.gitignore

from transformers import pipeline
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import re
import numpy as np
from typing import Dict, List, Tuple

class GiftRecommender:
    def __init__(self):
        # Initialize NER pipeline for entity extraction
        self.ner = pipeline("ner", model="dbmdz/bert-large-cased-finetuned-conll03-english")
        
        # Initialize sentiment analysis pipeline
        self.sentiment = pipeline("sentiment-analysis")
        
        # Initialize zero-shot classification for interest categorization
        self.zero_shot = pipeline("zero-shot-classification")
        
        # Define interest categories
        self.interest_categories = [
            "art", "music", "sports", "technology", "reading",
            "travel", "cooking", "gaming", "fashion", "outdoor activities"
        ]
        
        # Initialize gift rules database
        self.gift_rules = {
            "art": ["art supplies set", "digital drawing tablet", "museum membership", "art classes"],
            "music": ["wireless headphones", "concert tickets", "vinyl records", "music streaming subscription"],
            "sports": ["fitness tracker", "sports equipment", "team merchandise", "gym membership"],
            "technology": ["smart gadgets", "latest electronics", "tech accessories", "coding courses"],
            "reading": ["e-reader", "book subscription", "rare book editions", "bookstore gift card"],
            "travel": ["travel accessories", "language learning subscription", "travel guides", "luggage"],
            "cooking": ["cooking classes", "premium ingredients set", "kitchen gadgets", "cookbook collection"],
            "gaming": ["gaming console", "gaming accessories", "game subscription", "collectible items"],
            "fashion": ["designer accessories", "fashion subscription box", "custom jewelry", "premium clothing"],
            "outdoor": ["camping gear", "hiking equipment", "outdoor experiences", "adventure gear"]
        }

    def extract_age(self, text: str) -> int:
        """Extract age from text using regex."""
        age_pattern = r'\b(\d{1,2})\s*-?\s*years?\s*-?\s*old\b|\b(\d{1,2})\b'
        matches = re.findall(age_pattern, text)
        if matches:
            # Return the first number found
            age = next(int(num) for nums in matches for num in nums if num)
            return age if 0 < age < 120 else None
        return None

    def extract_gender(self, text: str) -> str:
        """Extract gender from text using keywords."""
        text = text.lower()
        gender_indicators = {
            'female': ['she', 'her', 'sister', 'girlfriend', 'wife', 'daughter', 'mom', 'mother'],
            'male': ['he', 'him', 'brother', 'boyfriend', 'husband', 'son', 'dad', 'father']
        }
        
        for gender, indicators in gender_indicators.items():
            if any(indicator in text for indicator in indicators):
                return gender
        return "unknown"

    def extract_interests(self, text: str) -> List[Dict]:
        """Extract and categorize interests using zero-shot classification."""
        # First, extract potential interest phrases
        interest_pattern = r'loves?\s+([^,.]+)|\blikes?\s+([^,.]+)'
        matches = re.findall(interest_pattern, text.lower())
        
        interests = []
        for match in matches:
            phrase = next(m for m in match if m)
            # Classify the interest into predefined categories
            result = self.zero_shot(
                phrase,
                candidate_labels=self.interest_categories,
                multi_label=False
            )
            
            # Get sentiment score for the interest
            sentiment_score = self.sentiment(phrase)[0]
            
            interests.append({
                'phrase': phrase,
                'category': result['labels'][0],
                'confidence': result['scores'][0],
                'sentiment': sentiment_score['label'],
                'sentiment_score': sentiment_score['score']
            })
        
        return interests

    def extract_dislikes(self, text: str) -> List[str]:
        """Extract dislikes from text."""
        dislike_pattern = r'hates?\s+([^,.]+)|dislikes?\s+([^,.]+)'
        matches = re.findall(dislike_pattern, text.lower())
        return [next(m for m in match if m) for match in matches]

    def get_gift_recommendations(self, text: str) -> Dict:
        """Process text and return gift recommendations."""
        # Extract basic information
        age = self.extract_age(text)
        gender = self.extract_gender(text)
        interests = self.extract_interests(text)
        dislikes = self.extract_dislikes(text)
        
        # Generate recommendations based on interests
        recommendations = []
        for interest in interests:
            category = interest['category']
            if category in self.gift_rules:
                # Weight recommendations by sentiment and confidence
                weight = interest['confidence'] * (2 if interest['sentiment'] == 'POSITIVE' else 1)
                recommendations.extend([
                    {
                        'gift': gift,
                        'category': category,
                        'weight': weight,
                        'reason': f"Based on their interest in {interest['phrase']}"
                    }
                    for gift in self.gift_rules[category]
                ])
        
        # Sort recommendations by weight
        recommendations.sort(key=lambda x: x['weight'], reverse=True)
        
        return {
            'profile': {
                'age': age,
                'gender': gender,
                'interests': interests,
                'dislikes': dislikes
            },
            'recommendations': recommendations[:5]  # Return top 5 recommendations
        }

    def format_recommendations(self, results: Dict) -> str:
        """Format the recommendations into a readable string."""
        output = []
        output.append("🎁 Gift Recommendations\n")
        
        profile = results['profile']
        output.append(f"Profile Summary:")
        output.append(f"- Age: {profile['age'] or 'Unknown'}")
        output.append(f"- Gender: {profile['gender'].title()}")
        output.append("- Interests: " + ", ".join(i['phrase'] for i in profile['interests']))
        if profile['dislikes']:
            output.append("- Dislikes: " + ", ".join(profile['dislikes']))
        
        output.append("\nTop Recommendations:")
        for i, rec in enumerate(results['recommendations'], 1):
            output.append(f"{i}. {rec['gift']}")
            output.append(f"   • {rec['reason']}")
        
        return "\n".join(output)

# Example usage
if __name__ == "__main__":
    recommender = GiftRecommender()
    
    # Example input
    text = """I'm looking for a gift for my 25-year-old sister. 
              She loves painting and traveling, especially in Japan. 
              She hates loud noises and doesn't like spicy food."""
    
    results = recommender.get_gift_recommendations(text)
    formatted_output = recommender.format_recommendations(results)
    print(formatted_output)