File size: 14,530 Bytes
fd50325 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 | import faiss
import numpy as np
from pymongo import MongoClient
import os
from dotenv import load_dotenv
import pickle
from typing import List, Dict, Tuple, Optional
import logging
load_dotenv()
# Configure logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
class FAISSIndexManager:
"""Manages FAISS indices for text and visual embeddings"""
def __init__(self, mongo_uri: str, db_name: str = None):
self.mongo_client = MongoClient(mongo_uri)
self.db = self.mongo_client.get_default_database() if not db_name else self.mongo_client[db_name]
# Collection references
self.event_descriptions = self.db.event_description
self.events = self.db.event
# FAISS indices
self.text_index = None
self.visual_index = None
# Index metadata
self.text_index_metadata = {} # Maps FAISS ID to MongoDB document ID
self.visual_index_metadata = {} # Maps FAISS ID to MongoDB document ID
# Embedding dimensions (adjust based on your embedding model)
self.text_embedding_dim = 384 # Common for sentence-transformers
self.visual_embedding_dim = 512 # Common for visual embeddings
# Index file paths
self.text_index_path = "faiss_text_index.bin"
self.visual_index_path = "faiss_visual_index.bin"
self.text_metadata_path = "faiss_text_metadata.pkl"
self.visual_metadata_path = "faiss_visual_metadata.pkl"
self._initialize_indices()
def _initialize_indices(self):
"""Initialize or load existing FAISS indices"""
try:
# Try to load existing indices
if os.path.exists(self.text_index_path) and os.path.exists(self.text_metadata_path):
self._load_text_index()
logger.info("Loaded existing text index")
else:
self._create_text_index()
logger.info("Created new text index")
if os.path.exists(self.visual_index_path) and os.path.exists(self.visual_metadata_path):
self._load_visual_index()
logger.info("Loaded existing visual index")
else:
self._create_visual_index()
logger.info("Created new visual index")
except Exception as e:
logger.error(f"Error initializing indices: {e}")
# Fallback to creating new indices
self._create_text_index()
self._create_visual_index()
def _create_text_index(self):
"""Create a new FAISS index for text embeddings"""
self.text_index = faiss.IndexFlatIP(self.text_embedding_dim) # Inner product for cosine similarity
self.text_index_metadata = {}
self._save_text_index()
def _create_visual_index(self):
"""Create a new FAISS index for visual embeddings"""
self.visual_index = faiss.IndexFlatIP(self.visual_embedding_dim) # Inner product for cosine similarity
self.visual_index_metadata = {}
self._save_visual_index()
def _load_text_index(self):
"""Load text index from disk"""
self.text_index = faiss.read_index(self.text_index_path)
with open(self.text_metadata_path, 'rb') as f:
self.text_index_metadata = pickle.load(f)
def _load_visual_index(self):
"""Load visual index from disk"""
self.visual_index = faiss.read_index(self.visual_index_path)
with open(self.visual_metadata_path, 'rb') as f:
self.visual_index_metadata = pickle.load(f)
def _save_text_index(self):
"""Save text index to disk"""
if self.text_index is not None:
faiss.write_index(self.text_index, self.text_index_path)
with open(self.text_metadata_path, 'wb') as f:
pickle.dump(self.text_index_metadata, f)
def _save_visual_index(self):
"""Save visual index to disk"""
if self.visual_index is not None:
faiss.write_index(self.visual_index, self.visual_index_path)
with open(self.visual_metadata_path, 'wb') as f:
pickle.dump(self.visual_index_metadata, f)
def rebuild_text_index(self):
"""Rebuild text index from MongoDB data"""
logger.info("Rebuilding text index from MongoDB...")
# Create new index
self._create_text_index()
# Fetch all event descriptions with embeddings
cursor = self.event_descriptions.find(
{"text_embedding": {"$exists": True, "$ne": []}},
{"_id": 0, "description_id": 1, "text_embedding": 1}
)
embeddings = []
metadata = {}
for doc in cursor:
embedding = np.array(doc["text_embedding"], dtype=np.float32)
if len(embedding) == self.text_embedding_dim:
faiss_id = len(embeddings)
embeddings.append(embedding)
metadata[faiss_id] = doc["description_id"]
if embeddings:
embeddings_array = np.vstack(embeddings)
self.text_index.add(embeddings_array)
self.text_index_metadata = metadata
self._save_text_index()
logger.info(f"Rebuilt text index with {len(embeddings)} embeddings")
else:
logger.warning("No text embeddings found in MongoDB")
def rebuild_visual_index(self):
"""Rebuild visual index from MongoDB data"""
logger.info("Rebuilding visual index from MongoDB...")
