Skip to content

Database Backends

Amgix supports multiple database backends, allowing you to choose based on your performance, cost, and infrastructure requirements. You can use your existing database (PostgreSQL, MariaDB) or a specialized vector database (Qdrant) for maximum performance.

Qdrant: The Specialized Vector Database

Qdrant is a high-performance vector similarity search engine that serves as the default backend for Amgix.

  • Optimized for Vectors: Purpose-built for efficient vector storage and similarity search.
  • High Performance: Best choice for applications requiring high query throughput and low latency.

SQL Databases: Hybrid Search on Your Existing Infrastructure

Amgix supports hybrid search on traditional SQL databases, offering several advantages:

  • Cost Efficiency: Use databases you already operate and manage.
  • Operational Simplicity: Consolidate data storage within your existing relational database infrastructure.
  • Data Compliance & Governance: Keep your data within your established compliance boundaries and security policies.
  • Sparse Vector Support: Amgix implements sparse vector search directly within SQL, complementing dense vectors where supported.

MariaDB

MariaDB is an open-source relational database that Amgix supports for hybrid search.

Notes: MariaDB 11.7.1+ supports dense vector columns, limited to one dense vector per collection. Sparse vectors (model-based and lexical) are fully supported.

PostgreSQL

PostgreSQL is an open-source object-relational database that Amgix supports for hybrid search.

Notes: Requires pgvector extension for dense vector support. Sparse vectors (model-based and lexical) are fully supported.

Choosing Your Backend

The best backend for your Amgix deployment depends on your specific requirements:

  • Choose Qdrant if:

    • You need maximum performance for vector search.
    • You want a dedicated vector database.
    • You're comfortable managing a separate database instance.

  • Choose a SQL Database (MariaDB, PostgreSQL) if:

    • You want to use your existing database infrastructure.
    • Cost efficiency and operational simplicity are priorities.
    • You prefer to keep data within a relational database.
    • Your primary data is already in SQL.
    • Note on Scale: SQL backends are recommended for modest data volumes. Dedicated vector databases like Qdrant typically offer better scalability for large datasets.

Amgix provides a consistent API across all backends.