@kudzujs/core 0.16.13 → 0.16.15

package/MIGRATION_ROADMAP.md

@@ -57,7 +57,7 @@ Syntax compatibility does not mean reproducing React wholesale. Kudzu accepts th
 - stronger JavaScript failure resilience;
 - no remaining material large-route build-scaling weakness.
 
-The completed 0.9 plan defines the cross-framework comparison and release gate. The active application packet is `0.20.3` in the capability release plan below. Benchmark-only feature omission, unmatched accessibility, weighted scores that hide losses, and unrecorded environment differences do not count as proof.
+The completed 0.9 plan defines the cross-framework comparison and release gate. The active application packet is `0.21.1` in the capability release plan below. Benchmark-only feature omission, unmatched accessibility, weighted scores that hide losses, and unrecorded environment differences do not count as proof.
 
 ### 0.10.0 Through 0.21.x: Application Capability Release Train
 

package/PERFORMANCE.md

@@ -1,6 +1,131 @@
 # Performance Records
 
-Reproducibility classes: `npm run benchmark`, `npm run benchmark:keyed`, `npm run benchmark:native`, `npm run benchmark:module-cache`, `npm run benchmark:project-navigation`, `npm run benchmark:project-state`, and `npm run benchmark:source-scale` are maintained in this repository; `npm run benchmark:commerce` is a maintained paired runner over the public external storefront; older excluded-workspace sections are historical provenance only and are not current framework rankings.
+Reproducibility classes: `npm run benchmark`, `npm run benchmark:keyed`, `npm run benchmark:native`, `npm run benchmark:module-cache`, `npm run benchmark:project-navigation`, `npm run benchmark:project-state`, `npm run benchmark:source-scale`, and `npm run benchmark:ai-delivery` are maintained in this repository; `npm run benchmark:commerce` is a maintained paired runner over the public external storefront; older excluded-workspace sections are historical provenance only and are not current framework rankings.
+
+## 0.21.1 Compiler And Route Scale (Partial)
+
+Measured 2026-09-01 on Linux x64 with Node 24.14.0, an Intel i5-9500, six
+logical CPUs, and 31.2 GiB RAM. Each recorded scale uses one warm-up and three
+measured fresh-process samples. The generated topology contains nine imported
+modules plus one page per route and 1,011 lines per route. Fixture generation is
+excluded. These are absolute candidate measurements, not regression claims.
+
+| Routes | Modules / lines | Graph median | TypeScript parse median | Normalize median | Compile/transform median | Render median | Write median | Clean median / RSS | Incremental median / retained-session peak RSS |
+|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
+| 100 | 1,000 / 101,100 | 1,184.4 ms | 1,796.7 ms | 37.9 ms | 2,428.1 ms | 456.5 ms | 70.4 ms | 7,444.3 ms / 672.3 MiB | 1,838.0 ms / 680.9 MiB |
+| 1,000 | 10,000 / 1,011,000 | 12,152.6 ms | 18,717.6 ms | 302.9 ms | 23,681.5 ms | 4,919.4 ms | 521.7 ms | 77,515.9 ms / 3,182.3 MiB | 18,132.1 ms / 3,281.1 MiB |
+
+The exact 100-route runs are source read `[30.5, 30.3, 24.7]`, graph
+`[1273.8, 1138.8, 1184.4]`, parse `[1749.8, 1911.6, 1796.7]`, normalize
+`[37.2, 37.9, 41.4]`, compile/transform `[2428.1, 2828.6, 2381.1]`, render
+`[516.7, 416.9, 456.5]`, write `[95.3, 53.1, 70.4]`, clean build
+`[7444.3, 7269.2, 8156.1]`, and incremental build
+`[1631.9, 1907.7, 1838.0]` ms. Clean-build RSS is
+`[672.3, 658.6, 675.1]` MiB and retained-session peak RSS is
+`[707.1, 673.9, 680.9]` MiB.
+
+The exact 1,000-route runs are source read `[296.7, 315.7, 271.2]`, graph
+`[12152.6, 12308.1, 10011.3]`, parse `[19099.5, 18717.6, 16398.2]`,
+normalize `[259.6, 302.9, 315.8]`, compile/transform
+`[27200.6, 23681.5, 21608.7]`, render `[4919.4, 4454.6, 5238.7]`, write
+`[748.6, 521.7, 487.8]`, clean build `[77515.9, 70196.9, 82980.2]`, and
+incremental build `[22359.7, 18132.1, 17097.2]` ms. Clean-build RSS is
+`[3181.5, 3182.3, 3188.5]` MiB and retained-session peak RSS is
+`[3281.0, 3281.1, 3284.5]` MiB.
+
+Graph time excludes measured TypeScript parsing. Compile/transform excludes
+measured TypeScript parse and Kudzu normalization, but includes the indivisible
+esbuild parse/transform call for proven import-free modules. Render includes
+provisional page materialization; write includes artifact finalization and
+staged-output promotion. Clean build is an end-to-end wall clock and is not the
+sum of the reported phase boundaries.
+
+Every incremental sample recompiles exactly 10 modules, renders one page,
+matches a clean build of the changed source byte-for-byte, rejects invalid
+source in the retained session without changing output, and recovers in that
+same session after source restoration.
+At 100 routes, clean output is 100 files / 21,980 B with digest
+`61525962c70c5e69a975a656781e598cf6665137d648be82f1c3d5e73f0b1b48`;
+the source change produces digest
+`4c7e0f098f3f3236b6e04fa96e8057341c3aa8bdb35a72fe701e78be06160154`.
