Design Patterns

Overview of the three design patterns implemented in ChatbotForCse

Patterns Overview

This project demonstrates the practical application of three core design patterns from the Gang of Four catalog.

Strategy

Behavioral pattern for dynamic algorithm selection

Factory

Creational pattern for OCP-compliant object creation

Singleton

Creational pattern for resource management

---

Why These Patterns?

Strategy Pattern

Problem: Multiple classification algorithms needed with runtime switching

Solution: Encapsulate each algorithm in a separate class implementing a common interface

Benefit: Add new classifiers without modifying existing code

Factory Pattern

Problem: Complex object creation logic scattered throughout codebase

Solution: Centralize creation in factory functions with dictionary registry

Benefit: OCP compliance - extend without modifying

Singleton Pattern

Problem: Expensive database/LLM client initialization on every request

Solution: Single instance with lazy initialization and thread safety

Benefit: Connection pooling, resource efficiency

---

Pattern Interaction

1. User Sends Message

   User sends “bugün menüde ne var?” via WhatsApp

2. Factory Creates Classifier

   get_classifier() returns cached HybridClassifier (Singleton-like caching)

3. Strategy Classifies Intent

   HybridClassifier.classify() returns "dining"

4. Factory Creates Strategy

   get_context_strategy("dining") returns new DiningStrategy instance

5. Strategy Fetches Context

   DiningStrategy.fetch() queries MongoDB (Singleton) and formats data

6. LLM Generates Response

   Gemini generates response with context and sends back to user

---

Sequence Diagram

---

Summary Table

Pattern	Category	Components	Problem Solved
Strategy	Behavioral	IntentClassifier, ContextStrategy	Multiple algorithms
Factory	Creational	get_classifier(), get_context_strategy()	Object creation
Singleton	Creational	MongoDB, OllamaClient	Resource management