payman / src /storage /vector_db.py
satyamdev404's picture
Upload 31 files
e0aa230 verified
raw
history blame
25.6 kB
"""
Vector Database Module
This module is responsible for storing and indexing vector embeddings
for efficient retrieval using Pinecone with complete functionality.
Technology: Pinecone
"""
import logging
import os
import time
import uuid
import hashlib
from datetime import datetime
from typing import Dict, List, Any, Optional, Union
# Import Pinecone and related libraries
try:
import pinecone
from pinecone import Pinecone, ServerlessSpec
except ImportError as e:
logging.warning(f"Pinecone library not installed: {e}")
from utils.error_handler import VectorStorageError, error_handler, ErrorType
class VectorDB:
"""
Stores and indexes vector embeddings for efficient retrieval using Pinecone with full functionality.
Features:
- Complete Pinecone integration
- Index management (create, update, delete)
- Batch upsert operations with optimization
- Advanced similarity search with metadata filtering
- Statistics and monitoring
"""
def __init__(self, config: Optional[Dict[str, Any]] = None):
"""
Initialize the VectorDB with configuration.
Args:
config: Configuration dictionary with Pinecone parameters
"""
self.config = config or {}
self.logger = logging.getLogger(__name__)
# Configuration settings
self.api_key = self.config.get("api_key", os.environ.get("PINECONE_API_KEY"))
self.environment = self.config.get("environment", "us-west1-gcp")
self.index_name = self.config.get("index_name", "rag-ai-index")
self.dimension = self.config.get(
"dimension", 3072
) # ✅ Fixed: Match Gemini embedding dimension
self.metric = self.config.get("metric", "cosine")
self.batch_size = self.config.get("batch_size", 100)
# Performance settings
self.max_metadata_size = self.config.get(
"max_metadata_size", 40960
) # 40KB limit
self.upsert_timeout = self.config.get("upsert_timeout", 60)
self.query_timeout = self.config.get("query_timeout", 30)
# Statistics tracking
self.stats = {
"vectors_stored": 0,
"vectors_queried": 0,
"vectors_deleted": 0,
"batch_operations": 0,
"failed_operations": 0,
"start_time": datetime.now(),
}
# Initialize Pinecone client
self.pc = None
self.index = None
self._initialize_client()
def _initialize_client(self):
"""Initialize Pinecone client and index with validation."""
if not self.api_key:
self.logger.warning(
"No Pinecone API key provided. Vector storage will not be available."
)
return
try:
# Initialize Pinecone client
self.pc = Pinecone(api_key=self.api_key)
# Check if index exists, create if not
self._ensure_index_exists()
# Connect to index
self.index = self.pc.Index(self.index_name)
# Test connection
self._test_connection()
self.logger.info(
f"Pinecone client initialized successfully with index: {self.index_name}"
)
except Exception as e:
self.logger.error(f" Failed to initialize Pinecone client: {str(e)}")
self.pc = None
self.index = None
def _ensure_index_exists(self):
"""Ensure the Pinecone index exists, create if necessary."""
try:
# List existing indexes
existing_indexes = [index.name for index in self.pc.list_indexes()]
if self.index_name not in existing_indexes:
self.logger.info(f"Creating new Pinecone index: {self.index_name}")
# Create index with serverless spec
self.pc.create_index(
name=self.index_name,
dimension=self.dimension,
metric=self.metric,
spec=ServerlessSpec(cloud="aws", region=self.environment),
)
# Wait for index to be ready
self._wait_for_index_ready()
self.logger.info(f"Index {self.index_name} created successfully")
else:
self.logger.info(f"Index {self.index_name} already exists")
except Exception as e:
raise VectorStorageError(f"Failed to ensure index exists: {str(e)}")
def _wait_for_index_ready(self, max_wait_time: int = 300):
"""Wait for index to be ready for operations."""
start_time = time.time()
while time.time() - start_time < max_wait_time:
try:
index_stats = self.pc.describe_index(self.index_name)
if index_stats.status.ready:
self.logger.info(f"Index {self.index_name} is ready")
return
self.logger.info(f"Waiting for index to be ready...")
time.sleep(10)
except Exception as e:
self.logger.warning(f"Error checking index status: {str(e)}")
time.sleep(5)
raise VectorStorageError(
f"Index {self.index_name} not ready after {max_wait_time}s"
)
def _test_connection(self):
"""Test connection to Pinecone index."""
try:
# Get index stats
stats = self.index.describe_index_stats()
self.logger.info(f"Connection test successful. Index stats: {stats}")
except Exception as e:
raise VectorStorageError(f"Connection test failed: {str(e)}")
@error_handler(ErrorType.VECTOR_STORAGE)
def store_embeddings(self, items: List[Dict[str, Any]]) -> bool:
"""
Store embeddings in the vector database with full functionality.