+At 1,000 routes, clean output is 1,000 files / 221,780 B with digest
+`ccdd2630bcaf9fac4a481b64c69982c341e4798205bb9fbb4cac67d528246635`;
+the source change produces digest
+`024811a5f4598d5025abc57320695503b2d63d0fd6aa54bcf79efde848c2a94e`.
+Both failure probes reject invalid source without changing the prior deploy and
+recover to the clean digest after source restoration.
+
+The default 10,000-route topology would contain 100,000 modules and 10,110,000
+lines. Extrapolating from the measured 1,000-route RSS is not acceptance
+evidence, and this 31.2 GiB host is not used for an unsafe or reduced substitute.
+The packet remains incomplete until the full topology runs on a separately
+provisioned host. Opt-in internal timing adds no semantic primitive, compiler
+pass, runtime concept, public API, deploy file, or browser byte.
+
+## 0.21.0 Functional Parity Matrix
+
+Measured 2026-09-01 on Linux x64 with Node 24.14.0 and required Chrome. The
+300-test suite passes the maintained content, authentication, forms, CRUD,
+shared-data, large-list, overlay, editor, lazy-load, realtime, error,
+accessibility, and navigation browser journeys. Greenfield `/help`, Apache
+Answer and Memos `/public`, editor `/static`, and the direct 10,000-row route
+remain the corresponding static exclusion controls.
+
+The previously benchmark-only 10,000-item pagination decision now runs once as
+ordinary browser acceptance. Enter advances the native control to rows 101-200,
+focus remains on the control, returning to rows 1-100 releases the old row and
+restores fresh input state, the DOM remains bounded to 100 rows, and no browser
+exception occurs. Its deploy contains 19 files totaling 1,900,048 raw / 242,968
+aggregate gzip B with digest
+`1bbaf3479ac90f923deb469c73eb93c0dd7364e7fe906919b167f2f854deda70`.
+The selected pagination route owns 30,143 raw / 11,193 gzip B JavaScript; the
+complete direct 10,000-row HTML control owns 0 B JavaScript. This freezes
+behavior rather than making a new timing claim.
+
+No fixture source, compiler, IR, normalization pass, runtime concept, public API,
+or browser artifact changed. One focused browser acceptance test and one
+artifact-baseline entry were added. `npm run check`, required-Chrome 300/300,
+package smoke, and the artifact inventory pass, so no release version is
+consumed.
+
+## 0.20.5 Tooling Cost Validation
+
+Measured 2026-09-01 on Linux x64 with Node 24.14.0, OpenCode 1.18.25, and
+`openai/gpt-5.6-sol`. Five interleaved attempts per condition repaired the same
+Apache Answer-derived static React Bootstrap breakpoint under one prompt,
+budget, public context, starter, and acceptance digest. Baseline exposed plain
+`build`; tool-assisted exposed structured `build`, `inspect`, and `explain`.
+Both conditions passed 5/5 with the same 234 raw / 181 gzip B HTML artifact,
+98.94% byte-identical source retention, and zero browser JavaScript.
+
+Baseline total-token samples were `[75485, 75459, 66141, 60644, 61570]`: a
+66,141 median, 60,644/75,485 range, and 67,860 failure-inclusive tokens per
+success. Tool-assisted samples were `[75983, 87615, 108588, 75828, 75950]`: a
+75,983 median, 75,828/108,588 range, and 84,793 tokens per success. Structured
+tools therefore cost 16,933 more tokens per success on this task. Monetary cost
+was reported as zero by the subscription provider in every attempt.
+
+Baseline versus tool-assisted medians and ranges were: file reads 2 (1/3)
+versus 2 (1/3), tool calls 8 (5/9) versus 9 (8/12), builds 2 (2/2) versus 2
+(2/3), correction cycles 1 (1/1) versus 1 (1/2), and completion time 45,811 ms
+(43,194/46,390) versus 60,736 ms (52,313/76,963). There were no failed
+attempts. This rejects an AI cost-reduction claim for the structured tool bundle
+on this task; it does not remove the independently useful machine-readable
+commands or claim results for other tasks.
+
+## 0.16.14 Fair AI Delivery Evidence
+
+Measured 2026-09-01 on Linux x64 with Node 24.14.0. The deterministic protocol
+fixture records two interleaved attempts per variant, including one success and
+one retained failure each. Both variants therefore report a 50% fixture success
+rate, 50 nanos median successful cost, and 100 nanos total cost per success when
+failed attempts remain in the numerator. Successful artifacts are one 63 raw /
+66 gzip B HTML file with digest
+`0d22f579d793d0034c967c7ffa08e0b2c377bc36c959c7b4bafab587fd5dda60`.
+This validates evidence accounting only and is not a model-driven framework
+comparison. Compiler passes, runtime concepts, representative deploy output, and
+browser bytes are unchanged. Browser-disabled and required-Chrome suites pass
+297/297 tests, package smoke passes, and the runner adds no dependency.
 
 ## 0.16.13 Route Artifact Explanations
 

package/README.md

@@ -14,7 +14,7 @@ Kudzu compiles ordinary React-shaped TypeScript and TSX into complete static HTM
 
 > Experimental `0.16.x`: the compiler API and supported TSX surface may change.
 
