Batched Ingestion

For write-heavy workloads, enable the async batch queue:

grafomem serve --batch --batch-size 64

Writes are enqueued and flushed in batches, with backpressure when the queue is full. flush() forces a drain. Tune --batch-size and flush interval to your throughput/latency trade-off.

Source: src/aml/server/ingestion.py