Spaces:
Sleeping
Sleeping
File size: 7,781 Bytes
76bba0b |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 |
"""
RAG (Retrieval Augmented Generation) store for fraud pattern matching
"""
import os
import json
import logging
from typing import List, Dict, Any, Optional
import pickle
from sentence_transformers import SentenceTransformer
import numpy as np
from sklearn.metrics.pairwise import cosine_similarity
logger = logging.getLogger(__name__)
class RAGStore:
"""Simple RAG store using sentence transformers and local file storage"""
def __init__(self, collection_dir: str, model_name: str = "sentence-transformers/all-MiniLM-L6-v2"):
self.collection_dir = collection_dir
self.model_name = model_name
self.embeddings_file = os.path.join(collection_dir, "embeddings.pkl")
self.texts_file = os.path.join(collection_dir, "texts.json")
self.metadata_file = os.path.join(collection_dir, "metadata.json")
os.makedirs(collection_dir, exist_ok=True)
# Initialize sentence transformer
try:
self.encoder = SentenceTransformer(model_name)
logger.info(f"Initialized SentenceTransformer: {model_name}")
except Exception as e:
logger.error(f"Failed to initialize SentenceTransformer: {e}")
self.encoder = None
# Load existing data
self.texts = []
self.metadatas = []
self.embeddings = None
self._load_data()
def _load_data(self):
"""Load existing embeddings, texts, and metadata"""
try:
if os.path.exists(self.texts_file):
with open(self.texts_file, 'r') as f:
self.texts = json.load(f)
if os.path.exists(self.metadata_file):
with open(self.metadata_file, 'r') as f:
self.metadatas = json.load(f)
if os.path.exists(self.embeddings_file):
with open(self.embeddings_file, 'rb') as f:
self.embeddings = pickle.load(f)
logger.info(f"Loaded {len(self.texts)} existing documents")
except Exception as e:
logger.error(f"Error loading RAG data: {e}")
self.texts = []
self.metadatas = []
self.embeddings = None
def _save_data(self):
"""Save embeddings, texts, and metadata to files"""
try:
with open(self.texts_file, 'w') as f:
json.dump(self.texts, f)
with open(self.metadata_file, 'w') as f:
json.dump(self.metadatas, f, default=str)
if self.embeddings is not None:
with open(self.embeddings_file, 'wb') as f:
pickle.dump(self.embeddings, f)
logger.info(f"Saved {len(self.texts)} documents to storage")
except Exception as e:
logger.error(f"Error saving RAG data: {e}")
def add(self, texts: List[str], metadatas: List[Dict[str, Any]]):
"""Add new documents to the RAG store"""
if not self.encoder:
logger.warning("No encoder available, cannot add documents")
return
if len(texts) != len(metadatas):
logger.error("Texts and metadatas must have the same length")
return
try:
# Generate embeddings for new texts
new_embeddings = self.encoder.encode(texts)
# Add to existing data
self.texts.extend(texts)
self.metadatas.extend(metadatas)
if self.embeddings is None:
self.embeddings = new_embeddings
else:
self.embeddings = np.vstack([self.embeddings, new_embeddings])
# Save to disk
self._save_data()
logger.info(f"Added {len(texts)} new documents to RAG store")
except Exception as e:
logger.error(f"Error adding documents to RAG store: {e}")
def query(self, query: str, k: int = 5) -> List[Dict[str, Any]]:
"""Query the RAG store for similar documents"""
if not self.encoder or self.embeddings is None or len(self.texts) == 0:
logger.warning("RAG store is empty or encoder unavailable")
return []
try:
# Encode the query
query_embedding = self.encoder.encode([query])
# Calculate similarities
similarities = cosine_similarity(query_embedding, self.embeddings)[0]
# Get top k results
top_indices = np.argsort(similarities)[::-1][:k]
results = []
for idx in top_indices:
if similarities[idx] > 0.1: # Minimum similarity threshold
results.append({
"text": self.texts[idx],
"metadata": self.metadatas[idx],
"similarity": float(similarities[idx])
})
logger.info(f"Query returned {len(results)} results")
return results
except Exception as e:
logger.error(f"Error querying RAG store: {e}")
return []
def get_stats(self) -> Dict[str, Any]:
"""Get statistics about the RAG store"""
return {
"total_documents": len(self.texts),
"has_embeddings": self.embeddings is not None,
"encoder_available": self.encoder is not None,
"collection_dir": self.collection_dir
}
def clear(self):
"""Clear all data from the RAG store"""
try:
self.texts = []
self.metadatas = []
self.embeddings = None
# Remove files
for file_path in [self.embeddings_file, self.texts_file, self.metadata_file]:
if os.path.exists(file_path):
os.remove(file_path)
logger.info("RAG store cleared")
except Exception as e:
logger.error(f"Error clearing RAG store: {e}")
# Utility functions for fraud-specific RAG queries
def build_fraud_context(transaction_data: Dict[str, Any]) -> str:
"""Build a searchable text representation of transaction data"""
parts = []
if 'amount' in transaction_data:
parts.append(f"amount:{transaction_data['amount']}")
if 'merchant' in transaction_data:
parts.append(f"merchant:{transaction_data['merchant']}")
if 'category' in transaction_data:
parts.append(f"category:{transaction_data['category']}")
if 'description' in transaction_data:
parts.append(f"description:{transaction_data['description']}")
if 'timestamp' in transaction_data:
parts.append(f"time:{transaction_data['timestamp']}")
return " ".join(parts)
def extract_fraud_patterns(rag_results: List[Dict[str, Any]]) -> List[str]:
"""Extract common fraud patterns from RAG results"""
patterns = []
for result in rag_results:
metadata = result.get('metadata', {})
similarity = result.get('similarity', 0)
if similarity > 0.7: # High similarity threshold
if 'merchant' in metadata:
patterns.append(f"Similar merchant: {metadata['merchant']}")
if 'amount' in metadata:
patterns.append(f"Similar amount: ${metadata['amount']}")
if 'category' in metadata:
patterns.append(f"Similar category: {metadata['category']}")
return list(set(patterns)) # Remove duplicates |