-**Latest release: 0.16.13 - Route artifact explanations.** `kudzu explain --route <route> --json` now traces one exact emitted route from authored source and semantic ownership through its capability family and selected runtime, handler, chunk, style, and Worker bytes without exposing raw IR. Read the [release notes](./RELEASES.md#01613---route-artifact-explanations), open the [release page](https://github.com/kudzujs/kudzu/releases/tag/v0.16.13), or follow the [architecture packet](./docs/next-architecture/README.md).
+**Latest release: 0.16.15 - Production evidence gates.** Kudzu now records compiler phase and retained-session scale evidence, rejects incomplete AI attribution, and runs the 10,000-item pagination decision as ordinary browser acceptance. Read the [release notes](./RELEASES.md#01615---production-evidence-gates), open the [release page](https://github.com/kudzujs/kudzu/releases/tag/v0.16.15), or follow the [architecture packet](./docs/next-architecture/README.md).
 
 - [Documentation](https://kudzujs.cloud/docs)
 - [Installation guide](https://kudzujs.cloud/docs#install)

package/RELEASES.md

@@ -1,5 +1,75 @@
 # Kudzu Releases
 
+## 0.16.15 - Production Evidence Gates
+
+Kudzu 0.16.15 turns compiler scale, large-list parity, and AI attribution gaps
+into maintained executable evidence instead of release-time assumptions.
+
+### Changed in 0.16.15
+
+- Adds opt-in internal source-read, graph, TypeScript parse, normalization,
+  compile/transform, render, and write timing to the existing project session.
+- Extends `npm run benchmark:source-scale` with retained-session incremental
+  timing and RSS, clean changed-source equivalence, output digests, and clean
+  and retained-session failure recovery.
+- Records repeated 100-route / 1,000-module and 1,000-route / 10,000-module
+  measurements. Every incremental sample recompiles 10 modules and renders one
+  page while matching a clean changed-source build byte-for-byte.
+- Keeps the full 10,000-route / 100,000-module run explicitly pending a
+  separately provisioned host; no projection or reduced fixture is called a
+  pass.
+- Promotes the 10,000-item pagination decision to ordinary Chrome acceptance,
+  including Enter navigation, focus retention, bounded 100-row DOM, state
+  release, and a zero-JavaScript direct-list control.
+- Marks missing AI attribution traces incomplete and withholds token and cost
+  per-success claims instead of treating missing data as zero.
+- Records the equal-condition OpenCode tooling trial: both conditions passed
+  5/5, while structured tools used 16,933 more tokens per success, so no AI cost
+  reduction is claimed.
+- Adds no semantic primitive, compiler pass, runtime concept, application API,
+  dependency, deploy artifact, or browser byte.
+- Passes `npm run check`, required-Chrome 300/300, package smoke, tarball
+  inspection, and both npm publish dry-runs on the exact versioned tree.
+- Updates `create-kudzu@0.1.140` to generate projects on
+  `@kudzujs/core@^0.16.15`.
+
+### Upgrade
+
+```sh
+npm install @kudzujs/core@^0.16.15
+```
+
+## 0.16.14 - Fair AI Delivery Evidence
+
+Kudzu 0.16.14 adds a repository-owned protocol and runner for equal-condition,
+fully retained Kudzu and React+Vite delivery attempts.
+
+### Changed in 0.16.14
+
+- Adds `npm run benchmark:ai-delivery`, backed only by Node standard-library
+  process, hash, filesystem, timing, and compression APIs.
+- Pins model revision, pricing, tool permissions, prompt, budgets, adapter,
+  acceptance command, starter trees, public context, and paired attempt order.
+- Runs every attempt in an isolated starter copy, then independently reruns the
+  declared build and framework-neutral acceptance command.
+- Preserves raw adapter/build/acceptance output, final authored source, browser
+  artifacts, hashes, token/cost/tool/file/build metrics, and source retention.
+- Includes deliberate failures in the evidence and total cost-per-success
+  numerator rather than cherry-picking successful attempts.
+- Adds no compiler pass, semantic primitive, runtime concept, application API,
+  dependency, or browser byte.
+- Uses a deterministic fixture only to validate the protocol. It does not claim
+  that Kudzu outperforms React+Vite; measured tooling-cost validation remains the
+  next packet.
+- Updates `create-kudzu@0.1.139` to generate projects on
+  `@kudzujs/core@^0.16.14`.
+
+### Upgrade
+
+```sh
+npm install @kudzujs/core@^0.16.14
+```
+
 ## 0.16.13 - Route Artifact Explanations
 
 Kudzu 0.16.13 lets tools ask why one exact emitted route owns its browser bytes

package/docs/next-architecture/README.md

@@ -15,7 +15,7 @@ The completed `0.9.0` milestone is recorded in [`0.9-semantic-compression.md`](.
 
 [`1.0-large-application-compatibility-audit.md`](./1.0-large-application-compatibility-audit.md) records the first post-0.9 probes against Memos, Apache Answer, and Actual Budget. The audit finds that reduced slices build but whole-application source retention and behavior parity do not yet pass.
 
