Skip to content

I1119 — The budgets are measured, and a regression fails rather than being noticed

Measure what this architecture costs — discovery, reconciliation, the environment entry, the generated artifacts — and hold each to a budget derived from its measured baseline.

BLOCKED wave 7 · p2 · implementation profile · parallel safe

Part of two-folder-integration — A repository that adopts Majordomus gains two directories and two lines, and can give them back.

Blocked. This issue cannot start until I1113 is done. The status is derived from that, not declared.

Objective

Measure what this architecture costs — discovery, reconciliation, the environment entry, the generated artifacts — and hold each to a budget derived from its measured baseline.

Why

A spine every consumer reads is a cost every consumer pays. Measured once and then never again, it grows until somebody works around the tool instead of using it.

Current state

The benchmark infrastructure records distributions per command against a committed baseline. The new paths are not among its targets.

Desired state

Each new path is a benchmark target with a recorded baseline and a threshold that catches a material regression while tolerating ordinary growth.

Scope

  • .ai/repo/benchmarks
  • .ai/repo/policy.yaml

Out of scope

  • A second benchmark system
  • A threshold so tight that ordinary content growth fails the build

Dependencies

What waits on this

Acceptance criteria

  • Discovery, reconciliation, the dry run and the environment entry are benchmark targets with recorded baselines
  • The thresholds are policy data, and the check names the target, both values and the threshold when it fails
  • The environment it was measured on is recorded with each baseline
  • A material regression fails; a byte does not

Validation

  • majordomus bench --check

Evidence required

  • budgets_measured
  • regression_refused

Evidence

None recorded. Every token above needs a command or an artifact behind it before this issue can be completed; narrative is refused.

Risk

Wall-clock thresholds are noisy across machines. The structural counters gate strictly and the timings conservatively, as the existing performance doctrine already decided.

Timeline

started
—
verified
—
completed
—

Those three fields, the evidence above and the state of the dependencies are all the status is made of. There is no status field to disagree with them.

Canonical record: .ai/repo/project/issues/I1119.yaml. Read it back with majordomus plan show I1119.