SQL-Based Vector Stores
SQL-based vector store components provide vector storage capabilities using traditional SQL databases with vector extensions, combining familiar SQL operations with modern vector search.
PGVector
PGVector provides vector storage and similarity search capabilities using PostgreSQL with the pgvector extension.
Usage
PGVector features:
PostgreSQL integration
SQL-based vector operations
ACID transactions
Complex queries
Mature ecosystem
Inputs
connection_string
Connection String
PostgreSQL connection string
table_name
Table Name
Table for vector storage
vector_column
Vector Column
Column name for vector data
text_column
Text Column
Column name for text content
metadata_columns
Metadata Columns
Additional metadata columns
Outputs
vector_store
Vector Store
Configured PGVector vector store
Redis
Redis provides vector search capabilities through its RediSearch module with vector similarity search.
Usage
Redis vector features:
In-memory performance
Real-time search
Hybrid data structures
Pub/sub messaging
Cluster support
Inputs
host
Host
Redis server host
port
Port
Redis port (default: 6379)
password
Password
Redis authentication password
index_name
Index Name
Vector index name
key_prefix
Key Prefix
Prefix for Redis keys
Outputs
vector_store
Vector Store
Configured Redis vector store
Vector Operations
PGVector Capabilities
Distance Functions:
<->Euclidean distance (L2)<#>Negative inner product<=>Cosine distanceCustom distance functions
Index Types:
IVFFlat: Inverted file with flat compression
HNSW: Hierarchical Navigable Small World graphs
Custom Indexes: Application-specific indexing
Query Examples:
-- Similarity search
SELECT * FROM documents
ORDER BY embedding <-> '[0.1,0.2,0.3]'
LIMIT 10;
-- Threshold filtering
SELECT * FROM documents
WHERE embedding <-> '[0.1,0.2,0.3]' < 0.5;
-- Hybrid search with metadata
SELECT * FROM documents
WHERE category = 'technical'
ORDER BY embedding <-> '[0.1,0.2,0.3]'
LIMIT 5;Redis Vector Features
Search Capabilities:
Vector Similarity: KNN and range queries
Hybrid Search: Combine vector and traditional search
Filtering: Pre and post-filtering options
Aggregation: Search result aggregation
Data Types:
FLOAT32: Standard floating-point vectors
FLOAT64: Double precision vectors
Binary: Binary vector representations
Sparse: Sparse vector support
Advanced Features
PostgreSQL Integration Benefits
ACID Compliance:
Atomicity: All-or-nothing transactions
Consistency: Data integrity constraints
Isolation: Concurrent transaction handling
Durability: Persistent data storage
SQL Capabilities:
Complex Queries: JOIN operations with vector search
Window Functions: Advanced analytical queries
CTEs: Common table expressions for complex logic
Views: Create vector-enabled views
Ecosystem Integration:
ORM Support: Integration with popular ORMs
Backup/Recovery: Standard PostgreSQL tools
Replication: Streaming and logical replication
Extensions: Rich extension ecosystem
Redis Performance Benefits
In-Memory Speed:
Sub-millisecond: Ultra-fast query response
High Throughput: Handle thousands of queries/second
Real-time: Instant updates and searches
Low Latency: Consistent performance
Scalability:
Cluster Mode: Distributed vector storage
Sharding: Automatic data distribution
Replication: Master-slave replication
Persistence: Configurable persistence options
Use Cases
PGVector Applications
Document Management: Full-text and semantic search
E-commerce: Product recommendation systems
Analytics: Complex analytical queries with vectors
Data Warehousing: Integration with existing data infrastructure
Redis Applications
Real-time Recommendations: Instant recommendation engines
Session Storage: User session and preference vectors
Caching: Vector result caching for performance
Pub/Sub Systems: Real-time vector updates
Deployment Options
PostgreSQL Deployment
Self-hosted: On-premise PostgreSQL instances
Cloud Managed: AWS RDS, Google Cloud SQL, Azure Database
Container: Docker and Kubernetes deployments
Serverless: Aurora Serverless and similar services
Redis Deployment
Redis Cloud: Managed Redis service
Self-hosted: On-premise Redis instances
Container: Docker and Kubernetes deployments
Edge: Edge computing deployments
Usage Notes
SQL Familiarity: Leverage existing SQL knowledge and tools
Transaction Support: ACID compliance for critical applications
Performance: Balance between consistency and speed
Integration: Easy integration with existing database infrastructure
Tooling: Use familiar database administration tools
Migration: Smooth migration from traditional databases
Last updated