-[`application-capability-release-plan.md`](./application-capability-release-plan.md) is the authoritative post-0.9 execution queue, currently at `0.20.3`. It assigns one application-capability section to each minor release and one independently accepted evidence packet to each patch release from `0.10.0` through the `1.0.0` gate. It supersedes the provisional 0.10/0.11/0.12 tool-first ordering in the completed 0.9 handoff without rewriting that historical record.
+[`application-capability-release-plan.md`](./application-capability-release-plan.md) is the authoritative post-0.9 execution queue, currently at `0.21.1`. It assigns one application-capability section to each minor release and one independently accepted evidence packet to each patch release from `0.10.0` through the `1.0.0` gate. It supersedes the provisional 0.10/0.11/0.12 tool-first ordering in the completed 0.9 handoff without rewriting that historical record.
 
 ## Required Invariants
 

package/docs/next-architecture/application-capability-release-plan.md

@@ -1881,6 +1881,21 @@ HTML, generated code, captures, and state values.
 - **Stop condition:** canonical compiler source replaces ordinary authored TSX.
 - **Done condition:** each fix has one diagnostic code and explicit preconditions.
 
+Deferred by the evidence gate without production changes or a release. A complete
+repository inventory found no recorded model prompt, model/tool version, raw
+attempt, correction cycle, token/cost trace, or repeated source edit proving any
+automatic fix useful. Existing diagnostics offer migration guidance but do not
+identify a uniquely safe edit; replacing a non-null ref initializer would change
+authored behavior. The existing lazy-state-literal normalization could support a
+narrow exact-range rewrite while preserving comments and output bytes, but no AI
+trial records it as a correction or cost reduction. Adding check/preview/write
+infrastructure now would therefore be speculative and violate the packet purpose.
+Resume this packet only after `0.20.4` records a repeated correction with one
+stable diagnostic code, a unique semantics-preserving replacement, and
+compiler-proven preconditions. No fixture, analyzer, fix command, formatter,
+compiler pass, runtime concept, browser module, package version, or release was
+added.
+
 ### `0.20.4`: AI Delivery Protocol And Runner
 
 - **Purpose:** track fair Kudzu and React+Vite delivery attempts.
@@ -1891,6 +1906,20 @@ HTML, generated code, captures, and state values.
   comparator.
 - **Done condition:** every attempt is reproducible or fully attributable.
 
+Completed by `@kudzujs/core@0.16.14` and `create-kudzu@0.1.139`. The
+repository-owned Node standard-library runner pins the model revision, pricing,
+tools, prompt, budgets, adapter, acceptance command, starter digests, optional
+public context, and paired schedule before creating isolated workspaces. It
+retains raw adapter/build/acceptance streams, final source, browser artifacts,
+hashes, attribution, token/cost/tool/file/build metrics, source retention, and
+both successful and failed attempts. The deterministic two-attempt-per-variant
+fixture proves equal ordering, one retained failure per variant, independent
+final builds, framework-neutral acceptance, and failure-inclusive cost per
+success. It is protocol evidence only and makes no Kudzu versus React+Vite
+delivery claim. No dependency, compiler pass, semantic primitive, runtime
+concept, application API, or browser byte was added. Browser-disabled and
+required-Chrome suites pass 297/297 tests, and package smoke passes.
+
 ### `0.20.5`: Tooling Cost Validation
 
 - **Purpose:** prove that diagnostics/inspect/explain/fix lower total cost per
@@ -1902,6 +1931,19 @@ HTML, generated code, captures, and state values.
 - **Done condition:** useful tools are retained and non-improving tools are
   removed or deferred.
 
+Measured on 2026-09-01 with five interleaved attempts per condition against one
+Apache Answer-derived static Bootstrap correction. Both conditions passed 5/5
+with identical accepted behavior, source retention, and browser artifacts, but
+plain-build baseline cost 67,860 tokens per success while structured
+build/inspect/explain cost 84,793. The structured bundle is therefore deferred
+as an AI default rather than credited with a cost reduction. Existing JSON,
+inspection, and explanation commands remain for their independently accepted
+machine-readable contracts; no speculative `fix` command, compiler pass,
+runtime concept, application API, dependency, or browser byte was added. Exact
+samples, medians, and ranges are recorded in `PERFORMANCE.md`, while the pinned
+protocol reproduces and retains raw attempt evidence under the ignored
+`test-results/ai-delivery/` path.
+
 ## `0.21.x`: Production-Scale Proof
 
 ### `0.21.0`: Functional Parity Matrix
@@ -1913,6 +1955,28 @@ HTML, generated code, captures, and state values.
 - **Done condition:** every claimed capability has a browser journey and static
   exclusion control.
 
+**Result:** closed through existing application semantics without a release.
+The required-Chrome 300-test matrix passes content, authentication, forms, CRUD,
+shared data, large lists, overlays, editor ownership, lazy loading, realtime,
+errors, accessibility, and navigation across the maintained greenfield, Apache
+Answer, Memos, and focused package fixtures. The one material matrix gap was the
+10,000-item pagination decision: its release/remount behavior existed only in an
+optional benchmark. One ordinary browser acceptance now proves Enter-key page
+advance, focus retention, the 101-200 and restored 1-100 ranges, bounded 100-row
+DOM, fresh off-range input state, and no browser exceptions. Greenfield `/help`,
+Answer and Memos `/public`, package `/static` routes, and the complete direct
+10,000-row route provide zero-JavaScript controls.
+
+The artifact inventory now includes the large-list fixture: 19 files,
+1,900,048 raw / 242,968 aggregate gzip bytes, deploy SHA-256
+`1bbaf3479ac90f923deb469c73eb93c0dd7364e7fe906919b167f2f854deda70`.
+Pagination owns 30,143 raw / 11,193 gzip bytes of existing JavaScript while the
+direct static route owns zero. Semantic primitives, IR kinds, compiler passes,
+production compiler/runtime LOC, normalization rules, adapters, runtime
+concepts, public APIs, and browser artifacts change by zero. `npm run check`,
+required-Chrome 300/300, package smoke, focused acceptance, and artifact
+inventory pass. No production feature changed, so no release is consumed.
+
 ### `0.21.1`: Compiler And Route Scale
 
 - **Purpose:** measure clean/incremental compilation and artifact emission.
@@ -1923,6 +1987,33 @@ HTML, generated code, captures, and state values.
 - **Done condition:** repeated measurements and all material regressions are
   explained or fixed.
 
