@namitjain.india/agent-memory-postgres

Postgres adapter for @namitjain.india/agent-memory. Provides production-scale persistence with pgvector support for semantic search.

GitHub | Report Bug | Request Feature

Installation

npm install @namitjain.india/agent-memory-postgres pg pgvector

Usage

import { AgentMemory } from "@namitjain.india/agent-memory";
import { PostgresAdapter } from "@namitjain.india/agent-memory-postgres";

const memory = new AgentMemory({
  adapter: new PostgresAdapter({
    connectionString: "postgresql://user:pass@localhost:5432/memory"
  }),
  embedding: {
    embedFn: async (text) => [text.length / 100, 0.5]
  }
});

PostgresAdapter Options

| Option | Type | Default | Description | |--------|------|---------|-------------| | connectionString | string | - | PostgreSQL connection string | | client | PgClientLike | - | Custom pg client/pool instance | | tableName | string | memory_items | Custom table name | | autoCreateExtension | boolean | true | Auto-create vector extension |

Features

• Production Scale: Designed for high-traffic production environments
• pgvector Support: Native vector similarity search using <=> operator
• Connection Pooling: Uses pg Pool for efficient connections
• Type Support: Full support for entries, facts, and summaries
• Schema Auto-Creation: Tables and indexes created automatically
• Async Initialization: Lazy initialization with promise caching

Database Schema

CREATE EXTENSION IF NOT EXISTS vector;

CREATE TABLE memory_items (
  id TEXT PRIMARY KEY,
  kind TEXT NOT NULL,
  session_id TEXT NOT NULL,
  timestamp BIGINT NOT NULL,
  importance REAL NOT NULL DEFAULT 0.5,
  role TEXT,
  content TEXT,
  key_name TEXT,
  value_text TEXT,
  embedding vector,
  metadata JSONB,
  from_timestamp BIGINT,
  to_timestamp BIGINT,
  replaced_entry_ids TEXT[]
);

CREATE INDEX idx_memory_items_session ON memory_items(session_id, timestamp);

Example with Custom Table

const memory = new AgentMemory({
  adapter: new PostgresAdapter({
    connectionString: "postgresql://...",
    tableName: "user_memory", // Custom table name
    autoCreateExtension: true
  }),
  embedding: {
    embedFn: async (text) => [/* embeddings */]
  }
});

Example with Custom Client

import { Pool } from "pg";

const pool = new Pool({ connectionString: "postgresql://..." });

const memory = new AgentMemory({
  adapter: new PostgresAdapter({
    client: pool // Reuse existing connection pool
  }),
  embedding: {
    embedFn: async (text) => [/* embeddings */]
  }
});

Methods

close()

Close the connection pool.

const adapter = new PostgresAdapter({ connectionString: "postgresql://..." });
// ... use adapter ...
await adapter.close();

Requirements

• Node.js >= 18.0.0
• PostgreSQL >= 14.0 (for vector support)
• pg >= 8.0.0
• pgvector >= 0.2.0

Contributing & Collaboration

We welcome contributions, feedback, and feature requests!

• Bug Reports: Open an issue
• Feature Requests: Share it
• Pull Requests: Submit a PR

If this project helps you, please consider starring it on GitHub!

License

MIT