Skip to Content
kusaw docsv1.0.0
CLI reference

kusaw bench

Capture a run and replay it as a benchmark

kusaw bench

Capture a run and replay it as a benchmark

Usage

Text
kusaw bench [flags]

Flags

NameShortTypeDefaultDescription
--configstringPath to kusaw.yaml (overrides KUSAW_CONFIG and parent search)
--factorystringfactory name from kusaw list
--help-hboolfalsehelp for bench

Subcommands

kusaw bench add

Capture a finished run as a benchmark task

Text
kusaw bench add <run> [flags]
NameShortTypeDefaultDescription
--checkstringArrayHeld-out check as = (repeatable)
--configstringPath to kusaw.yaml (overrides KUSAW_CONFIG and parent search)
--factorystringfactory name from kusaw list
--help-hboolfalsehelp for add
--namestringBenchmark name (default: the run id)
--originstringunknownTask origin: fresh, private, public, or unknown

Example:

Text
kusaw bench add demo --name demo-bench --origin fresh

kusaw bench add-review

Store a planted-bug review benchmark

Text
kusaw bench add-review <name> [flags]
NameShortTypeDefaultDescription
--basestringBase commit or ref the patch applies to
--bugsstringbugs.yaml with planted bugs
--configstringPath to kusaw.yaml (overrides KUSAW_CONFIG and parent search)
--factorystringfactory name from kusaw list
--help-hboolfalsehelp for add-review
--originstringfreshTask origin: fresh, private, public, or unknown
--patchstringUnified diff of the change under review
--repostringRepo label from kusaw.yaml

Example:

Text
kusaw bench add-review demo --repo demo --base HEAD --patch change.patch --bugs bugs.yaml --origin fresh

kusaw bench cancel

Stop a benchmark batch

Text
kusaw bench cancel <batch> [flags]
NameShortTypeDefaultDescription
--configstringPath to kusaw.yaml (overrides KUSAW_CONFIG and parent search)
--factorystringfactory name from kusaw list
--help-hboolfalsehelp for cancel

Example:

Text
kusaw bench cancel demo-batch

kusaw bench init-recipes

Copy the routing and reviewer-model overlays into the variants directory

Text
kusaw bench init-recipes [flags]
NameShortTypeDefaultDescription
--configstringPath to kusaw.yaml (overrides KUSAW_CONFIG and parent search)
--factorystringfactory name from kusaw list
--help-hboolfalsehelp for init-recipes

Example:

Text
kusaw bench init-recipes

kusaw bench label-finding

Mark a review finding as real or noise

Text
kusaw bench label-finding <batch> <try> <finding-no> real|noise [flags]
NameShortTypeDefaultDescription
--configstringPath to kusaw.yaml (overrides KUSAW_CONFIG and parent search)
--factorystringfactory name from kusaw list
--help-hboolfalsehelp for label-finding
--notestringOptional note stored with the label

Example:

Text
kusaw bench label-finding demo-batch 1 1 real

kusaw bench list

List benchmark tasks

Text
kusaw bench list [flags]
NameShortTypeDefaultDescription
--configstringPath to kusaw.yaml (overrides KUSAW_CONFIG and parent search)
--factorystringfactory name from kusaw list
--help-hboolfalsehelp for list

Example:

Text
kusaw bench list

kusaw bench report

Print a benchmark comparison table

Text
kusaw bench report [batch] [flags]
NameShortTypeDefaultDescription
--configstringPath to kusaw.yaml (overrides KUSAW_CONFIG and parent search)
--factorystringfactory name from kusaw list
--help-hboolfalsehelp for report
--jsonboolfalsePrint JSON
--kint3k for pass@k and pass^k
--markdownboolfalseAlso write docs/benchmarks/-.md
--tasksstringComma-separated benchmark task names
--variantsstringComma-separated variant names

Example:

Text
kusaw bench report demo-batch

kusaw bench run

Queue baseline and variant replays

Text
kusaw bench run <name> [flags]
NameShortTypeDefaultDescription
--budgetfloat6410Batch budget in USD
--configstringPath to kusaw.yaml (overrides KUSAW_CONFIG and parent search)
--factorystringfactory name from kusaw list
--help-hboolfalsehelp for run
--repeatint3Replays of each variant, from 1 to 20
--variantsstringComma-separated variant names, after baseline

Example:

Text
kusaw bench run demo-bench

kusaw bench run-review

Replay a review step against planted bugs

Text
kusaw bench run-review <name> [flags]
NameShortTypeDefaultDescription
--budgetfloat645Batch budget in USD
--configstringPath to kusaw.yaml (overrides KUSAW_CONFIG and parent search)
--count-infoboolfalseCount severity info findings as claimed bugs
--factorystringfactory name from kusaw list
--help-hboolfalsehelp for run-review
--matchstringCustom matcher command (replaces the location matcher)
--repeatint3Replays of each variant, from 1 to 20
--stepstringReview step id to replay
--variantsstringComma-separated variant names, after baseline

Example:

Text
kusaw bench run-review demo --step review-standards --repeat 3 --budget 5.00

kusaw bench set-origin

Set a benchmark task's origin

Text
kusaw bench set-origin <name> <origin> [flags]
NameShortTypeDefaultDescription
--configstringPath to kusaw.yaml (overrides KUSAW_CONFIG and parent search)
--factorystringfactory name from kusaw list
--help-hboolfalsehelp for set-origin

Example:

Text
kusaw bench set-origin demo-bench fresh

kusaw bench show

Show one benchmark task and its batches

Text
kusaw bench show <name> [flags]
NameShortTypeDefaultDescription
--configstringPath to kusaw.yaml (overrides KUSAW_CONFIG and parent search)
--factorystringfactory name from kusaw list
--help-hboolfalsehelp for show

Example:

Text
kusaw bench show demo-bench

Exit codes

CodeNameDescription
0okSuccess.
1errorRuntime error.
2usageUsage or validation error.
3rejectedControl request rejected; the error string is printed verbatim.
4not picked upControl request not picked up within the wait time while serve runs.

Example

Text
kusaw bench list

On this page