+**Progress:** opt-in internal telemetry now records source reads, reachable
+graph discovery, TypeScript parse, normalize, compile/transform, render, and
+write boundaries without changing deploy output. Compile/transform excludes
+the measured TypeScript parse and Kudzu normalization but includes indivisible
+esbuild parse/transform for proven import-free modules. Render includes
+provisional page materialization, write includes finalization and staged-output
+promotion, and the end-to-end clean build is not their sum. The runner also proves that an
+incremental source edit matches a clean build of the changed source, compares
+paired incremental outputs when a baseline is supplied, names retained-session
+rather than incremental-only peak RSS, and fails the process when either clean
+or retained-session invalid-source recovery changes output. This adds no
+semantic primitive, compiler pass, runtime concept, public API, deploy artifact,
+or browser byte; production compiler/build instrumentation is 35 added and four
+removed lines.
+
+One warm-up and three fresh-process samples pass at 100 routes / 1,000 modules /
+101,100 lines and 1,000 routes / 10,000 modules / 1,011,000 lines. Clean-build
+medians are 7,444.3 and 77,515.9 ms with 672.3 and 3,182.3 MiB peak RSS;
+incremental medians are 1,838.0 and 18,132.1 ms, recompiling exactly 10 modules
+and rendering one page in every sample. Clean and changed digests differ,
+incremental output equals a clean changed-source build, and clean and retained
+failure probes preserve then recover the expected digest. Exact phase samples,
+bytes, and digests are recorded in `PERFORMANCE.md`. The full 10,000-route,
+100,000-module, 10,110,000-line run remains explicitly incomplete pending a
+separately provisioned host; no projection or reduced topology is counted as a
+pass, so this packet does not yet authorize a version or release.
+
 ### `0.21.2`: Browser Performance And Memory
 
 - **Purpose:** protect interaction latency and long-running stability.
@@ -2091,3 +2182,8 @@ release transaction where possible or document and publish a forward-fix patch.
 | `0.20.0` | Released as `0.16.11` | Preserve stable diagnostic schema/codes, authored ranges, human errors, compatibility reuse, and zero browser/output delta. | Patch release retained the `0.16.x` public version line |
 | `0.20.1` | Released as `0.16.12` | Preserve bounded deterministic reachable inventory, first-blocker retention, existing-record reuse, and zero browser/output delta. | Patch release retained the `0.16.x` public version line |
 | `0.20.2` | Released as `0.16.13` | Preserve exact emitted-route lookup, authored provenance, semantic ownership, capability/artifact byte reasons and hashes, bounded deterministic output, structured missing-route diagnostics, and zero-JavaScript explanation. | Patch release retained the `0.16.x` public version line |
+| `0.20.3` | Deferred by evidence gate | Resume only after recorded AI trials prove one repeated correction safe and useful with a stable diagnostic code and compiler-proven preconditions. | No recorded AI trials; no production change or release consumed |
+| `0.20.4` | Released as `0.16.14` | Preserve pinned equal conditions, isolated paired attempts, raw failures, independent acceptance, attribution, failure-inclusive costs, source retention, and browser artifact evidence. | Deterministic fixture validates the protocol only; no framework ranking claimed |
+| `0.20.5` | Closed by measured stop condition | Preserve the negative tooling-cost result and do not present structured tools as an AI cost reduction. | Tool-assisted attempts cost 16,933 more tokens per success; no release consumed |
+| `0.21.0` | Closed by existing semantics | Preserve the complete browser parity matrix, large-list keyboard/release acceptance, and static exclusion controls. | No production change; no release consumed |
+| `0.21.1` | Active; telemetry released as `0.16.15` | Preserve the 100/1,000-route phase, RSS, output, digest, incremental-equivalence, and recovery evidence, then run the full separately provisioned 10,000-route topology. | 10,000-route / 100,000-module run remains pending; no projection is accepted |

package/docs/next-architecture/compiler-current-architecture.md

@@ -1,6 +1,6 @@
 # Current Compiler Architecture
 
-This maps the current `0.16.13` architecture, built on the completed `0.9.0` semantic-compression release and `0.8.23` Goal A compiler foundation. The active application packet is `0.20.3`; file and function names are the stable references, while line numbers are intentionally omitted because later work may still move code.
+This maps the current `0.16.15` architecture, built on the completed `0.9.0` semantic-compression release and `0.8.23` Goal A compiler foundation. The active application packet is `0.21.1`; file and function names are the stable references, while line numbers are intentionally omitted because later work may still move code.
 
