CLI Latency Benchmark — Three Processes vs One
- Published
Overview
A measurement of the wall-clock latency of a guardrailed chat turn: three one-shot CLI processes against a single persistent process. Methodology, raw runs and the plotted result are published rather than summarised.
Key features
- Compares process-per-turn against one persistent process on the same workload
- Wall-clock measurement, repeated runs, results plotted with matplotlib
- Raw data and the script published alongside the chart
Technical details
- GitHub Pages
- MIT
- Python
- matplotlib