senkani eval
Live
Quality gates: bench savings + KB health + regression detection.
Syntax
senkani eval [--update-baseline]
Behavior
Runs BenchmarkTasks.all() through SavingsTestRunner, then layers KBGateComputer health checks and a live-multiplier regression check (default 7-day window, optionally filtered by --agent). Each gate appends a row to the report; the savings regression always fails the run, the rest fail only with --strict. --update-baseline locks the current numbers as the new baseline; --json emits the full report for scripting.
Example
$ senkani eval --strict
Flags
Flag
Description
--update-baseline
Lock in current numbers as the new baseline.
--strict
Nonzero exit on regression.
--json
JSON output.
See also
Source:
Sources/CLI/EvalCommand.swift, Sources/Bench/SavingsTestRunner.swift, Sources/Bench/KBEvalRunner.swift (KBGateComputer)