 ## Responsibility Map
 

package/docs/next-architecture/large-application-ai-native-roadmap.md

@@ -2,9 +2,9 @@
 
 ## Status
 
-Completed compiler-foundation record and longer-term plan after `0.8.35`. The [`0.9-semantic-compression.md`](./0.9-semantic-compression.md) execution queue is complete. The active queue for current work is [`application-capability-release-plan.md`](./application-capability-release-plan.md) at packet `0.20.3`. This document does not mark any remaining planned capability as supported and does not authorize a React runtime, VDOM, hydration, retained browser component tree, generic rerenderer, public store/query/resource API, SPA router, or islands.
+Completed compiler-foundation record and longer-term plan after `0.8.35`. The [`0.9-semantic-compression.md`](./0.9-semantic-compression.md) execution queue is complete. The active queue for current work is [`application-capability-release-plan.md`](./application-capability-release-plan.md) at packet `0.21.1`. This document does not mark any remaining planned capability as supported and does not authorize a React runtime, VDOM, hydration, retained browser component tree, generic rerenderer, public store/query/resource API, SPA router, or islands.
 
-[`MIGRATION_ROADMAP.md`](../../MIGRATION_ROADMAP.md) remains the product authority for product invariants and fixture-driven feature selection. [`application-capability-release-plan.md`](./application-capability-release-plan.md) is authoritative for current work order and evidence at packet `0.20.3`; this plan retains the completed foundation, deferred program, and long-term production gates. If implementation evidence changes either boundary, update the relevant document before broadening a patch.
+[`MIGRATION_ROADMAP.md`](../../MIGRATION_ROADMAP.md) remains the product authority for product invariants and fixture-driven feature selection. [`application-capability-release-plan.md`](./application-capability-release-plan.md) is authoritative for current work order and evidence at packet `0.21.1`; this plan retains the completed foundation, deferred program, and long-term production gates. If implementation evidence changes either boundary, update the relevant document before broadening a patch.
 
 ## Product Outcome
 
@@ -483,4 +483,4 @@ The first comparison is Kudzu versus React + Vite using the same agent, model, t
 
 ## Immediate Decision
 
-All listed foundation and `0.9` slices are complete. Current work continues at `0.20.3` deterministic normalize and fix under the [`application-capability-release-plan.md`](./application-capability-release-plan.md); keep ResourceIR limited to qualifying independent fixtures, and do not add range ownership, virtualization, optimistic transactions, a public adapter/store API, or a router before evidence justifies them.
+All listed foundation and `0.9` slices are complete. Current work continues at `0.21.1` compiler and route scale under the [`application-capability-release-plan.md`](./application-capability-release-plan.md); keep ResourceIR limited to qualifying independent fixtures, and do not add range ownership, virtualization, optimistic transactions, a public adapter/store API, or a router before evidence justifies them.

package/docs/next-architecture/versioning.md

@@ -1,6 +1,6 @@
 # Planned Version Sequence
 
-This is an execution sequence, not release history. `0.8.16` through `0.8.62` are completed scopes represented by package/release records. The `0.10.0` through `1.0.0` minor/patch sequence is maintained in [`application-capability-release-plan.md`](./application-capability-release-plan.md), currently at `0.20.3`; that plan supersedes the provisional tool-first 0.10/0.11/0.12 ordering in the completed 0.9 handoff.
+This is an execution sequence, not release history. `0.8.16` through `0.8.62` are completed scopes represented by package/release records. The `0.10.0` through `1.0.0` minor/patch sequence is maintained in [`application-capability-release-plan.md`](./application-capability-release-plan.md), currently at `0.21.1`; that plan supersedes the provisional tool-first 0.10/0.11/0.12 ordering in the completed 0.9 handoff.
 
 Keep each patch behavior-preserving and independently reviewable. If a boundary proves inseparable, revise this plan before combining releases; do not silently broaden a patch.
 
@@ -66,7 +66,7 @@ Keep each patch behavior-preserving and independently reviewable. If a boundary
 
 ## Generator Versions
 
-`create-kudzu@0.1.138` retains the explicit install instructions and generates projects with `@kudzujs/core@^0.16.13`.
+`create-kudzu@0.1.140` retains the explicit install instructions and generates projects with `@kudzujs/core@^0.16.15`.
 
 ## Release Boundary
 

package/framework/build.mjs

@@ -60,7 +60,9 @@ export async function buildWithSession(project, { changedFiles, quiet = false, m
     await recoverOutput(outputDirectory, backupOutput)
     await rm(stagedOutput, { recursive: true, force: true })
     const { result, pageCount, behaviorCount, cache, inspectionData, explanation } = await buildInto(project, stagedOutput, { changedFiles, minify, quiet, retainCache, explanationRoute })
+    const promoteStarted = project.timings ? performance.now() : 0
     await promoteOutput(stagedOutput, outputDirectory, backupOutput)
+    addTiming(project.timings, "writeMs", promoteStarted)
     project.buildCache = retainCache ? cache : undefined
     if (!quiet) console.log(`Built ${pageCount} page(s), ${behaviorCount} interactive page(s) into dist/`)
     return explanationRoute ? explanation : inspection ? createInspectionReport(inspectionData) : result
@@ -110,6 +112,7 @@ async function buildInto(project, outputDirectory, { changedFiles, minify, quiet
   await mkdir(workDirectory, { recursive: true })
   await mkdir(outputDirectory, { recursive: true })
 
