Find the repeated work.
Remove only that.

The architecture optimization Goal B starts with measured improvements to large keyed restoration and large multi-route compilation without changing Kudzu's HTML-first architecture.

See the results Raw measurements

19.77%Faster keyed restore

6.26%Faster 1,000-product build

172/172Tests passing

GOAL B

Optimize measured work.
Preserve every contract.

KEYED

Batch large restoration

Safe top-level flat lists discover mount hooks once when additions dominate both the list and its parent.

BUILD

Skip no-op AST repair

Normalization repairs parent pointers only when a pass returns a structurally changed source root.

BENCHMARK

Fresh-profile checks

The maintained 2,000-row fixture measures append, filter, restore, and reverse while checking identity, state, and handlers.

EXTERNAL

Real route scale

A public 1,000-product fixture validates 1,011 complete pages across 21 alternating clean builds.

OUTPUT

35 B gzip cost

The keyed route adds 35 B aggregate gzip; normalization changes no emitted bytes.

Repeated route transforms

Identical generated route entries remain the next measured Goal B investigation, not a speculative cache.

UPGRADE

Keep the contracts. Drop the repeated work.

$npm install @kudzujs/core@^0.8.24

Find the repeated work.Remove only that.

Optimize measured work.Preserve every contract.