Skip to main content

MongoDB Atlas Vector Search Configuration

This guide covers detailed configuration options for MongoDB Atlas Vector Search integration with RecoAgent.

Environment Variables​

Required Configuration​

# MongoDB Atlas Connection
MONGODB_URI=mongodb+srv://username:password@cluster.mongodb.net/
MONGODB_DATABASE=recoagent
MONGODB_COLLECTION=documents
MONGODB_VECTOR_SEARCH_INDEX=vector_index

# Vector Store Type
VECTOR_STORE_TYPE=mongodb_atlas
EMBEDDING_DIMENSION=3072

Connection Pool Configuration​

# Connection Pool Settings
MONGODB_MAX_POOL_SIZE=100
MONGODB_MIN_POOL_SIZE=10
MONGODB_MAX_IDLE_TIME_MS=30000
MONGODB_CONNECT_TIMEOUT_MS=10000
MONGODB_SERVER_SELECTION_TIMEOUT_MS=10000

Advanced Configuration​

# Performance Tuning
MONGODB_READ_PREFERENCE=secondaryPreferred
MONGODB_READ_CONCERN=available
MONGODB_WRITE_CONCERN=1

# Retry Configuration
MONGODB_RETRY_READS=true
MONGODB_RETRY_WRITES=true
MONGODB_MAX_RETRY_TIME_MS=30000

Configuration in Code​

Basic Configuration​

from packages.rag.stores import MongoDBAtlasVectorStore

vector_store = MongoDBAtlasVectorStore(
    uri="mongodb+srv://username:password@cluster.mongodb.net/",
    database="recoagent",
    collection="documents",
    vector_search_index="vector_index",
    embedding_dim=3072
)

Advanced Configuration​

vector_store = MongoDBAtlasVectorStore(
    uri=uri,
    database=database,
    collection=collection,
    vector_search_index=index_name,
    embedding_dim=embedding_dim,
    
    # Connection Pool Settings
    max_pool_size=100,
    min_pool_size=10,
    max_idle_time_ms=30000,
    connect_timeout_ms=10000,
    server_selection_timeout_ms=10000,
    
    # Additional Options
    read_preference="secondaryPreferred",
    read_concern="available",
    write_concern=1,
    retry_reads=True,
    retry_writes=True,
    max_retry_time_ms=30000
)

Using Configuration Settings​

from config.settings import get_config

config = get_config()

vector_store = MongoDBAtlasVectorStore(
    uri=config.vector_store.mongodb_uri,
    database=config.vector_store.mongodb_database,
    collection=config.vector_store.mongodb_collection,
    vector_search_index=config.vector_store.mongodb_vector_search_index,
    embedding_dim=config.llm.embedding_dimension,
    max_pool_size=config.vector_store.mongodb_max_pool_size,
    min_pool_size=config.vector_store.mongodb_min_pool_size,
    max_idle_time_ms=config.vector_store.mongodb_max_idle_time_ms,
    connect_timeout_ms=config.vector_store.mongodb_connect_timeout_ms,
    server_selection_timeout_ms=config.vector_store.mongodb_server_selection_timeout_ms
)

Vector Search Index Configuration​

Index Definition​

index_definition = {
    "name": "vector_index",
    "type": "vectorSearch",
    "definition": {
        "fields": [
            {
                "type": "vector",
                "path": "embedding",
                "numDimensions": 3072,
                "similarity": "cosine"
            },
            {
                "type": "filter",
                "path": "metadata"
            }
        ]
    }
}

Similarity Metrics​

Supported similarity metrics:

  • cosine: Cosine similarity (default)
  • dotProduct: Dot product similarity
  • euclidean: Euclidean distance
# Cosine similarity (default)
{
    "type": "vector",
    "path": "embedding",
    "numDimensions": 3072,
    "similarity": "cosine"
}

# Dot product similarity
{
    "type": "vector",
    "path": "embedding",
    "numDimensions": 3072,
    "similarity": "dotProduct"
}

# Euclidean distance
{
    "type": "vector",
    "path": "embedding",
    "numDimensions": 3072,
    "similarity": "euclidean"
}

Text Index for Hybrid Search​

# Create text index
vector_store.create_text_index(['content', 'title', 'description'])

# Custom text index configuration
text_index_definition = {
    "content": "text",
    "title": "text",
    "description": "text",
    "metadata.tags": "text"
}

Production Configuration​

High-Availability Setup​

# Multi-region configuration
uri = "mongodb+srv://username:password@cluster0.abc123.mongodb.net,cluster1.def456.mongodb.net,cluster2.ghi789.mongodb.net/"

vector_store = MongoDBAtlasVectorStore(
    uri=uri,
    database="recoagent",
    collection="documents",
    vector_search_index="vector_index",
    