+  const sourceReadStarted = project.timings ? performance.now() : 0
   const projectFiles = await walk(sourceDirectory)
   const allSourceFiles = projectFiles.filter(file => /\.(?:ts|tsx)$/.test(file) && !file.endsWith(".d.ts")).sort()
   const configuredStyleSources = new Set(configuredStyles.sources.map(style => style.source))
@@ -118,10 +121,13 @@ async function buildInto(project, outputDirectory, { changedFiles, minify, quiet
   const sourceIndex = project.sourceIndex
   for (const file of sourceIndex.keys()) if (file.startsWith(`${sourceDirectory}${sep}`) && !allSourceFileSet.has(file)) sourceIndex.delete(file)
   for (const [file, source] of await Promise.all(allSourceFiles.map(async file => [file, await readFile(file, "utf8")]))) sourceIndex.set(file, source)
+  addTiming(project.timings, "sourceReadMs", sourceReadStarted)
   const pageFiles = allSourceFiles.filter(file => file.startsWith(`${pagesDirectory}${sep}`) && file.endsWith(".tsx"))
   if (!pageFiles.length) throw new Error("No pages found in src/pages/")
+  const graphStarted = project.timings ? performance.now() : 0
   const pageSources = new Map(pageFiles.map(file => [file, new Set(reachableSourceFiles([file], allSourceFileSet, sourceIndex))]))
   const sourceFiles = [...new Set([...pageSources.values()].flatMap(files => [...files]))].sort()
+  addTiming(project.timings, "graphMs", graphStarted)
   const compatibility = createCompatibilityReport(sourceFiles.map(file => ({ file: relative(root, file).replaceAll(sep, "/"), source: sourceIndex.get(file) })))
   await writePrettyJson(join(workDirectory, "kudzu-compatibility.json"), compatibility)
   const sourceFileSet = project.sourceFiles
@@ -193,6 +199,7 @@ async function buildInto(project, outputDirectory, { changedFiles, minify, quiet
   const routePageFiles = new Map()
   let renderedPages = 0
 
+  const renderStarted = project.timings ? performance.now() : 0
   for (const pageFile of pageFiles) {
     const cached = !affectedPages.has(pageFile) ? previous?.pageRenders.get(pageFile) : undefined
     if (cached) {
@@ -328,7 +335,9 @@ async function buildInto(project, outputDirectory, { changedFiles, minify, quiet
       rewrites: rewrites.slice(rewriteOffset)
     })
   }
+  addTiming(project.timings, "renderMs", renderStarted)
 
+  const writeStarted = project.timings ? performance.now() : 0
   for (const group of navigationGroups) for (const route of group.routes) if (!emittedApplicationRoutes.has(route)) throw new Error(`${group.label} route ${JSON.stringify(route)} is not an emitted route`)
   rejectNavigationOverlap(navigationGroups)
   for (const group of navigationGroups) {
@@ -541,6 +550,7 @@ async function buildInto(project, outputDirectory, { changedFiles, minify, quiet
     if (typeof config.afterBuild !== "function") throw new Error("kudzu.config afterBuild must be a function")
     await config.afterBuild({ root, outDir: outputDirectory, sourceDir: sourceDirectory, base, routes: plans.map(plan => plan.route), plans, rewrites: sortedRewrites, artifacts })
   }
+  addTiming(project.timings, "writeMs", writeStarted)
 
   let explanation
   if (explanationRoute) {
@@ -563,6 +573,10 @@ async function buildInto(project, outputDirectory, { changedFiles, minify, quiet
   }
 }
 
+function addTiming(timings, name, started) {
+  if (timings) timings[name] = (timings[name] ?? 0) + performance.now() - started
+}
+
 function affectedPageFiles({ changedFiles, pageFiles, pageSources, previous, sourceDirectory }) {
   if (!previous || changedFiles === undefined) return new Set(pageFiles)
   if (changedFiles.some(file => typeof file !== "string")) return new Set(pageFiles)

package/framework/compiler/project-session.mjs

@@ -5,13 +5,15 @@ import { assetPath } from "./path-helpers.mjs"
 import { createSourceGraph } from "./source-graph.mjs"
 import { createWorkerCompiler } from "./worker-compiler.mjs"
 
-export function createProjectSession(projectRoot = process.cwd(), { counters, sourceIndex = new Map() } = {}) {
+export function createProjectSession(projectRoot = process.cwd(), { counters, sourceIndex = new Map(), timings } = {}) {
   const root = resolve(projectRoot)
   const sourceDirectory = join(root, "src")
   const graph = createSourceGraph(root)
-  const modules = createModuleCache(root, sourceIndex, graph, counters)
+  if (timings && !ts.performance.isEnabled()) ts.performance.enable()
+  const modules = createModuleCache(root, sourceIndex, graph, counters, timings)
   return {
     counters,
+    timings,
     root,
     sourceDirectory,
     pagesDirectory: join(sourceDirectory, "pages"),
@@ -25,7 +27,7 @@ export function createProjectSession(projectRoot = process.cwd(), { counters, so
   }
 }
 