Args:
items: List of dictionaries containing content, metadata, and embeddings
Returns:
True if successful, False otherwise
"""
if not self.index or not items:
self.logger.warning("No index available or empty items list")
return False
# Filter and validate items
valid_items = self._validate_items(items)
if not valid_items:
self.logger.warning("No valid embeddings to store")
return False
self.logger.info(f"Storing {len(valid_items)} embeddings in Pinecone")
start_time = time.time()
try:
# Process in batches
total_batches = (len(valid_items) + self.batch_size - 1) // self.batch_size
successful_batches = 0
for i in range(0, len(valid_items), self.batch_size):
batch_num = (i // self.batch_size) + 1
batch = valid_items[i : i + self.batch_size]
self.logger.info(
f"Processing batch {batch_num}/{total_batches} ({len(batch)} vectors)"
)
success = self._store_batch(batch)
if success:
successful_batches += 1
self.stats["vectors_stored"] += len(batch)
else:
self.stats["failed_operations"] += 1
self.logger.error(f" Batch {batch_num} failed")
# Rate limiting between batches
if i + self.batch_size < len(valid_items):
time.sleep(0.1)
self.stats["batch_operations"] += total_batches
processing_time = time.time() - start_time
success_rate = successful_batches / total_batches * 100
self.logger.info(
f"Storage completed: {successful_batches}/{total_batches} batches successful ({success_rate:.1f}%) in {processing_time:.2f}s"
)
return successful_batches > 0
except Exception as e:
self.stats["failed_operations"] += 1
raise VectorStorageError(f"Failed to store embeddings: {str(e)}")
def _validate_items(self, items: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
"""
Validate and prepare items for storage.
Args:
items: List of items to validate
Returns:
List of valid items
"""
valid_items = []
for i, item in enumerate(items):
try:
# Check required fields
if not isinstance(item, dict):
self.logger.warning(f"Item {i} is not a dictionary")
continue
if "embedding" not in item or not item["embedding"]:
self.logger.warning(f"Item {i} missing embedding")
continue
embedding = item["embedding"]
if not isinstance(embedding, list) or len(embedding) != self.dimension:
self.logger.warning(
f"Item {i} has invalid embedding dimension: {len(embedding)} != {self.dimension}"
)
continue
# Prepare item
processed_item = self._prepare_item_for_storage(item, i)
valid_items.append(processed_item)
except Exception as e:
self.logger.warning(f"Error validating item {i}: {str(e)}")
continue
return valid_items
def _prepare_item_for_storage(
self, item: Dict[str, Any], index: int
) -> Dict[str, Any]:
"""
Prepare item for Pinecone storage.
Args:
item: Item to prepare
index: Item index for ID generation
Returns:
Prepared item
"""
# 🆔 Generate unique ID
item_id = item.get("id")
if not item_id:
# Create ID from content hash + timestamp
content = item.get("content", "")
timestamp = str(int(time.time() * 1000))
content_hash = hashlib.md5(content.encode()).hexdigest()[:8]
item_id = f"doc_{content_hash}_{timestamp}_{index}"
# Prepare metadata
metadata = item.get("metadata", {}).copy()
# Add essential fields to metadata
metadata.update(
{
"content_preview": item.get("content", "")[:500], # First 500 chars
"content_length": len(item.get("content", "")),
"stored_at": datetime.now().isoformat(),
"source": item.get("source", "unknown"),
"document_type": item.get("document_type", "text"),
}
)
# Ensure metadata size limit
metadata = self._truncate_metadata(metadata)
return {"id": item_id, "values": item["embedding"], "metadata": metadata}
def _truncate_metadata(self, metadata: Dict[str, Any]) -> Dict[str, Any]:
"""
Truncate metadata to fit Pinecone size limits.
Args:
metadata: Original metadata
Returns:
Truncated metadata
"""
import json
# 📏 Check current size
metadata_str = json.dumps(metadata, default=str)
if len(metadata_str.encode()) <= self.max_metadata_size:
return metadata
# Truncate large fields
truncated = metadata.copy()
# Truncate text fields progressively
text_fields = ["content_preview", "text", "description", "summary"]
for field in text_fields:
if field in truncated:
while (
len(json.dumps(truncated, default=str).encode())
> self.max_metadata_size
):
current_length = len(str(truncated[field]))
if current_length <= 50:
break
truncated[field] = (
str(truncated[field])[: current_length // 2] + "..."
