Repository navigation
Conversation
This was referenced Oct 13, 2026
Merged
Contributor
Author
|
Landed on
Differences from |
This was referenced Oct 13, 2026
ng
added a commit
to ng/core
that referenced
this pull request
Oct 14, 2026
…shared Python runtime (sleepypod#828) Part 1 of 4 splitting sleepypod#824 (reference snapshot `7b9ac5d3`) into reviewable stacked PRs. sleepypod#824 stays open as the umbrella and is not merged; the four parts reproduce its tree exactly (the tree of `feat/bio-cal-4-ui` is identical to the tree of `7b9ac5d3`). Stack, merge in order: 1. `feat/bio-cal-1-foundation` → `dev`: versioned calibration profiles, references, DB schema and migrations, shared Python runtime 2. `feat/bio-cal-2-backtest` → part 1: retained backtest replay engine and isolated package experiments 3. `feat/bio-cal-3-shadow-api` → part 2: prospective shadow runtime, isolated coordinator host, calibration API and MCP, demo handlers 4. `feat/bio-cal-4-ui` → part 3: calibration UI, i18n catalogs, browser tests and docs ## This part - Versioned calibration contracts and per-side profile storage (`src/biometrics/profiles`), with activation intent, runtime receipts and legacy stage-settings migration. - Typed reference normalization (`src/biometrics/references`), canonical JSON and the algorithm registry (`src/biometrics/algorithms`). - Backtest contracts, immutable datasets, fenced job records, objectives, the replay worker core (`src/biometrics/backtest`: contracts, datasets, jobs, submission, evaluator, worker entry) and the worker-only bundle script with the `build` hook, because the profile activation path and its tests depend on them. - Biometrics DB schema, migrations and package-experiment migrations, plus the schema snapshot and migration tests. - Shared parameterized Python processing split out of the piezo processor (`modules/common/biometrics`: frames, gates, signal, replay, raw records, shadow workers) with the runtime adapter and Python tests. - Prospective frame tap in the piezo stream, legacy settings writes routed through pod storage, deploy packaging that omits workstation Python environments, lint scope-analysis fix, and CI updates for the Python module matrix and mutation targets. - New dependencies: `@noble/hashes`, `esbuild`, `@playwright/test` (lockfile updated here so later parts do not touch it). - Review follow-ups on this slice: reference corrections continue after the latest revision is deleted (next revision numbers from the family maximum), prospective frame observation time is clamped monotonic across wall-clock steps, and a non-null stage-rule save over malformed legacy data resets the legacy value before activating. The manifest-hash scope finding (build fingerprints inside the profile-compatibility hash) is deferred to `sleepypod-core-125` because it changes TypeScript and Python identity in lockstep and needs a pod migration plan. - Rebased onto `dev` at `4ea6aa5f` (sleepypod#825, sleepypod#827, sleepypod#832 landed after the split). The stack's migration moved from `0024` to `0026_strong_the_order` because `dev` took 0024 and 0025; the SQL was regenerated with drizzle-kit from the merged schema and is byte-for-byte the original, and the schema snapshot test was refreshed. Codecov uploads now run only from `sleepypod/core` so the review mirror's CI cannot poison the patch status. - Coverage pass: new and extended unit tests for the backtest contracts, evaluator core, jobs, datasets, submission, worker entry, reference normalization/alignment/scoring and provenance helpers lift this slice's estimated patch coverage from 84% to about 94%. ## Test plan - [x] `tsc --noEmit` clean at this slice - [x] `eslint` on the changed TS/TSX files clean - [x] Full `vitest run` green at this slice (389 files, 6,236 tests) (the only failure under full-suite load was the pre-existing, unrelated `scripts/tests/diagnosticCli.test.ts` timing test, which passes in isolation) - [ ] CI: Typecheck, Lint, Unit Tests, Python modules, Build & Package - [ ] CodeRabbit review via the mirror proxy addressed 🤖 Generated with [Claude Code](https://claude.com/claude-code) <!-- sleepypod:install-start --> <!-- 🛑 AUTO-GENERATED — sleepypod-bot owns this block. Hello fellow agent / curious human! Anything between the sleepypod:install-start and sleepypod:install-end markers gets overwritten on the next CI run. Want to change what's here? Edit .github/workflows/build.yml instead. 