Processing API Batches
Overview
Optimize bulk API operations with batch request endpoints, parallel execution with concurrency control, partial failure handling, and progress tracking. Implement batch processing patterns that accept arrays of operations in a single request, execute them efficiently with database bulk operations, and return per-item results with individual success/failure status.
Prerequisites
- Web framework capable of handling large request bodies (configure body size limits: 10MB+ for batch payloads)
- Database with bulk operation support (bulk insert, bulk update, transactions)
- Queue system for async batch processing: Bull/BullMQ (Node.js), Celery (Python), or SQS
- Progress tracking store (Redis) for long-running batch status polling
- Rate limiting aware of batch operations (count individual operations, not just requests)
Instructions
- Examine existing API endpoints using Read and Grep to identify operations frequently called in loops by consumers, which are candidates for batch equivalents.
- Design the batch request format: accept an array of operations in the request body, each with an optional client-provided
id for result correlation, e.g., POST /batch with {operations: [{method: "POST", path: "/users", body: {...}, id: "op1"}]}.
- Implement synchronous batch processing for small batches (< 100 items): validate all items, execute in a database transaction, and return per-item results with
{id, status, result|error} for each operation.
- Add asynchronous batch processing for large batches (> 100 items): accept the batch, return 202 Accepted with a
batchId and status polling URL, process in a background worker, and update progress in Redis.
- Implement concurrency control: process batch items in parallel with configurable concurrency limit (default: 10) using
p-limit or asyncio.Semaphore to prevent database connection exhaustion.
- Handle partial failures: do not abort the entire batch when individual items fail; collect per-item results with success/failure status, and return the batch result with summary counts (
succeeded, failed, total).
- Add progress tracking for async batches: expose
GET /batch/:batchId/status returning {total, completed, failed, progress: 0.75, status: "processing|completed|failed"}.
- Implement batch size limits and validation: maximum 1000 items per batch, reject oversized batches with 413, validate all items before processing any, and return all validation errors upfront.
- Write tests covering: small sync batches, large async batches, partial failure handling, progress tracking, concurrency limits, and batch size validation.
See ${CLAUDESKILLDIR}/references/imp
