lab-kit 0.5.36

Lab kit 0.5.36 admits client0.5.19 in its optional peer range. Runtime kit behavior is unchanged.

The package passed the normal gated suite, a fresh client19/HF4.17 consumer and strict declarations. A clean export of signed source reproduced the complete archive byte for byte. The attached manifest records exact source, CI, archive hashes and member inventory.

This GitHub package artifact does not upgrade a game or site automatically. Host adoption and live checks remain separate. No npm publication or SLSA attestation is claimed.

Lab kit 0.5.35

@adam0white/lab-kit 0.5.35

Immutable GitHub package artifact prepared from merged source 7605561564998230d2632fb8209f7c6a902d9e09. This is not an npm publication.

The Lab client supports per-instance deferred recording only with preserving initialization. Before explicit opt-in it does not retry pending linked records; unshared feedback sends an unlinked note without the supplied run. Feedback names are scoped per game, leaving the old unscoped value untouched. Kit adds compatibility with client 0.5.18. HF engine stays at 4.17.0.

Existing loaded tabs require refresh and dependent games require explicit adoption. Local controls and hosted PR checks do not establish live recording delivery, game deployment, or player acceptance.

Lab kit 0.5.34

@adam0white/lab-kit 0.5.34

Immutable GitHub package artifact from signed merged source 33ed47e7822abb52064d940807ec44699cce3b26. This is not an npm publication.

Lab-client queue retries now require the record's original game to match the sending game. Older rows recover that identity from their compressed envelope; foreign rows retain their bytes and still count toward the global capacity limit. Kit only permits the corrected client peer version. No HF engine version change.

Deterministic queue, preservation, capacity and package controls passed. No new admission, custody or complete campaign qualification is claimed. Existing loaded tabs require refresh; dependent bundles need explicit adoption.

lab-client 0.5.19

Lab client 0.5.19 removes its executable HF3 clock dependency and uses exact npm HF4.17.0 for chapter projection and portable people. The existing coarse-life record format, quantized clock, bounded opportunities, retained fields and supported refusal behavior remain intact.

Qualification includes the normal gated suite, authentic published-client18 boundary comparisons, retained preparation/admission/claim and pinned-clock recomputation, a fresh HF4-only consumer with strict declarations, and byte-identical packages reproduced from signed source. The attached manifest records the exact source, CI, archive hashes and member inventory.

This artifact does not upgrade a game or site automatically. Consumer adoption, game-host replay and live verification are separate gates. No core HF engine version changes. This is a GitHub package artifact; no npm SDK publication or SLSA attestation is claimed.

Lab client 0.5.18

@adam0white/lab-client 0.5.18

Immutable GitHub package artifact prepared from merged source 7605561564998230d2632fb8209f7c6a902d9e09. This is not an npm publication.

The Lab client supports per-instance deferred recording only with preserving initialization. Before explicit opt-in it does not retry pending linked records; unshared feedback sends an unlinked note without the supplied run. Feedback names are scoped per game, leaving the old unscoped value untouched. Kit adds compatibility with client 0.5.18. HF engine stays at 4.17.0.

Existing loaded tabs require refresh and dependent games require explicit adoption. Local controls and hosted PR checks do not establish live recording delivery, game deployment, or player acceptance.

Lab client 0.5.17

@adam0white/lab-client 0.5.17

Immutable GitHub package artifact from signed merged source 33ed47e7822abb52064d940807ec44699cce3b26. This is not an npm publication.

Lab-client queue retries now require the record's original game to match the sending game. Older rows recover that identity from their compressed envelope; foreign rows retain their bytes and still count toward the global capacity limit. Kit only permits the corrected client peer version. No HF engine version change.

Deterministic queue, preservation, capacity and package controls passed. No new admission, custody or complete campaign qualification is claimed. Existing loaded tabs require refresh; dependent bundles need explicit adoption.

Lab kit 0.5.33

@adam0white/lab-kit 0.5.33

Built by the normal package prepack from signed merged source 843f76fcd6ce6dd1a579ddcb01e316b8e2623964. Signed annotated tag lab-kit-v0.5.33 (8d3c2e221d18f461988ee5713221c58ae507296a) points to that exact source. Main CI42122960371, Bench42122960370 and CodeQL42122960380 passed. Central deployment533643b2 is live at exact source, HF4.17.0 and dirty:false. UPKEEP released the freeze before publication.

