Synthetic example walkthrough

Run the demo:

uv run python examples/quickstart.py

It generates these files:

• examples/output/comparison.parquet

• examples/output/latency_vs_capacity.png

• examples/output/required_units_distribution.png

• examples/output/scenario_benefit.png

• examples/output/percentile_tradeoff.png

What the fake workload is doing

The synthetic trace contains three request classes:

• chat

• rag

• reasoning

The optimized scenario applies four changes:

• prompt compression

• more caching

• tighter generation caps

• reduced thinking-token budgets

Current synthetic results

examples/output/comparison.parquet is generated directly by the demo and is the authoritative result table. The command also prints that table for quick inspection, so no second set of numbers is maintained in the documentation.

Rendered plots

Latency vs capacity

Distribution of required units

Optimization benefit

Percentile vs slack trade-off

The important pattern is not the exact number. It is that stricter tail planning tends to buy more slack, while prompt/token optimizations can collapse the tail and shrink the reserved-capacity bill.