)
return truncated
def _store_batch(self, batch: List[Dict[str, Any]]) -> bool:
"""
Store a batch of embeddings in Pinecone.
Args:
batch: List of prepared items
Returns:
True if successful
"""
try:
# Upsert vectors to Pinecone
upsert_response = self.index.upsert(
vectors=batch, timeout=self.upsert_timeout
)
# Verify upsert success
if hasattr(upsert_response, "upserted_count"):
expected_count = len(batch)
actual_count = upsert_response.upserted_count
if actual_count != expected_count:
self.logger.warning(
f"Upsert count mismatch: {actual_count}/{expected_count}"
)
return False
self.logger.info(f"Successfully stored batch of {len(batch)} vectors")
return True
except Exception as e:
self.logger.error(f" Error storing batch: {str(e)}")
return False
@error_handler(ErrorType.VECTOR_STORAGE)
def search(
self,
query_embedding: List[float],
top_k: int = 5,
filter: Optional[Dict[str, Any]] = None,
include_metadata: bool = True,
include_values: bool = False,
) -> List[Dict[str, Any]]:
"""
Search for similar vectors with advanced filtering.
Args:
query_embedding: Query vector to search for
top_k: Number of results to return
filter: Optional metadata filter
include_metadata: Whether to include metadata in results
include_values: Whether to include vector values in results
Returns:
List of search results with scores and metadata
"""
if not self.index or not query_embedding:
self.logger.warning("No index available or empty query embedding")
return []
# Validate query embedding
if len(query_embedding) != self.dimension:
raise VectorStorageError(
f"Query embedding dimension {len(query_embedding)} != {self.dimension}"
)
self.logger.info(f"Searching for similar vectors (top_k={top_k})")
start_time = time.time()
try:
# Perform similarity search
search_response = self.index.query(
vector=query_embedding,
top_k=top_k,
filter=filter,
include_metadata=include_metadata,
include_values=include_values,
timeout=self.query_timeout,
)
# Process results
results = []
if hasattr(search_response, "matches"):
for match in search_response.matches:
result = {
"id": match.id,
"score": float(match.score),
}
if include_metadata and hasattr(match, "metadata"):
result["metadata"] = (
dict(match.metadata) if match.metadata else {}
)
if include_values and hasattr(match, "values"):
result["values"] = match.values
results.append(result)
self.stats["vectors_queried"] += len(results)
search_time = time.time() - start_time
self.logger.info(
f"Search completed: {len(results)} results in {search_time:.3f}s"
)
return results
except Exception as e:
self.stats["failed_operations"] += 1
raise VectorStorageError(f"Search failed: {str(e)}")
@error_handler(ErrorType.VECTOR_STORAGE)
def delete(
self,
ids: Optional[List[str]] = None,
filter: Optional[Dict[str, Any]] = None,
delete_all: bool = False,
) -> bool:
"""
Delete vectors from the database.
Args:
ids: Optional list of vector IDs to delete
filter: Optional metadata filter for vectors to delete
delete_all: Whether to delete all vectors
Returns:
True if successful
"""
if not self.index:
self.logger.warning("No index available")
return False
try:
if delete_all:
# Delete all vectors
self.index.delete(delete_all=True)
self.logger.info("Deleted all vectors from index")
self.stats["vectors_deleted"] += 1 # Approximate
elif ids:
# Delete by IDs
self.index.delete(ids=ids)
self.logger.info(f"Deleted {len(ids)} vectors by ID")
self.stats["vectors_deleted"] += len(ids)
elif filter:
# Delete by filter
self.index.delete(filter=filter)
self.logger.info(f"Deleted vectors by filter: {filter}")
self.stats["vectors_deleted"] += 1 # Approximate
else:
self.logger.warning("No deletion criteria provided")
return False
return True
except Exception as e:
self.stats["failed_operations"] += 1
raise VectorStorageError(f"Delete operation failed: {str(e)}")
def get_index_stats(self) -> Dict[str, Any]:
"""
Get comprehensive index statistics.
