---
title: "Storage Configuration"
description: "Configure memory, Postgres, SQLite, ORM, or custom GraphRAG storage."
canonical_url: "https://grag.farming-labs.dev/docs/storage-configuration"
markdown_url: "https://grag.farming-labs.dev/docs/storage-configuration.md"
last_updated: "2018-10-20"
---

# Storage Configuration
URL: /docs/storage-configuration
LLM index: /llms.txt
Description: Configure memory, Postgres, SQLite, ORM, or custom GraphRAG storage.

# Storage Configuration

GRAG storage is configured by passing a `GraphRagStore` into `createGraphRagService`.

```ts
import { createGraphRagService } from "@farming-labs/grag";

const grag = createGraphRagService({ store });
```

## Which Database Can I Use?

| Storage                        | Best for                                   | Package surface                                      | Notes                                                                                     |
| ------------------------------ | ------------------------------------------ | ---------------------------------------------------- | ----------------------------------------------------------------------------------------- |
| Memory                         | tests, demos, local prototypes             | `MemoryGraphRagStore`, `createMemoryGraphRagService` | No persistence after process exit.                                                        |
| Postgres                       | production, multi-tenant AI infrastructure | `SqlGraphRagStore` + Kysely                          | Recommended production default. Add `pgvector` later for native vector indexes.           |
| SQLite                         | local apps, desktop tools, small teams     | `SqlGraphRagStore` + Kysely                          | Good simple persistent option.                                                            |
| Any `@farming-labs/orm` driver | portable app storage                       | `OrmGraphRagStore`, `graphRagOrmSchema`              | Use when another service already standardizes on Farming Labs ORM.                        |
| Custom store                   | hosted DBs, existing infra, service APIs   | implement `GraphRagStore`                            | Useful for Supabase, Neon, Turso, PlanetScale-style service boundaries, or internal APIs. |

The current SQL migration helper has explicit dialect names for:

- `postgres`
- `sqlite`

The ORM adapter is the portability layer when you want to reuse the same schema through a Farming Labs ORM driver.

## Memory

Use this in tests and demos:

```ts
import { createMemoryGraphRagService } from "@farming-labs/grag";

const grag = createMemoryGraphRagService();

await grag.ingestTextDocuments({
  title: "Storage note",
  text: "Postgres stores GraphRAG documents, text units, entities, relationships, and communities.",
});
```

## Postgres With Kysely

Install:

```bash
npm install @farming-labs/grag kysely pg
```

Configure:

```ts
import { Kysely, PostgresDialect } from "kysely";
import { Pool } from "pg";
import {
  applyGraphRagMigrations,
  createGraphRagService,
  SqlGraphRagStore,
  type GraphRagSqlDatabase,
} from "@farming-labs/grag";

const db = new Kysely<GraphRagSqlDatabase>({
  dialect: new PostgresDialect({
    pool: new Pool({
      connectionString: process.env.DATABASE_URL,
    }),
  }),
});

await applyGraphRagMigrations(db, "postgres");

const grag = createGraphRagService({
  store: new SqlGraphRagStore({ db }),
});
```

This creates normalized `grag_*` tables plus link tables for related chunks, entities, relationships, communities, claims, reports, and embeddings.

## SQLite With Kysely

Install:

```bash
npm install @farming-labs/grag kysely better-sqlite3
```

Configure:

```ts
import Database from "better-sqlite3";
import { Kysely, SqliteDialect } from "kysely";
import {
  applyGraphRagMigrations,
  createGraphRagService,
  SqlGraphRagStore,
  type GraphRagSqlDatabase,
} from "@farming-labs/grag";

const db = new Kysely<GraphRagSqlDatabase>({
  dialect: new SqliteDialect({
    database: new Database("grag.sqlite"),
  }),
});

await applyGraphRagMigrations(db, "sqlite");

const grag = createGraphRagService({
  store: new SqlGraphRagStore({ db }),
});
```

SQLite is good for local-first tools, desktop previews, eval runners, and small internal demos.

## Farming Labs ORM

Use this when a service already relies on `@farming-labs/orm` and you want GRAG to share that storage system.

```ts
import { createOrm } from "@farming-labs/orm";
import { createGraphRagService } from "@farming-labs/grag";
import { graphRagOrmSchema, OrmGraphRagStore } from "@farming-labs/grag/orm";

const orm = createOrm({
  schema: graphRagOrmSchema,
  driver,
});

const grag = createGraphRagService({
  store: new OrmGraphRagStore({ orm }),
});
```

The ORM schema maps the same graph tables:

- documents and text units;
- entities and relationships;
- covariates/claims;
- communities and reports;
- embeddings;
- join tables for related chunks and graph neighborhoods.

## Custom Store

For a hosted database or internal persistence service, implement `GraphRagStore`.

```ts
import type { GraphRagStore } from "@farming-labs/grag";

export class MyGraphRagStore implements GraphRagStore {
  // implement upsertGraph, getSnapshot, listEntities, listTextUnits, etc.
}
```

This is the right boundary when your app already has tenancy, authorization, audit logs, or a storage API that GRAG should respect.

## Recommended Defaults

- Use memory for tests.
- Use SQLite for local persistent demos.
- Use Postgres for production.
- Use `@farming-labs/orm` when integrating into a service that already uses that ORM.
- Use a custom `GraphRagStore` when storage must go through an internal platform API.

For production Postgres, add later:

- `tenant_id`, `source_id`, and `index_run_id` columns or attributes;
- full-text indexes over text/title/description fields;
- `pgvector` columns beside `vector_json`;
- HNSW indexes for text unit, entity, and community report embeddings;
- retention/versioning for old index runs.

## Sitemap

See the full [sitemap](/sitemap.md) for all pages.
Docs-scoped sitemap: [/docs/sitemap.md](/docs/sitemap.md).
Well-known sitemap: [/.well-known/sitemap.md](/.well-known/sitemap.md).