File size: 7,222 Bytes
d37e0aa
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
import pandas as pd
import joblib
import os
import time
from sqlalchemy import create_engine
from sklearn.metrics.pairwise import cosine_similarity
from urllib.parse import quote_plus
from text_utils import TextProcessor
from functools import lru_cache

# --- CONFIGURATION ---
# For cloud deployment (HF/Production), use DATABASE_URL. 
# Fallback to local construction if not present.
DATABASE_URL = os.getenv("DATABASE_URL")
if not DATABASE_URL:
    DB_USER = os.getenv("DB_USER", "postgres")
    DB_PASSWORD = quote_plus(os.getenv("DB_PASSWORD", "subisu"))
    DB_HOST = os.getenv("DB_HOST", "localhost")
    DB_PORT = os.getenv("DB_PORT", "5432")
    DB_NAME = os.getenv("DB_NAME", "ppd_project_db")
    DB_URI = f'postgresql+psycopg2://{DB_USER}:{DB_PASSWORD}@{DB_HOST}:{DB_PORT}/{DB_NAME}'
else:
    # Ensure URL is compatible with SQLAlchemy if it starts with postgres://
    if DATABASE_URL.startswith("postgres://"):
        DATABASE_URL = DATABASE_URL.replace("postgres://", "postgresql+psycopg2://", 1)
    elif "postgresql://" in DATABASE_URL and "+psycopg2" not in DATABASE_URL:
        DATABASE_URL = DATABASE_URL.replace("postgresql://", "postgresql+psycopg2://", 1)
    DB_URI = DATABASE_URL


class RecommenderCore:
    def __init__(self):
        self.engine = create_engine(DB_URI)
        self.vectorizer = None
        self.tfidf_matrix = None
        self.df = None
        self.load_model()

    def load_model(self):
        try:
            if os.path.exists('vectorizer.pkl') and os.path.exists('tfidf_matrix.pkl'):
                self.vectorizer = joblib.load('vectorizer.pkl')
                self.tfidf_matrix = joblib.load('tfidf_matrix.pkl')
                print("💾 Model Loaded into Memory.")
            
            self.df = pd.read_sql("SELECT * FROM articles WHERE status = 'Approved' ORDER BY article_id", self.engine)
            self.df = self.df.reset_index(drop=True)
            print(f"📚 Indexed {len(self.df)} articles.")
        except Exception as e:
            print(f"Load Error: {e}")

    @lru_cache(maxsize=128)
    def recommend_articles(self, symptoms_text, crisis_level, top_n=5):
        """Modular requirement: Main entry point with caching."""
        if self.df is None or self.vectorizer is None:
            return []

        # 1. Preprocess user query
        query_raw = symptoms_text
        query_norm = TextProcessor.normalize(symptoms_text)
        query_phased = TextProcessor.detect_phrases(query_norm)

        # 2. Filter by Crisis Level (Safety First)
        risk_map = {
            "High": ["High", "Critical", "Moderate", "All"],
            "Moderate": ["Moderate", "Low", "All"],
            "Low": ["Low", "All"]
        }
        allowed = risk_map.get(crisis_level, ["All"])
        
        # Determine the filtered subset
        mask = self.df['risk_level'].apply(
            lambda x: any(level.strip() in allowed for level in str(x).split(','))
        )
        filtered_df = self.df[mask].copy()

        if filtered_df.empty: return []

        # 3. Primary ML Scoring (Cosine Similarity)
        user_vec = self.vectorizer.transform([query_phased])
        all_cos_scores = cosine_similarity(user_vec, self.tfidf_matrix).flatten()
        
        # 4. Final Ranking
        # Correctly align scores using the original dataframe's index
        # SAFETY: Ensure we don't exceed the bounds of the scores array (mismatch protection)
        max_idx = len(all_cos_scores)
        cos_scores_for_filtered = []
        for i in filtered_df.index:
            if i < max_idx:
                cos_scores_for_filtered.append(all_cos_scores[i])
            else:
                cos_scores_for_filtered.append(0.0)
        
        filtered_df['cosine_score'] = cos_scores_for_filtered
        
        # Apply the hybrid ranking engine
        ranked_results = self.apply_ranking(filtered_df, query_raw)

        
        # Format for output
        final_list = ranked_results.head(top_n).to_dict('records')
        
        # 5. Live Fallback if needed
        # Requirement: If results are too few, fetch fresh content
        K = 3
        if len(final_list) < K:
            try:
                from ingestion_service import IngestionService
                service = IngestionService()
                live_arts = service.fetch_from_pubmed(query_raw, limit=K)
                for art in live_arts:
                    if len(final_list) >= top_n: break
                    final_list.append({
                        "article_id": -1,
                        "title": art['title'],
                        "category": "Live Fallback",
                        "format_type": "pubmed",
                        "external_url": art['url'],
                        "content": art['content'],
                        "risk_level": "All"
                    })
                # Background ingestion (optional here, but requested in strategy)
                if live_arts: service.store_articles(live_arts)
            except Exception as e:
                print(f"Fallback error: {e}")

        for item in final_list:
            item['access_type'] = 'External Link' if item.get('format_type') == 'pubmed' else 'Direct Text'
            if 'created_at' in item and item['created_at']:
                item['created_at'] = str(item['created_at'])
            
        return final_list

    def apply_ranking(self, df, raw_query):
        """Modular requirement: Hybrid ranking engine."""
        # Constants for weighting
        SOURCE_WEIGHT = 1.15  # 15% boost for contributor articles
        EXACT_MATCH_BOOST = 0.2
        
        tokens = TextProcessor.normalize(raw_query).split()
        
        now = pd.Timestamp.now()

        def calculate_hybrid_score(row):
            score = row['cosine_score']
            
            # A. Source Weighting (Trusted Contributors)
            if row['format_type'] == 'text':
                score *= SOURCE_WEIGHT
            
            # B. Exact Symptom Overlap Boost
            # Check how many user tokens appear exactly in the normalized title
            norm_title = TextProcessor.normalize(row['title'])
            matches = sum(1 for t in tokens if t in norm_title)
            score += (matches * EXACT_MATCH_BOOST)
            
            # C. Recency Boost (PubMed only, newer is better)
            if row['format_type'] == 'pubmed' and row['created_at']:
                age_days = (now - pd.to_datetime(row['created_at'])).days
                # Decaying boost: max 0.1 for brand new, goes to 0 over 365 days
                recency_boost = max(0, 0.1 * (1 - (min(age_days, 365) / 365)))
                score += recency_boost
                
            return score

        df['final_score'] = df.apply(calculate_hybrid_score, axis=1)
        return df.sort_values(by='final_score', ascending=False)

    def get_article_by_id(self, article_id):
        """Fetches a single article by its ID."""
        if self.df is None: return None
        article = self.df[self.df['article_id'] == article_id]
        return article.iloc[0].to_dict() if not article.empty else None

# Singleton instance to be used by main.py
recommender = RecommenderCore()