Database layer for the VoltX framework. Provides a unified interface for relational databases (Drizzle ORM + Neon Postgres) and vector stores (Pinecone, pgvector, in-memory).

Installation

npm install @voltx/db

Quick Start

Relational Database (Drizzle + Neon)

import { createDB } from "@voltx/db";

const db = createDB({ url: process.env.DATABASE_URL });

// Use Drizzle ORM queries
const users = await db.select().from(usersTable);

Vector Store

import { createVectorStore } from "@voltx/db";

// In-memory (development)
const store = createVectorStore();

// Pinecone (production)
const store = createVectorStore("pinecone", {
  apiKey: process.env.PINECONE_API_KEY,
  indexName: "my-index",
});

// pgvector (self-hosted)
const store = createVectorStore("pgvector", {
  connectionString: process.env.DATABASE_URL,
});

Vector Operations

// Store documents with embeddings
await store.upsert([
  { id: "doc-1", content: "Hello world", embedding: [0.1, 0.2, ...], metadata: {} },
]);

// Search by similarity
const results = await store.search(queryEmbedding, 5);
// → [{ document, score }, ...]

// Delete documents
await store.delete(["doc-1"]);

Supported Backends

| Backend | Type | Use Case | |---------|------|----------| | Neon Postgres | Relational | Production database via Drizzle ORM | | Pinecone | Vector | Managed vector search at scale | | pgvector | Vector | Self-hosted vector search (same Postgres) | | In-memory | Vector | Development and testing |

Schema Helpers

Built-in Drizzle schema for common AI app tables:

import { conversationsTable, messagesTable, documentsTable } from "@voltx/db";

Environment Variables

DATABASE_URL=postgresql://user:[email protected]/dbname?sslmode=require
PINECONE_API_KEY=pc-...
PINECONE_INDEX=voltx-embeddings

Part of VoltX

This package is part of the VoltX framework. See the monorepo for full documentation.

License

MIT — Made by the Promptly AI Team