasePlugin

Purpose:

Plugin for filtering profanity

Methods:

_create_metadata(self)

Lines 292-299

Purpose: Create and return a new instance or configuration.

Parameters:

Returns: PluginMetadata

_on_initialize(self)

Lines 301-303

Purpose:

Initialize filter

Parameters:

execute(self, input_data, context)

Lines 305-324

Purpose:

Filter content

Parameters: – input_data: Input data – context: Context

Returns: Any

Class: PluginRegistry

Lines 327-457

Purpose:

Manages plugin registration and lifecycle

Methods:

__init__(self)

Lines 330-334

Purpose:

Initialize plugin registry

Parameters:

register_plugin(self, plugin, config)

Lines 336-362

Purpose:

Register a plugin

Parameters: – plugin: Plugin – config: Config

_update_execution_order(self, plugin_name, priority)

Lines 364-378

Purpose:

Update execution order based on priority

Parameters: – plugin_name: Plugin name – priority: Priority

unregister_plugin(self, plugin_name)

Lines 380-393

Purpose:

Unregister a plugin

Parameters: – plugin_name: Plugin name

Returns: bool

get_plugin(self, plugin_name)

Lines 395-397

Purpose:

Get plugin by name

Parameters: – plugin_name: Plugin name

Returns: <ast.Subscript object at 0x000001A06D7CC850>

execute_plugin(self, plugin_name, input_data, context)

Lines 399-412

Purpose:

Execute specific plugin

Parameters: – plugin_name: Plugin name – input_data: Input data – context: Context

Returns: Any

execute_pipeline(self, input_data, plugin_types, context)

Lines 414-438

Purpose:

Execute plugins in pipeline

Parameters: – input_data: Input data – plugin_types: Plugin types – context: Context

Returns: Any

list_plugins(self)

Lines 440-452

Purpose:

List all registered plugins

Parameters:

Returns: <ast.Subscript object at 0x000001A06D6F6290>

shutdown_all(self)

Lines 454-457

Purpose:

Shutdown all plugins

Parameters:

Functions

demonstrate_plugin_architecture()

Lines 460-635

Purpose:

Demonstrate plugin/extension architecture

Key Concepts Summary

Pattern Overview

The Plugin Extension pattern provides a structured approach to building AI agents with specific capabilities and behaviors. This implementation leverages LangChain’s composable components to create a robust and maintainable solution.

Core Components

  1. Agent Class: Main implementation of the pattern logic
  2. Prompt Engineering: Carefully crafted prompts to guide LLM behavior
  3. Tool Integration: Optional tools for extended functionality
  4. Memory Management: State handling across interactions
  5. Output Parsing: Structured response processing

When to Use

✅ Good for: – Scenarios requiring this specific pattern’s strengths – Applications needing structured agent behavior – Systems where explainability is important – Production environments requiring reliability

❌ Not ideal for: – Simple single-prompt tasks – Scenarios requiring different pattern characteristics – Resource-constrained environments (if computationally intensive)

Best Practices Demonstrated

1. Environment Configuration

  • ✅ Use .env files for API keys
  • ✅ Never hardcode credentials
  • ✅ Validate environment variables on startup

2. Error Handling

  • ✅ Graceful degradation on failures
  • ✅ Informative error messages
  • ✅ Logging for debugging

3. Prompt Engineering

  • ✅ Clear, specific instructions
  • ✅ Structured output formats
  • ✅ Few-shot examples when beneficial

4. Code Organization

  • ✅ Separation of concerns (class-based design)
  • ✅ Type hints for clarity
  • ✅ Docstrings for documentation

5. Performance Optimization

  • ✅ Appropriate temperature settings
  • ✅ Token budget management
  • ✅ Caching where applicable

LangChain Components Used

  • ChatOpenAI: OpenAI model interface
  • Chains: Sequential component composition

Why LangChain? – Composable abstractions for LLM applications – Rich ecosystem of integrations – Production-ready patterns – Active community and development

Execution Flow

1. Initialize agent with configuration
   ↓
2. Load environment and dependencies
   ↓
3. Prepare prompt with input data
   ↓
4. Invoke LLM through LangChain
   ↓
5. Parse and validate response
   ↓
6. Return formatted result

Key Decision Points: – Model selection (temperature, model type) – Prompt structure (few-shot vs zero-shot) – Error handling strategy – Output format validation

Performance Considerations

Computational Cost

  • LLM Calls: Primary cost factor (based on tokens)
  • Latency: Depends on model and input size
  • Memory: Minimal for most patterns

Optimization Strategies

  1. Prompt Optimization: Minimize tokens while maintaining clarity
  2. Caching: Cache repeated queries/responses
  3. Model Selection: Use smaller models when appropriate
  4. Batch Processing: Group similar requests
  5. Streaming: For long responses, stream tokens

Cost Estimates (Approximate)

  • GPT-3.5-turbo: ~$0.002 per request
  • GPT-4: ~$0.03 per request
  • GPT-4-turbo: ~$0.01 per request

Actual costs vary based on input/output token counts

Common Pitfalls to Avoid

1. Prompt Design Issues

❌ Vague or ambiguous instructions
✅ Clear, specific, testable prompts

2. Error Handling

❌ Ignoring API errors or rate limits
✅ Implement retry logic and graceful degradation

3. Token Management

❌ Exceeding model token limits
✅ Truncate inputs, chunk processing

4. Security

❌ Exposing API keys in code
✅ Use environment variables

5. Testing

❌ Only testing happy paths
✅ Test edge cases, failures, malformed inputs

6. Cost Management

❌ Unmonitored API usage
✅ Implement budgets, alerts, and monitoring

Further Improvements

Production Enhancements

  1. Monitoring: Add logging, metrics, tracing
  2. Validation: Input/output schema validation
  3. Rate Limiting: Implement request throttling
  4. Caching: Redis or in-memory cache for responses
  5. Async Support: Non-blocking operations for scalability

Feature Extensions

  1. Multi-Model Support: Switch between providers
  2. Fallback Models: Graceful degradation strategy
  3. Custom Tools: Domain-specific function calling
  4. Advanced Memory: Vector stores for long-term memory
  5. Evaluation: Automated testing and benchmarking

Code Quality

  1. Testing: Unit tests, integration tests
  2. Documentation: API docs, usage examples
  3. Type Safety: Stricter type hints
  4. Configuration: Externalized config files

Related Patterns

  • Pattern 157
  • Pattern 159

Pattern Families: – Reasoning Patterns (001-011) – Multi-Agent Patterns (012-019) – Tool Use Patterns (020-025) – Memory Patterns (026-031) – And many more…

Summary

The Plugin Extension pattern provides a powerful approach to building AI agents with specific capabilities. This implementation demonstrates:

✅ Key Strengths: – Clean, maintainable code structure – LangChain integration for composability – Type hints and documentation – Production-ready patterns

🎯 Best Use Cases: – Applications requiring this pattern’s characteristics – Systems needing structured agent behavior – Production environments requiring reliability

📚 Learning Resources: – LangChain Documentation: https://python.langchain.com/ – Pattern Documentation: See related patterns – OpenAI API Reference: https://platform.openai.com/docs/

File: 158_plugin_extension.py
Total Lines: 640
Pattern Category: AI Agent Design Pattern #158

Generated from actual implementation code
For educational and reference purposes

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *

More posts