# Create new index
self._create_visual_index()
# Fetch all events with visual embeddings
cursor = self.events.find(
{"visual_embedding": {"$exists": True, "$ne": []}},
{"_id": 0, "event_id": 1, "visual_embedding": 1}
)
embeddings = []
metadata = {}
for doc in cursor:
embedding = np.array(doc["visual_embedding"], dtype=np.float32)
if len(embedding) == self.visual_embedding_dim:
faiss_id = len(embeddings)
embeddings.append(embedding)
metadata[faiss_id] = doc["event_id"]
if embeddings:
embeddings_array = np.vstack(embeddings)
self.visual_index.add(embeddings_array)
self.visual_index_metadata = metadata
self._save_visual_index()
logger.info(f"Rebuilt visual index with {len(embeddings)} embeddings")
else:
logger.warning("No visual embeddings found in MongoDB")
def add_text_embedding(self, description_id: str, embedding: List[float]) -> bool:
"""Add a text embedding to the index"""
try:
embedding_array = np.array(embedding, dtype=np.float32).reshape(1, -1)
if embedding_array.shape[1] != self.text_embedding_dim:
logger.error(f"Text embedding dimension mismatch: expected {self.text_embedding_dim}, got {embedding_array.shape[1]}")
return False
faiss_id = self.text_index.ntotal
self.text_index.add(embedding_array)
self.text_index_metadata[faiss_id] = description_id
self._save_text_index()
logger.info(f"Added text embedding for description_id: {description_id}")
return True
except Exception as e:
logger.error(f"Error adding text embedding: {e}")
return False
def add_visual_embedding(self, event_id: str, embedding: List[float]) -> bool:
"""Add a visual embedding to the index"""
try:
embedding_array = np.array(embedding, dtype=np.float32).reshape(1, -1)
if embedding_array.shape[1] != self.visual_embedding_dim:
logger.error(f"Visual embedding dimension mismatch: expected {self.visual_embedding_dim}, got {embedding_array.shape[1]}")
return False
faiss_id = self.visual_index.ntotal
self.visual_index.add(embedding_array)
self.visual_index_metadata[faiss_id] = event_id
self._save_visual_index()
logger.info(f"Added visual embedding for event_id: {event_id}")
return True
except Exception as e:
logger.error(f"Error adding visual embedding: {e}")
return False
def search_text_embeddings(self, query_embedding: List[float], k: int = 10) -> List[Dict]:
"""Search for similar text embeddings"""
try:
if self.text_index.ntotal == 0:
return []
query_array = np.array(query_embedding, dtype=np.float32).reshape(1, -1)
if query_array.shape[1] != self.text_embedding_dim:
logger.error(f"Query embedding dimension mismatch: expected {self.text_embedding_dim}, got {query_array.shape[1]}")
return []
# Search FAISS
scores, indices = self.text_index.search(query_array, min(k, self.text_index.ntotal))
# Fetch corresponding documents from MongoDB
results = []
for score, idx in zip(scores[0], indices[0]):
if idx in self.text_index_metadata:
description_id = self.text_index_metadata[idx]
doc = self.event_descriptions.find_one(
{"description_id": description_id},
{"_id": 0}
)
if doc:
doc["similarity_score"] = float(score)
results.append(doc)
return results
except Exception as e:
logger.error(f"Error searching text embeddings: {e}")
return []
def search_visual_embeddings(self, query_embedding: List[float], k: int = 10) -> List[Dict]:
"""Search for similar visual embeddings"""
try:
if self.visual_index.ntotal == 0:
return []
query_array = np.array(query_embedding, dtype=np.float32).reshape(1, -1)
if query_array.shape[1] != self.visual_embedding_dim:
logger.error(f"Query embedding dimension mismatch: expected {self.visual_embedding_dim}, got {query_array.shape[1]}")
return []
# Search FAISS
scores, indices = self.visual_index.search(query_array, min(k, self.visual_index.ntotal))
# Fetch corresponding documents from MongoDB
results = []
for score, idx in zip(scores[0], indices[0]):
if idx in self.visual_index_metadata:
event_id = self.visual_index_metadata[idx]
doc = self.events.find_one(
{"event_id": event_id},
{"_id": 0}
)
if doc:
doc["similarity_score"] = float(score)
results.append(doc)
return results
except Exception as e:
logger.error(f"Error searching visual embeddings: {e}")
return []
def get_index_stats(self) -> Dict:
"""Get statistics about the indices"""
return {
"text_index_size": self.text_index.ntotal if self.text_index else 0,
"visual_index_size": self.visual_index.ntotal if self.visual_index else 0,
"text_embedding_dim": self.text_embedding_dim,
"visual_embedding_dim": self.visual_embedding_dim
}
def close(self):
"""Close the index manager and save indices"""
self._save_text_index()
self._save_visual_index()
self.mongo_client.close()
# Global instance
faiss_manager = None
def get_faiss_manager() -> FAISSIndexManager:
"""Get the global FAISS manager instance"""
global faiss_manager
if faiss_manager is None:
mongo_uri = os.getenv("MONGO_URI")
faiss_manager = FAISSIndexManager(mongo_uri)
return faiss_manager
def generate_text_embedding(text: str) -> List[float]:
"""
Generate text embeddings using SentenceTransformer.
Uses all-mpnet-base-v2 for compatibility with NLP search (query_retreival.py).
Model is lazy-loaded and cached on first call.
"""
global _text_embedding_model
if '_text_embedding_model' not in globals() or _text_embedding_model is None:
try:
from sentence_transformers import SentenceTransformer
_text_embedding_model = SentenceTransformer('all-mpnet-base-v2')
logger.info("✅ Loaded SentenceTransformer (all-mpnet-base-v2) for text embeddings")
except Exception as e:
logger.error(f"Failed to load SentenceTransformer: {e}")
# Fallback to deterministic random for graceful degradation
np.random.seed(hash(text) % 2**32)
return np.random.randn(768).astype(np.float32).tolist()
try:
embedding = _text_embedding_model.encode(text, normalize_embeddings=True)
return embedding.astype(np.float32).tolist()
except Exception as e:
logger.error(f"Failed to generate embedding for text: {e}")
np.random.seed(hash(text) % 2**32)
return np.random.randn(768).astype(np.float32).tolist()
# Global model cache
_text_embedding_model = None
def generate_visual_embedding(image_data: bytes = None) -> List[float]:
"""
Placeholder function to generate visual embeddings.
Replace this with your actual visual embedding model.
"""
# For now, return a random embedding of the correct dimension
# In production, use a proper visual embedding model
np.random.seed(42) # Fixed seed for demo
return np.random.randn(512).astype(np.float32).tolist()
|