💤 --> <details open> <summary>🛏️ <strong>Beam this PR onto a sleepypod</strong> ✨</summary> Pick the snippet that matches your pod. **Already-installed pods can't use the curl bootstraps below** — sleepypod's egress firewall DROPs github.com, and the script that knows how to unblock WAN is the very file curl is trying to fetch. Use `sp-update` instead; it's on disk and opens WAN as its first step. 🔄 **Already on a sleepypod** (most common — review a PR on a running pod): ```bash sudo sp-update feat/bio-cal-1-foundation ``` 🚀 **Fresh pod / first install** — bootstraps from GitHub, rebuilds every push: ```bash curl -fsSL https://raw.githubusercontent.com/sleepypod/core/feat/bio-cal-1-foundation/scripts/install \ | sudo bash -s -- --branch feat/bio-cal-1-foundation ``` 🎯 **Pin to this exact build** ([run 38518366373](https://github.com/sleepypod/core/actions/runs/38518366373) · `f170f13`) — fresh-pod path, locked to one CI artifact for reproducible review: ```bash curl -fsSL https://raw.githubusercontent.com/sleepypod/core/f170f13baad42c43ba91ce1f2a7f647c8b6fcd92/scripts/install \ | sudo bash -s -- \ --branch feat/bio-cal-1-foundation \ --artifact-url 'https://nightly.link/sleepypod/core/actions/runs/38518366373/sleepypod-core.zip' ``` 🧊 Artifact: [`sleepypod-core`](https://github.com/sleepypod/core/actions/runs/38518366373/artifacts/11873120351) · self-destructs in 30 days 💥 </details> <!-- sleepypod:install-end -->
ng
added a commit
to ng/core
that referenced
this pull request
Oct 14, 2026
…ckage experiments (sleepypod#829) Part 2 of 4 splitting sleepypod#824 (reference snapshot `7b9ac5d3`) into reviewable stacked PRs. sleepypod#824 stays open as the umbrella and is not merged; the four parts reproduce its tree exactly (the tree of `feat/bio-cal-4-ui` is identical to the tree of `7b9ac5d3`). Stack, merge in order: 1. `feat/bio-cal-1-foundation` → `dev`: versioned calibration profiles, references, DB schema and migrations, shared Python runtime 2. `feat/bio-cal-2-backtest` → part 1: retained backtest replay engine and isolated package experiments 3. `feat/bio-cal-3-shadow-api` → part 2: prospective shadow runtime, isolated coordinator host, calibration API and MCP, demo handlers 4. `feat/bio-cal-4-ui` → part 3: calibration UI, i18n catalogs, browser tests and docs ## This part - Retained backtest execution (`src/biometrics/backtest`): the job runner, retained report artifacts, resource admission, paired-night uncertainty, replay traces and waveform replay of retained RAW copies, on top of the contracts and worker core from part 1. - Isolated algorithm package experiments (`src/biometrics/packages`, `scripts/algorithm-experiments`): manifest validation, Docker-restricted runs, determinism checks, resource evidence and managed trials. - The `biometrics:experiment` script. The coordinator and shadow host bundles are added to `scripts/build/biometrics-worker.mjs` in part 3 together with their entry points. - Review follow-ups on this slice: the package-experiment transport ceiling is derived from the frozen input so a passing freeze cannot fail mid-stream after holdout exposure is recorded, and submission pins the coordinator bundle beside the packaged worker so execution hashes the same bundle. - Coverage pass: unit tests for waveform execution, durable trace reads, runner lease and cancellation paths, waveform selection and the child-process helpers lift this slice's estimated patch coverage from 81% to about 98%. ## Test plan - [x] `tsc --noEmit` clean at this slice - [x] `eslint` on the changed TS/TSX files clean - [x] Full `vitest run` green at this slice (429 files, 6,458 tests) (the only failure under full-suite load was the pre-existing, unrelated `scripts/tests/diagnosticCli.test.ts` timing test, which passes in isolation) - [ ] CI: Typecheck, Lint, Unit Tests, Python modules, Build & Package - [ ] CodeRabbit review via the mirror proxy addressed 🤖 Generated with [Claude Code](https://claude.com/claude-code) <!-- sleepypod:install-start --> <!-- 🛑 AUTO-GENERATED — sleepypod-bot owns this block. Hello fellow agent / curious human! Anything between the sleepypod:install-start and sleepypod:install-end markers gets overwritten on the next CI run. Want to change what's here? Edit .github/workflows/build.yml instead. 