An inspectable engineering example

Batchline

See how arrivals become batches, where a bounded queue rejects work, and how the real toy model scores a feature vector.

Shape the traffic

Arrival times, in simulated millisecondsComma-separated, sorted automatically. Up to 80 requests, each 0–1,000 ms.

Batch sizeQueue slotsGather windowModel time

This is a deterministic scheduling simulation, not a running Python HTTP server or a throughput benchmark. Requests arriving at the same time are ordered by input position; a due batch finishes before new arrivals.

Run the toy model

Feature vector1–256 finite JSON numbers. Same weighted mean and stable sigmoid as toy-logistic-v1.

The score is computed here using a JavaScript port checked against the Python model. Model time in the timeline is an assumption, not a measurement.

Follow each event

Time	Event	Waiting	Batch	Completed	Rejected