-function createModuleCache(root, sourceIndex, graph, counters) {
+function createModuleCache(root, sourceIndex, graph, counters, timings) {
   const records = new Map()
   let resolvedExports = new WeakMap()
   const moduleName = file => relative(root, file).replaceAll(sep, "/")
@@ -45,7 +47,9 @@ function createModuleCache(root, sourceIndex, graph, counters) {
     if (typeof source !== "string") throw new Error(`Source text is unavailable for ${file}`)
     const cached = records.get(file)
     if (cached?.source === source) return cached
+    const parseBefore = timings ? ts.performance.getDuration("Parse") : 0
     const sourceFile = graph.parseSourceFile(file, source)
+    if (timings) timings.parseMs = (timings.parseMs ?? 0) + ts.performance.getDuration("Parse") - parseBefore
     increment("parsedModules")
     const summary = summarizeModule(file, sourceFile, moduleSymbol, siteId)
     increment("exportSummaries")

package/framework/compiler/source-compiler.mjs

@@ -28,7 +28,7 @@ import { ownedLazyPackageImport } from "./source-graph.mjs"
 import { createZustandPass } from "./zustand-pass.mjs"
 
 export function createSourceCompiler(project) {
-const { root, sourceDirectory, pagesDirectory, workDirectory, workerCompiler, modules, counters } = project
+const { root, sourceDirectory, pagesDirectory, workDirectory, workerCompiler, modules, counters, timings } = project
 const buildDirectory = project.buildDirectory ?? workDirectory
 const { ordinaryRuntimeDependencies, resolveSourceImport, runtimeModuleReference } = project.graph
 const parseSourceFile = (file, source) => modules.read(file, source).sourceFile
@@ -42,6 +42,13 @@ async function compileSourceAsync(file, sourceFiles, sourceIndex, staticFiles, c
 }
 
 function compileSource(file, sourceFiles, sourceIndex, staticFiles, cssModules, base, importFreeOutput) {
+  const parseBefore = timings?.parseMs ?? 0
+  const typeScriptParseBefore = timings ? ts.performance.getDuration("Parse") : 0
+  const finishTiming = () => {
+    if (!timings) return
+    const recordedParse = (timings.parseMs ?? 0) - parseBefore
+    timings.parseMs = (timings.parseMs ?? 0) + Math.max(0, ts.performance.getDuration("Parse") - typeScriptParseBefore - recordedParse)
+  }
   const importedAssets = new Set()
   const source = sourceIndex.get(file)
   const semantic = createSemanticArtifact(relative(root, file).replaceAll(sep, "/"))
@@ -74,6 +81,7 @@ function compileSource(file, sourceFiles, sourceIndex, staticFiles, cssModules,
   const moduleHandlers = moduleIR.handlers.filter(handler => handler.kind === "module-export")
   if (!moduleHandlers.length && !moduleIR.bindings.length) {
     normalizeModulePaths(moduleIR)
+    finishTiming()
     return sourceResult
   }
   const moduleSource = printHandlerModule({ moduleIR, handlerPath })
@@ -85,6 +93,7 @@ function compileSource(file, sourceFiles, sourceIndex, staticFiles, cssModules,
   if (moduleErrors.length) throw new Error(moduleErrors.map(error => ts.flattenDiagnosticMessageText(error.messageText, "\n")).join("\n"))
   normalizeModulePaths(moduleIR)
   sourceResult.handlerModule = { path: handlerPath, code: moduleResult.outputText, hasNativeHandlers: moduleHandlers.some(handler => handler.role === "native"), hasEffects: moduleHandlers.some(handler => handler.role === "effect"), clientImports: moduleIR.clientModules, hasPackageImports: moduleIR.imports.some(entry => entry.package) }
+  finishTiming()
   return sourceResult
 }
 
@@ -358,7 +367,9 @@ function createKudzuTransformer({ semantic, handlerUrl, file, sourceFiles, sourc
     const importedStaticCollections = importedSerializableCollections(sourceFile, file, sourceFiles, sourceIndex, true)
     const importedCollections = new Set(importedStaticCollections.keys())
     const importedFallbackCollections = importedSerializableCollections(sourceFile, file, sourceFiles, sourceIndex)
+    const normalizeStarted = timings ? performance.now() : 0
     const normalized = normalizeCompilerSource(sourceFile, { base, context, file, importedCollections, importedFallbackCollections, importedStaticCollections, sourceFiles, sourceIndex })
+    if (timings) timings.normalizeMs = (timings.normalizeMs ?? 0) + performance.now() - normalizeStarted
     sourceFile = normalized.sourceFile
     const { customHookTimerStates } = normalized
     const bindingIndex = createBindingIndex(sourceFile)
@@ -417,7 +428,9 @@ function createKudzuTransformer({ semantic, handlerUrl, file, sourceFiles, sourc
     const importedSource = target => {
       let result = importedSourceCache.get(target)
       if (!result) {
+        const normalizeStarted = timings ? performance.now() : 0
         result = normalizeCompilerSource(modules.clone(target, context.factory, context), { base, context, file: target, sourceFiles, sourceIndex })
+        if (timings) timings.normalizeMs = (timings.normalizeMs ?? 0) + performance.now() - normalizeStarted
         importedSourceCache.set(target, result)
       }
       return result.sourceFile

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kudzujs/core",
-  "version": "0.16.13",
+  "version": "0.16.15",
   "description": "HTML-first TSX framework with synchronous state semantics and no virtual DOM",
   "type": "module",
   "license": "MIT",
@@ -65,6 +65,7 @@
     "benchmark:project-list-decision": "node test/project-list-decision-performance.mjs",
     "benchmark:project-state": "node test/project-state-performance.mjs",
     "benchmark:source-scale": "node test/source-scale-performance.mjs",
+    "benchmark:ai-delivery": "node test/ai-delivery-runner.mjs",
     "prepublishOnly": "npm run check && npm test",
     "deploy": "wrangler deploy",
     "preview": "wrangler dev"