Process independent documents or queries in parallel with a concurrency limit

Goal

Reduce latency for I/O-bound batch tasks without overwhelming downstream services.

Steps

1. Confirm each list item can be processed independently without depending on another item’s writes. [parallel]
2. Wrap the per-item function with @node(parallel_worker=True, max_parallel_workers=4); the worker receives one node_input. [parallel]
3. Set a conservative maximum concurrency and verify the input parameter matches the node contract. [parallel]
4. Check output order against input order and test cancellation/error propagation when an item fails. [parallel]

How to verify

Result count and order match input, concurrency stays within the limit, and failures are canceled or reported as configured.

Troubleshooting

If input is missing, check the node parameter name; if downstream throttles, lower concurrency and add a total deadline.