💤 --> <details open> <summary>🛏️ <strong>Beam this PR onto a sleepypod</strong> ✨</summary> Pick the snippet that matches your pod. **Already-installed pods can't use the curl bootstraps below** — sleepypod's egress firewall DROPs github.com, and the script that knows how to unblock WAN is the very file curl is trying to fetch. Use `sp-update` instead; it's on disk and opens WAN as its first step. 🔄 **Already on a sleepypod** (most common — review a PR on a running pod): ```bash sudo sp-update feat/bio-cal-2-backtest ``` 🚀 **Fresh pod / first install** — bootstraps from GitHub, rebuilds every push: ```bash curl -fsSL https://raw.githubusercontent.com/sleepypod/core/feat/bio-cal-2-backtest/scripts/install \ | sudo bash -s -- --branch feat/bio-cal-2-backtest ``` 🎯 **Pin to this exact build** ([run 38519463600](https://github.com/sleepypod/core/actions/runs/38519463600) · `b925d8f`) — fresh-pod path, locked to one CI artifact for reproducible review: ```bash curl -fsSL https://raw.githubusercontent.com/sleepypod/core/b925d8f6f6e2a7461925ff5f6b5c3bd42d0e0197/scripts/install \ | sudo bash -s -- \ --branch feat/bio-cal-2-backtest \ --artifact-url 'https://nightly.link/sleepypod/core/actions/runs/38519463600/sleepypod-core.zip' ``` 🧊 Artifact: [`sleepypod-core`](https://github.com/sleepypod/core/actions/runs/38519463600/artifacts/11872719480) · self-destructs in 30 days 💥 </details> <!-- sleepypod:install-end -->
ng
added a commit
to ng/core
that referenced
this pull request
Oct 14, 2026
…d calibration API (sleepypod#830) Part 3 of 4 splitting sleepypod#824 (reference snapshot `7b9ac5d3`) into reviewable stacked PRs. sleepypod#824 stays open as the umbrella and is not merged; the four parts reproduce its tree exactly (the tree of `feat/bio-cal-4-ui` is identical to the tree of `7b9ac5d3`). Stack, merge in order: 1. `feat/bio-cal-1-foundation` → `dev`: versioned calibration profiles, references, DB schema and migrations, shared Python runtime 2. `feat/bio-cal-2-backtest` → part 1: retained backtest replay engine and isolated package experiments 3. `feat/bio-cal-3-shadow-api` → part 2: prospective shadow runtime, isolated coordinator host, calibration API and MCP, demo handlers 4. `feat/bio-cal-4-ui` → part 3: calibration UI, i18n catalogs, browser tests and docs ## This part - Prospective shadow runtime (`src/biometrics/shadow`): bounded storage, worker transport and supervision, kernel budgets, admission bound to installed qualification and observed hardware, scheduling on pod boot, cancellation and recovery. - Isolated coordinator host for replay and shadow work (`src/biometrics/backtest/coordinator*`, `host*`) and Node-only instrumentation startup. - Calibration API surface (plus the profile diff, status, overview and reference comparison helpers that only the routers consume): tRPC routers (tuning, datasets, backtests, references, shadow), OpenAPI additions, RAW route changes and MCP tools, resources and prompts. - Demo handlers and the simulated browser evaluator that implement the new router contract, with the demo worker build identity and `prebuild` hook. - Pod benchmark runners for replay, waveform and shadow measurements. - Review follow-ups on this slice: an in-window shadow frame already dequeued at collection end is forwarded instead of silently dropped, and profile diffs report keys present on either side. - Rebase on `dev`: the app router, MCP tool and prompt lists, and the MCP README count now include both the saved-curve additions from sleepypod#825/sleepypod#827 and the biometric additions from this stack. - Coverage pass: tests for the shadow API and store, demo biometrics handlers, MCP biometric tools, waveform publication and Node instrumentation shutdown lift this slice's estimated patch coverage from 82% to about 93%. ## Test plan - [x] `tsc --noEmit` clean at this slice - [x] `eslint` on the changed TS/TSX files clean - [x] Full `vitest run` green at this slice (474 files, 6,959 tests) (the only failure under full-suite load was the pre-existing, unrelated `scripts/tests/diagnosticCli.test.ts` timing test, which passes in isolation) - [ ] CI: Typecheck, Lint, Unit Tests, Python modules, Build & Package - [ ] CodeRabbit review via the mirror proxy addressed 🤖 Generated with [Claude Code](https://claude.com/claude-code) <!-- sleepypod:install-start --> <!