    # High availability settings
    max_pool_size=200,
    min_pool_size=50,
    max_idle_time_ms=60000,
    connect_timeout_ms=5000,
    server_selection_timeout_ms=5000,
    
    # Read preferences for HA
    read_preference="secondaryPreferred",
    read_concern="majority",
    write_concern=2
)

Performance Optimization​

# High-performance configuration
vector_store = MongoDBAtlasVectorStore(
    uri=uri,
    database="recoagent",
    collection="documents",
    vector_search_index="vector_index",
    
    # Performance settings
    max_pool_size=500,
    min_pool_size=100,
    max_idle_time_ms=300000,  # 5 minutes
    connect_timeout_ms=2000,
    server_selection_timeout_ms=2000,
    
    # Optimized read/write settings
    read_preference="primary",
    read_concern="local",
    write_concern=1,
    
    # Retry settings
    retry_reads=True,
    retry_writes=True,
    max_retry_time_ms=10000
)

Security Configuration​

# Secure configuration
vector_store = MongoDBAtlasVectorStore(
    uri=uri,
    database="recoagent",
    collection="documents",
    vector_search_index="vector_index",
    
    # Security settings
    tls=True,
    tls_allow_invalid_certificates=False,
    tls_allow_invalid_hostnames=False,
    
    # Authentication
    auth_source="admin",
    auth_mechanism="SCRAM-SHA-256",
    
    # Connection security
    max_pool_size=50,
    min_pool_size=5,
    max_idle_time_ms=30000
)

Environment-Specific Configuration​

Development Environment​

# Development settings
MONGODB_URI=mongodb://localhost:27017
MONGODB_DATABASE=recoagent_dev
MONGODB_COLLECTION=documents
MONGODB_VECTOR_SEARCH_INDEX=vector_index_dev
MONGODB_MAX_POOL_SIZE=10
MONGODB_MIN_POOL_SIZE=2

Staging Environment​

# Staging settings
MONGODB_URI=mongodb+srv://username:password@staging-cluster.mongodb.net/
MONGODB_DATABASE=recoagent_staging
MONGODB_COLLECTION=documents
MONGODB_VECTOR_SEARCH_INDEX=vector_index_staging
MONGODB_MAX_POOL_SIZE=50
MONGODB_MIN_POOL_SIZE=10

Production Environment​

# Production settings
MONGODB_URI=mongodb+srv://username:password@production-cluster.mongodb.net/
MONGODB_DATABASE=recoagent
MONGODB_COLLECTION=documents
MONGODB_VECTOR_SEARCH_INDEX=vector_index
MONGODB_MAX_POOL_SIZE=200
MONGODB_MIN_POOL_SIZE=50

Configuration Validation​

Validate Configuration​

from config.settings import validate_config, get_config

# Validate configuration
config = get_config()
issues = validate_config(config)

if issues:
    print("Configuration issues found:")
    for issue in issues:
        print(f"  - {issue}")
else:
    print("Configuration is valid!")

Test Connection​

def test_mongodb_connection():
    """Test MongoDB connection and configuration."""
    try:
        vector_store = MongoDBAtlasVectorStore(uri=uri, database=database)
        
        # Test connection
        stats = vector_store.get_stats()
        print(f"✅ Connected to MongoDB: {stats['database']}")
        
        # Test vector search index
        vector_store._create_vector_search_index()
        print("✅ Vector search index ready")
        
        # Test text index
        vector_store.create_text_index(['content'])
        print("✅ Text index created")
        
        return True
        
    except Exception as e:
        print(f"❌ Connection failed: {e}")
        return False

# Run test
if test_mongodb_connection():
    print("MongoDB configuration is working correctly!")
else:
    print("Please check your MongoDB configuration.")

Troubleshooting Configuration​

Common Configuration Issues​

  1. Invalid URI Format

    Error: Invalid connection string
    Solution: Ensure URI follows format: mongodb+srv://username:password@cluster.mongodb.net/

  2. Authentication Failed

    Error: Authentication failed
    Solution: Check username, password, and database permissions

  3. Network Timeout

    Error: ServerSelectionTimeoutError
    Solution: Check network connectivity and firewall settings

  4. Index Not Found

    Error: Vector search index not found
    Solution: Create index manually or ensure auto-creation is enabled

Configuration Debugging​

import logging

# Enable debug logging
logging.basicConfig(level=logging.DEBUG)
logging.getLogger('pymongo').setLevel(logging.DEBUG)
logging.getLogger('motor').setLevel(logging.DEBUG)

# Test with debug logging
vector_store = MongoDBAtlasVectorStore(uri=uri, database=database)

This configuration guide provides comprehensive setup options for MongoDB Atlas Vector Search integration with RecoAgent.