Overview
Queue long-running AI tasks and publish progress/done/error events.
Problem
Synchronous calls time out and exceed rate limits when enriching thousands of records.
Solution
Push tasks to AiJobs; the cron worker handles backoff, retries, and persists results with usage metadata.
How it works
Split records into batches, enqueue prompts per batch, and consume AI.job.* events to update your UI. Results store type/content and token costs.
Who is this for
Ops
Data Engineering
Developers
Expected outcomes
- No timeouts during bulk operations
- Predictable throughput within provider limits
Key metrics
Batch success rate
Baseline
85 %
Target
99 %
Avg cost per record
Baseline
0.012 USD
Target
0.008 USD
Gallery
Downloads & templates
Case studies
Catalog team enriches 75k SKUs overnight
Queued batches finished with <0.5% failures.
Retail Enterprise NA
Security impact
- Selected fields from CRM/catalog; prompts/responses · PII: none by default
Compliance
- SOC2
Availability & next steps
Pro
Enterprise