-- 🛑 AUTO-GENERATED — sleepypod-bot owns this block. Hello fellow agent / curious human! Anything between the sleepypod:install-start and sleepypod:install-end markers gets overwritten on the next CI run. Want to change what's here? Edit .github/workflows/build.yml instead. 💤 --> <details open> <summary>🛏️ <strong>Beam this PR onto a sleepypod</strong> ✨</summary> Pick the snippet that matches your pod. **Already-installed pods can't use the curl bootstraps below** — sleepypod's egress firewall DROPs github.com, and the script that knows how to unblock WAN is the very file curl is trying to fetch. Use `sp-update` instead; it's on disk and opens WAN as its first step. 🔄 **Already on a sleepypod** (most common — review a PR on a running pod): ```bash sudo sp-update feat/bio-cal-3-shadow-api ``` 🚀 **Fresh pod / first install** — bootstraps from GitHub, rebuilds every push: ```bash curl -fsSL https://raw.githubusercontent.com/sleepypod/core/feat/bio-cal-3-shadow-api/scripts/install \ | sudo bash -s -- --branch feat/bio-cal-3-shadow-api ``` 🎯 **Pin to this exact build** ([run 38520827460](https://github.com/sleepypod/core/actions/runs/38520827460) · `338aabe`) — fresh-pod path, locked to one CI artifact for reproducible review: ```bash curl -fsSL https://raw.githubusercontent.com/sleepypod/core/338aabe8f6e790dcee6c1e088ce23a6159c304e6/scripts/install \ | sudo bash -s -- \ --branch feat/bio-cal-3-shadow-api \ --artifact-url 'https://nightly.link/sleepypod/core/actions/runs/38520827460/sleepypod-core.zip' ``` 🧊 Artifact: [`sleepypod-core`](https://github.com/sleepypod/core/actions/runs/38520827460/artifacts/11874181456) · self-destructs in 30 days 💥 </details> <!-- sleepypod:install-end -->
ng
added a commit
to ng/core
that referenced
this pull request
Oct 14, 2026
…od#831) Part 4 of 4 splitting sleepypod#824 (reference snapshot `7b9ac5d3`) into reviewable stacked PRs. sleepypod#824 stays open as the umbrella and is not merged; the four parts reproduce its tree exactly (the tree of `feat/bio-cal-4-ui` is identical to the tree of `7b9ac5d3`). Stack, merge in order: 1. `feat/bio-cal-1-foundation` → `dev`: versioned calibration profiles, references, DB schema and migrations, shared Python runtime 2. `feat/bio-cal-2-backtest` → part 1: retained backtest replay engine and isolated package experiments 3. `feat/bio-cal-3-shadow-api` → part 2: prospective shadow runtime, isolated coordinator host, calibration API and MCP, demo handlers 4. `feat/bio-cal-4-ui` → part 3: calibration UI, i18n catalogs, browser tests and docs ## This part - Calibration UI (`src/components/biometrics`): profile editing and revisions, reference imports and corrections, dataset freezing, backtest submission and monitoring, report review, replay traces, shadow planning and prospective observations. - Settings integration for sleep-stage rules and Watch imports with calibration profiles. - Spanish, English and pseudo catalogs for the new messages. - Playwright browser tests (`tests/browser`), the `test:browser` script and the Browser Tests CI job. - Review follow-ups on this slice: `signal_replay` experiments are labelled in the experiment list, CSV formula prefixes behind leading line breaks are neutralized, the correction and measurement panels in the reference library use distinct React keys, and the calibration components gained unit cases for their error, loading, paging and toggle paths so the per-PR patch-coverage gate holds without the browser suite. - Calibration docs: API, datasets, reference imports, backtest jobs, runtime activation, replay benchmark and readiness map. - Second coverage pass: CandidateStageFields and CalibrationOverview gained tests and the remaining Calibration components' branches are driven; estimated patch coverage is about 92% (remaining misses are i18n locale fallbacks and guards unreachable from the DOM). ## Test plan - [x] `tsc --noEmit` clean at this slice - [x] `eslint` on the changed TS/TSX files clean - [x] Full `vitest run` green at this slice (507 files, 7,164 tests) (the only failure under full-suite load was the pre-existing, unrelated `scripts/tests/diagnosticCli.test.ts` timing test, which passes in isolation) - [ ] CI: Typecheck, Lint, Unit Tests, Python modules, Build & Package - [ ] CodeRabbit review via the mirror proxy addressed 🤖 Generated with [Claude Code](https://claude.com/claude-code) <!-- sleepypod:install-start --> <!