Returns:
Dictionary with index statistics
"""
if not self.index:
return {}
try:
# Get Pinecone index stats
pinecone_stats = self.index.describe_index_stats()
# Combine with internal stats
runtime = datetime.now() - self.stats["start_time"]
return {
"pinecone_stats": {
"total_vector_count": pinecone_stats.total_vector_count,
"dimension": pinecone_stats.dimension,
"index_fullness": pinecone_stats.index_fullness,
"namespaces": (
dict(pinecone_stats.namespaces)
if pinecone_stats.namespaces
else {}
),
},
"internal_stats": {
**self.stats,
"runtime_seconds": runtime.total_seconds(),
"avg_vectors_per_batch": (
self.stats["vectors_stored"]
/ max(1, self.stats["batch_operations"])
),
"success_rate": (
(
self.stats["batch_operations"]
- self.stats["failed_operations"]
)
/ max(1, self.stats["batch_operations"])
* 100
),
},
"configuration": {
"index_name": self.index_name,
"dimension": self.dimension,
"metric": self.metric,
"batch_size": self.batch_size,
},
}
except Exception as e:
self.logger.error(f" Error getting index stats: {str(e)}")
return {"error": str(e)}
def health_check(self) -> Dict[str, Any]:
"""
Perform health check on the vector database.
Returns:
Health check results
"""
health = {
"status": "unknown",
"timestamp": datetime.now().isoformat(),
"checks": {},
}
try:
# Check API connection
if self.pc:
health["checks"]["api_connection"] = "Connected"
else:
health["checks"]["api_connection"] = " Not connected"
health["status"] = "unhealthy"
return health
# Check index availability
if self.index:
health["checks"]["index_available"] = "Available"
else:
health["checks"]["index_available"] = " Not available"
health["status"] = "unhealthy"
return health
# Test query operation
try:
test_vector = [0.1] * self.dimension
self.index.query(vector=test_vector, top_k=1, timeout=5)
health["checks"]["query_operation"] = "Working"
except Exception as e:
health["checks"]["query_operation"] = f" Failed: {str(e)}"
health["status"] = "degraded"
# Check index stats
try:
stats = self.index.describe_index_stats()
health["checks"]["index_stats"] = f"{stats.total_vector_count} vectors"
except Exception as e:
health["checks"]["index_stats"] = f" Failed: {str(e)}"
# 🎯 Overall status
if health["status"] == "unknown":
health["status"] = "healthy"
except Exception as e:
health["status"] = "unhealthy"
health["error"] = str(e)
return health
def reset_stats(self):
"""Reset internal statistics."""
self.stats = {
"vectors_stored": 0,
"vectors_queried": 0,
"vectors_deleted": 0,
"batch_operations": 0,
"failed_operations": 0,
"start_time": datetime.now(),
}
self.logger.info("Statistics reset")
def get_stats(self) -> Dict[str, Any]:
"""
Get simplified stats for UI display.
Returns:
Dictionary with basic statistics
"""
try:
if not self.index:
return {"total_vectors": 0, "status": "disconnected"}
# Get Pinecone stats
pinecone_stats = self.index.describe_index_stats()
return {
"total_vectors": pinecone_stats.total_vector_count,
"dimension": pinecone_stats.dimension,
"index_fullness": pinecone_stats.index_fullness,
"status": "connected",
}
except Exception as e:
self.logger.warning(f"Could not get stats: {e}")
return {"total_vectors": 0, "status": "error", "error": str(e)}
def get_unique_sources(self) -> List[Dict[str, Any]]:
"""
Get unique sources from stored vectors.
Returns:
List of unique sources with metadata
"""
try:
if not self.index:
return []
# This is a simplified approach - in a real implementation,
# you might want to maintain a separate metadata index
# For now, we'll return mock data based on what might be stored
# Try to get some sample vectors to extract sources
test_vector = [0.1] * self.dimension
results = self.index.query(
vector=test_vector,
top_k=100, # Get more results to find unique sources
include_metadata=True,
)
sources = {}
for match in results.matches:
if hasattr(match, "metadata") and match.metadata:
source = match.metadata.get("source", "Unknown")
if source not in sources:
sources[source] = {
"source": source,
"chunk_count": 1,
"added_date": match.metadata.get("stored_at", "Unknown"),
}
else:
sources[source]["chunk_count"] += 1
return list(sources.values())
except Exception as e:
self.logger.warning(f"Could not get unique sources: {e}")
return []
def list_documents(self) -> List[Dict[str, Any]]:
"""
List all documents in the vector database.
Returns:
List of document information
"""
try:
# Get unique sources and format as documents
sources = self.get_unique_sources()
documents = []
for source_info in sources:
documents.append(
{
"name": source_info["source"],
"chunks": source_info["chunk_count"],
"date": source_info["added_date"],
}
)
return documents
except Exception as e:
self.logger.warning(f"Could not list documents: {e}")
return []