Measures cold-start latency and steady-state execution latency for the Traverse CLI.
bash benchmarks/run.shResults are written to benchmarks/results/summary.json.
| Scenario | Command | What it captures |
|---|---|---|
| Cold start | traverse-cli expedition execute (first run) |
Capability resolution + WASM load + execution |
| Steady state | traverse-cli expedition execute (repeated) |
Repeated execution latency with trace enabled |
- macOS (Apple Silicon or Intel) or Linux x86-64
- No network calls — all inputs are local fixtures
- No background compilation — binary must be pre-built before measuring
- Results vary with hardware; always record the platform in
summary.json
- Input fixtures are pinned in
benchmarks/fixtures/ run.shrecords the git SHA at measurement time- Results are gitignored — check in
summary.jsonmanually when publishing benchmark data
See benchmarks/results/baseline-reference.md for the Docker comparison methodology.
After running benchmarks/run.sh, check for latency regressions against the checked-in baseline:
bash benchmarks/check-regression.shTo update the baseline after an intentional change: bash benchmarks/update-baseline.sh
docs/benchmarks.md— interpretation guidedocs/why-not-docker.md— when to use Traverse vs Docker