@batchactions/state-prisma

Prisma persistence adapter for @batchactions/core. Implements StateStore and DistributedStateStore ports for persisting job state, processed records, and distributed batch coordination to any Prisma-supported database.

Installation

npm install @batchactions/state-prisma @batchactions/core

Setup

1. Add models to your Prisma schema

Use the CLI helper:

npx batchactions-prisma init
# or specify schema path:
npx batchactions-prisma init --schema=prisma/schema.prisma

Or manually copy the models from prisma/batchactions.prisma into your schema.

2. Generate and migrate

npx prisma migrate dev --name add-batchactions
npx prisma generate

3. Use

import { PrismaClient } from '@prisma/client';
import { PrismaStateStore } from '@batchactions/state-prisma';
import { BulkImport, CsvParser, BufferSource } from '@batchactions/import';

const prisma = new PrismaClient();
const stateStore = new PrismaStateStore(prisma);

const importer = new BulkImport({
  schema: {
    fields: [
      { name: 'email', type: 'email', required: true },
      { name: 'name', type: 'string', required: true },
    ],
  },
  source: new BufferSource(new CsvParser(), Buffer.from(csvData)),
  stateStore,
  processor: async (record) => {
    // Save to your database
  },
});

await importer.start();

Tables

The adapter creates three tables:

| Table | Description | |-------|-------------| | batchactions_jobs | Job state, config, batch metadata | | batchactions_records | Individual processed records with status and errors | | batchactions_batches | Distributed batch coordination with optimistic locking |

Prisma compatibility

Works with both Prisma v6 and Prisma v7. The adapter accepts your PrismaClient via dependency injection and uses the standard CRUD API that is identical across versions.

Distributed support

Supports multi-worker distributed processing via DistributedStateStore:

• Atomic batch claiming with optimistic locking (version column)
• Stale batch recovery via reclaimStaleBatches()
• Exactly-once job finalization via tryFinalizeJob()

Limitation

Non-serializable schema fields (customValidator, transform, pattern) are stripped when saving. The consumer must re-inject them when restoring a job.

License

MIT