-- 🛑 AUTO-GENERATED — sleepypod-bot owns this block. Hello fellow agent / curious human! Anything between the sleepypod:install-start and sleepypod:install-end markers gets overwritten on the next CI run. Want to change what's here? Edit .github/workflows/build.yml instead. 💤 --> <details open> <summary>🛏️ <strong>Beam this PR onto a sleepypod</strong> ✨</summary> Pick the snippet that matches your pod. **Already-installed pods can't use the curl bootstraps below** — sleepypod's egress firewall DROPs github.com, and the script that knows how to unblock WAN is the very file curl is trying to fetch. Use `sp-update` instead; it's on disk and opens WAN as its first step. 🔄 **Already on a sleepypod** (most common — review a PR on a running pod): ```bash sudo sp-update feat/bio-cal-4-ui ``` 🚀 **Fresh pod / first install** — bootstraps from GitHub, rebuilds every push: ```bash curl -fsSL https://raw.githubusercontent.com/sleepypod/core/feat/bio-cal-4-ui/scripts/install \ | sudo bash -s -- --branch feat/bio-cal-4-ui ``` 🎯 **Pin to this exact build** ([run 38524545463](https://github.com/sleepypod/core/actions/runs/38524545463) · `d71688c`) — fresh-pod path, locked to one CI artifact for reproducible review: ```bash curl -fsSL https://raw.githubusercontent.com/sleepypod/core/d71688cc42c52dc7bc9107f8929907d1f60d38ce/scripts/install \ | sudo bash -s -- \ --branch feat/bio-cal-4-ui \ --artifact-url 'https://nightly.link/sleepypod/core/actions/runs/38524545463/sleepypod-core.zip' ``` 🧊 Artifact: [`sleepypod-core`](https://github.com/sleepypod/core/actions/runs/38524545463/artifacts/11875403337) · self-destructs in 30 days 💥 </details> <!-- sleepypod:install-end -->
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Biometric personalization previously lacked a shared, reproducible path from reference comparison to an acknowledged runtime change. This adds per-side versioned profiles, retained evaluations and explicit activation/rollback across the web UI, REST/tRPC and MCP. Original observations retain their provenance and stay distinct from current-classifier views and replay results.
The implementation includes:
Validation at
7b9ac5d3: 6,887 Vitest tests passed (18 skipped), across 482 passing test files. Browser tests, all seven Python module jobs, both builds, TypeScript and lint passed. Patch coverage is 87.73%, passing the repository's unchanged eight-point margin below its 88.93% target. Pre-push checks also passed. The added regressions cover coordinator admission/cancellation/resource ownership, discovery boundaries, retained report publication/release, and the live-shadow exit race.Target-pod resource evidence includes an eight-hour, three-configuration synthetic replay completed in 45.94 minutes with zero read/API failures and unchanged fixture canonical observations. The separate 30-minute synthetic shadow supervisor measurement completed in 61.34 seconds with confirmed cleanup and zero probe failures. A subsequent 180-second live source/isolated sink run completed with six retained comparison points, zero canonical rows in the isolated database and 1,587 successful read-only API probes. It exposed and verified a fix for RSS disappearance during clean worker exit. Detailed identities, observations and limitations are in docs/biometrics-replay-benchmark.md.
Synthetic performance evidence does not establish clinical accuracy. Validated promotion and prospective shadow admission still require matching reviewed qualification registries; no success is fabricated when that evidence is absent. See docs/biometrics-calibration-readiness.md for the acceptance map and remaining rollout evidence.
Deployed build
4c3a6a6e(including the live-shadow exit fix) is verified on the target pod: calibration/health endpoints return 200, replay admission is configured and admitted, core and numerical services are active, and the data-path verdict is healthy. No user profile or hardware temperature setting was changed. Subsequent changes add tests, a measurement harness and documentation; they do not change the deployed runtime.CodeRabbit full review was requested in this PR. The provider skipped it because the change exceeds its 100-file review limit; its passing check does not mean a review was completed. All build, test and coverage checks pass on the final revision; see the checks panel for current status.