| name | helm-bench |
| description | Use when the user wants to run or interpret Helm benchmark commands, compare benchmark reports, or summarize REST API performance regressions and improvements. |
Benchmark
Run REST API benchmarks against the Lando dev environment, save results, and compare with previous runs.
Steps
- Run the benchmark suite:
bun run bench
-
If there is a previous saved report, the output will include a comparison showing % change per endpoint. Regressions (slower) are shown in red, improvements (faster) in green.
-
To compare two specific reports:
bun run bench:compare [baseline.json] [current.json]
- For machine-readable output:
bun run bench --format=json
Interpreting Results
- avg — mean response time across all iterations
- p95 — 95th percentile response time
- ttfb — time to first byte (when headers arrive)
- size — response body size in bytes
Reports are saved to tmp/benchmarks/ with timestamp and git SHA in the filename.
When reporting results to the user, summarize:
- Which endpoints are fastest/slowest
- How embedded responses compare to non-embedded
- Any regressions compared to previous runs (highlight > 10% slower)
- Any notable improvements (highlight > 10% faster)