@memnest/store-postgres

Postgres + pgvector store for Memnest:

• Everything lives in its own schema (memnest by default).
• Full-text search on a tsvector column with a GIN index.
• HNSW vector search, with an index created per embedding dimension.
• A job queue claimed with FOR UPDATE SKIP LOCKED, so many workers can share one database.

import { createMemnest } from '@memnest/core';
import { ollamaEmbeddings } from '@memnest/providers';
import { createPostgresStore } from '@memnest/store-postgres';

const store = await createPostgresStore({ connectionString: process.env.MEMNEST_DATABASE_URL, autoMigrate: true });
const memnest = createMemnest({ store, queue: store.jobQueue(), embedder: ollamaEmbeddings({ model: 'nomic-embed-text' }) });

Requires the pgvector extension (pgvector/pgvector:pg16 works out of the box).