Lab client0.5.16 adds play-origin/path support and the reviewed receiver catalog. Lab kit0.5.33 includes the client0.5.16 optional peer. The game engine dependency stays exactly4.17.0; the client retains a separate legacy continuity decoder alias3.0.0.

Actual central Voice and external Murmur reader checks passed at360×690 and1280×800: named per-person skips preserve raw originals, ordinary native/file input/export/reload work, complete captured-prefix host/HF replays pass, and served assets match clean source builds. Synthetic unknown-protocol transport is explicitly isolated. Voice has five roles; Murmur has four. Empty putBatch RPCs occurred; no person envelopes were written. These receipts establish bounded receiver safety, not successful foreign admission, custody, a complete campaign or every writer. DNT observation still saw identity-linking requests.

The manifest retains exact source, tag, dependency, digest and reader-receipt identities. SHA256SUMS covers the tarball and both metadata files. This is a GitHub artifact release; no npm publication or build-provenance attestation is claimed.

Lab client 0.5.16

@adam0white/lab-client 0.5.16

Built by the normal package prepack from signed merged source 843f76fcd6ce6dd1a579ddcb01e316b8e2623964. Signed annotated tag lab-client-v0.5.16 (fa9d4a442491cde21ee9b96ac048e1f5e8e26acb) points to that exact source. Main CI42122960371, Bench42122960370 and CodeQL42122960380 passed. Central deployment533643b2 is live at exact source, HF4.17.0 and dirty:false. UPKEEP released the freeze before publication.

Lab client0.5.16 adds play-origin/path support and the reviewed receiver catalog. Lab kit0.5.33 includes the client0.5.16 optional peer. The game engine dependency stays exactly4.17.0; the client retains a separate legacy continuity decoder alias3.0.0.

Actual central Voice and external Murmur reader checks passed at360×690 and1280×800: named per-person skips preserve raw originals, ordinary native/file input/export/reload work, complete captured-prefix host/HF replays pass, and served assets match clean source builds. Synthetic unknown-protocol transport is explicitly isolated. Voice has five roles; Murmur has four. Empty putBatch RPCs occurred; no person envelopes were written. These receipts establish bounded receiver safety, not successful foreign admission, custody, a complete campaign or every writer. DNT observation still saw identity-linking requests.

The manifest retains exact source, tag, dependency, digest and reader-receipt identities. SHA256SUMS covers the tarball and both metadata files. This is a GitHub artifact release; no npm publication or build-provenance attestation is claimed.

HF 4.17.0

Engine: 4.17.0.

• Add another-person bodily preview conditional on the observer's held fatigue evidence and declared physical loads. The result discloses evidence, neutral priors and two named assumed-body samples, or an explicit unavailable reason. It does not read the target's hidden body.
• This is an assumed-body conditional, not a target prediction, clearance, departure decision, autonomous sleep-onset forecast or claim of predictive accuracy. Evidence and numerical work limits can refuse the projection. No overall CPU bound is claimed. See the SDK contract and port recipe.
• Migration: narrow bodyForecast.kind before reading its fields. Host/self results use conditional-body; another-person results use observer-conditional, whose projection.status must also be narrowed. Older 4.x saves require a fresh start under the exact-engine policy. No legacy continuation or archive conversion is added.

HF 4.16.3

Engine: 4.16.3.

• Avoid copying the retained queue when discovery found no events for the current phase. Discovery and the nonempty removal filter stay unchanged.
• In one instrumented seed-21 aborted 31-day prefix using fresh Node 26.12 Worker isolates, the private prototype was faster in all seven fixed pairs: paired median reduction 1.271188 s; median per-pair relative reduction 9.424452%. Complete native/raw-host values and counts matched in all 17 runs. See the experiment for the method and limits.
• This is a finite prototype result, not browser, full-game, save or two-second qualification. The fixture did not exercise nonempty discovery. Before the version bump, published 4.16.2 and the identity-preserving private prototype both replayed the 777-transaction native checkpoint; that is not a 4.16.3 save.
• Migration: older 4.x saves require a fresh start under the existing exact-engine policy. No legacy continuation, archive repair or retagging is added.