Skip to main content

PGLite Adapter for ElizaOS

A lightweight PostgreSQL-compatible database adapter for ElizaOS, providing vector search capabilities and efficient data persistence in a local environment.

Features​

  • Embedded PostgreSQL engine for local development and testing
  • Full vector search support via pgvector
  • Built-in fuzzy string matching
  • Automatic schema initialization
  • Memory-efficient caching system
  • Support for multiple embedding providers (OpenAI, Ollama, GaiaNet)
  • Transaction support with automatic rollback

Prerequisites​

  • Node.js 16 or later
  • ElizaOS installation
  • Sufficient disk space for local database storage

Installation​

npm install @elizaos-plugins/adapter-pglite

Configuration​

Add the adapter to your ElizaOS configuration:

{
  "plugins": ["@elizaos-plugins/adapter-pglite"],
  "settings": {
    "PGLITE_DATA_DIR": "/path/to/data/directory" // Required setting
  }
}

Required Environment Variables​

  • PGLITE_DATA_DIR: Path to the directory where PGLite will store its data
    • Use memory:// for in-memory database (useful for testing)

Features in Detail​

The adapter automatically configures vector search based on your embedding provider:

  • OpenAI (1536 dimensions)
  • Ollama (1024 dimensions)
  • GaiaNet (768 dimensions)
  • Default/Other (384 dimensions)

Schema Management​

The adapter automatically manages:

  • Database tables (memories, participants, goals, etc.)
  • Vector indexes for similarity search
  • Cache tables with TTL support
  • Relationship tracking tables

Caching System​

Built-in caching system features:

  • Automatic cache invalidation
  • Support for vector similarity caching
  • Levenshtein distance-based text matching
  • Transaction-safe cache updates

Performance Considerations​

  • Uses optimized vector indexes for fast similarity searches
  • Implements efficient batch processing for large datasets
  • Automatic cleanup of expired cache entries
  • Transaction support for data consistency

FAQ​

Can I use this adapter in production?​

PGLite is designed primarily for development and testing environments. For production use, consider using the full PostgreSQL adapter.

How does the vector search compare to full PostgreSQL?​

PGLite's vector search capabilities are identical to PostgreSQL when using pgvector, but with slightly lower performance due to its embedded nature.

Can I migrate from PGLite to full PostgreSQL?​

Yes, the schema and query structure are identical between PGLite and PostgreSQL, making migration straightforward.

Does it support concurrent connections?​

PGLite supports multiple connections but is optimized for single-process use. For heavy concurrent workloads, use the full PostgreSQL adapter.

How large can the database grow?​

PGLite is limited by available disk space and system memory. For datasets larger than a few GB, consider using the full PostgreSQL adapter.

Can I use custom embedding dimensions?​

Yes, the adapter automatically detects and configures the correct embedding dimensions based on your provider settings.

Is the data persistent?​

Yes, when using a file system path for PGLITE_DATA_DIR. Use memory:// for non-persistent, in-memory storage.

How do I backup the database?​

Simply copy the PGLITE_DATA_DIR directory. All database files are stored there.