@gsep/adapters-storage-postgres

PostgreSQL storage adapter for GSEP (Genomic Self-Evolving Prompts)

Installation

npm install @gsep/core @gsep/adapters-storage-postgres

Usage

import { PGA } from '@gsep/core';
import { ClaudeAdapter } from '@gsep/adapters-llm-anthropic';
import { PostgresAdapter } from '@gsep/adapters-storage-postgres';

const pga = new PGA({
  llm: new ClaudeAdapter({
    apiKey: process.env.ANTHROPIC_API_KEY,
  }),
  storage: new PostgresAdapter({
    connectionString: process.env.DATABASE_URL,
    autoInitialize: true, // Creates tables automatically
  }),
});

await pga.initialize();

const genome = await pga.createGenome({ name: 'my-assistant' });
const response = await genome.chat('Hello!', { userId: 'user123' });

console.log(response);

Configuration

PostgresAdapter(config)

| Option | Type | Default | Description | |--------|------|---------|-------------| | connectionString | string | required | PostgreSQL connection string | | maxConnections | number | 20 | Maximum pool size | | autoInitialize | boolean | true | Auto-create database schema |

Connection String Format

postgresql://user:password@host:port/database

Example:

postgresql://postgres:secret@localhost:5432/pga

Database Schema

The adapter automatically creates the following tables:

• pga_genomes - Genome metadata
• pga_alleles - Gene alleles (Layer 0, 1, 2)
• pga_user_dna - User cognitive profiles
• pga_interactions - Interaction logs
• pga_mutations - Mutation history
• pga_feedback - User feedback signals
• pga_analytics - Pre-aggregated stats

See sql/schema.sql for full schema.

Manual Schema Setup

If you prefer to manage migrations yourself:

const adapter = new PostgresAdapter({
  connectionString: process.env.DATABASE_URL,
  autoInitialize: false, // Don't auto-create tables
});

// Run your own migrations
await runMigrations();

// Then initialize GSEP (will skip schema creation)
await pga.initialize();

Performance

Connection Pooling

The adapter uses pg connection pooling by default:

const adapter = new PostgresAdapter({
  connectionString: process.env.DATABASE_URL,
  maxConnections: 50, // Increase for high load
});

Indexes

All critical queries are indexed:

• Genome lookups: pga_genomes(id)
• Allele selection: pga_alleles(genome_id, layer, gene, status)
• User DNA: pga_user_dna(user_id, genome_id)
• Analytics queries: All have composite indexes

Production Checklist

✅ Use connection pooling (maxConnections: 20+) ✅ Enable SSL for remote databases ✅ Set up database backups ✅ Monitor connection pool usage ✅ Use read replicas for analytics queries ✅ Consider partitioning pga_interactions table for high volume

Troubleshooting

"relation does not exist" error

// Ensure autoInitialize is true
const adapter = new PostgresAdapter({
  connectionString: process.env.DATABASE_URL,
  autoInitialize: true, // ← This creates tables
});

await pga.initialize();

Connection timeout

// Increase max connections
const adapter = new PostgresAdapter({
  connectionString: process.env.DATABASE_URL,
  maxConnections: 50, // Default is 20
});

SSL connection required

postgresql://user:pass@host:5432/db?sslmode=require

License

MIT © Luis Alfredo Velasquez Duran