Pure, DB-free cycle prediction and phase-assignment engine under
src/lib/cycle/. The math mirrors the offline-parity spec exactly so the
server and the native client compute byte-identical forecasts from the
same canonical inputs.
- day-math: noon-UTC anchored YYYY-MM-DD arithmetic (dayDiff, addDays,
round-half-up) that stays stable across timezones and DST transitions.
- prediction: robust median+MAD cycle-length estimator over a trailing
12-cycle window with 3-MAD outlier exclusion and missed-log detection;
period-length estimator; one-sigma Low/High date band widened by
logging sparsity; luteal back-calculated ovulation + six-day fertile
window; the CALENDAR -> symptothermal -> temperature-trend method
ladder (sensiplan 3-over-6 + mucus-peak re-implemented from the
published method) with documented blend precedence; a multiplicative
confidence scalar shrinking with length variance, low cycle count, and
sparse logging.
- phase: day-of-cycle -> MENSTRUAL/FOLLICULAR/OVULATORY/LUTEAL boundary
mapping with menstrual precedence on overlap.
- 48 unit tests covering the regular, irregular, cold-start, outlier,
symptothermal, temperature-trend, and every phase-boundary case; all
assertions are timezone-independent and pin exact dates + confidence.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Introduce the persistence layer for cycle tracking: eight enums
(CycleTrackingGoal, FlowLevel, OvulationTest, CervicalMucus,
HomeTestResult, ContraceptiveKind, CyclePhase, PredictionMethod) and
seven models.
CycleProfile is a 1:1 per-user settings row (lazily upserted on first
access via getOrCreateCycleProfile). MenstrualCycle and CycleDayLog are
the capture rows, carrying the MoodEntry soft-delete + syncVersion shape
so they ride the sync delta feed; CycleDayLog repeats the MoodEntry
dual-unique — the canonical (userId, date) key plus the NULL-distinct
(userId, source, externalId) key for HealthKit re-ingest. Free-text
notes encrypt at rest; queryable physiology (flow, BBT, OPK, mucus,
dates) stays plaintext for the rollup and correlation tier. The symptom
taxonomy (CycleSymptomCategory / CycleSymptom / CycleSymptomLink) clones
the mood-tag trio and seeds fifteen symptoms across three categories
with cycle.symptom.* i18n keys across all six locales. CyclePrediction
caches one forecast per user, regenerated debounced.
Migrations 0128-0130 are additive; the testcontainer integration suite
applies the full history from scratch and passes.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Medication cycle status: re-hydrate `nextDueAt` (it arrives as an ISO string
over the wire) before the Berlin day-bucketing, so "next dose in N days" no
longer drifts by a day near midnight or off Berlin. Covered by a new
string-input test.
- Wellness rings: drive the conic sheen off a fixed white so it always lightens
(a near-black light-mode foreground under overlay was darkening the tile into
a grey band); give the ring a transform transition so it scales smoothly with
the tile lift instead of snapping; cap the green-band glow at a few breaths
instead of an infinite drop-shadow repaint; deepen the light-mode stress and
recovery end-stops for non-text contrast; track the sm/md column count to the
active-score count so two or four scores never leave an empty cell.
- Mood factor crosstab: bound the snapshot's cross-metric read to the 365-day
window and exclude soft-deleted rows, so the factor deltas the model cites
match the visible board over the same canonical row set (no recency-biased
truncation, no soft-delete skew); thread the user's source priority so the
per-day canonical pick agrees with the screen; render the delta neutrally
rather than green/red, since a higher vital is good for steps but bad for
resting heart rate and the board is observational.
- Log-intake: surface an error when a pinned slot can't resolve to an instant
instead of silently logging the dose through the unscheduled path.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Rated mood factors (work / sleep-quality / stress, scored per entry) now
participate in the cross-metric correlation surface. Until now their score
series was captured but read by zero analytics.
Two paths, both on the existing FDR-controlled engine:
- A factor x vital crosstab on the mood insights page and the mood
snapshot: for each factor the user scores, split its days low vs high by
the factor's own median, then run the same Welch test + day floors +
Benjamini-Hochberg correction the tag crosstab uses. Surfaces lines like
"sleep runs ~38 min below baseline on the days you rated work low"
(n-gated, q <= 0.10, never fabricated on thin data).
- Factor channels (FACTOR:<key>) fold into the period-narrative discovery
matrix as both lag source and lag target, so the drivers the coach and
narrative reference can cite a factor relationship.
An inverse factor (higher rating = worse day) gets the documented sign-flip
once, at the series boundary, so "up" always reads as a better day and the
low/high split and phrasing stay honest. Strings land in all six locales;
the surface carries the standing "association, not cause" caption.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add an additive, non-breaking assessment field to GET /api/insights/derived:
{ text, source, updatedAt } | null, keyed to the requested metric id for the
five per-score ids (READINESS, SLEEP_SCORE, RECOVERY_SCORE, STRAIN_SCORE,
STRESS_SCORE). The field is null when the metric is not one of those or when
its status is not "ok". A deterministic text always fills it — composed from
the score's own contributors, so provider-less accounts and the demo are never
empty — and warmer provider prose overrides it once cached, warmed out of band
so the GET never blocks on a model round-trip. de/en, following the v1.12.12
deterministic-narrative model. The batch grid read returns assessment: null;
the detail sheets use the single-metric route.
Introduce a shared buildMetricSignal() helper: a token-lean, signal-rich
descriptor (current, own baseline, signed delta, spread, normal-swing verdict,
normal-range placement, contributors) computed server-side so the model states
the comparison instead of re-deriving it from raw buckets. Thread it into the
generic per-metric assessment snapshot and lead the base prompt from it. Add a
PROMPT_VERSION constant to the base system prompt so prompt changes are
traceable.
Regenerate docs/api/openapi.yaml.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Let a user shape their own mood-tag set on top of the global catalogue:
create custom tags (label encrypted at rest, icon from a fixed allow-list),
and hide built-in tags per user. Custom tags live in `mood_tags` with a
non-null `user_id` and a `custom:<uuid>` key; `GET /api/mood/tags` returns
the effective per-user set (catalogue − hidden + own customs) with new
`custom` + `label` fields, and `?include=hidden` surfaces hidden catalogue
tags for the management screen. The write path resolves a `custom:` key only
against the caller's own rows — a foreign custom key is dropped, never linked.
Delete soft-deactivates by default (history intact); `?purge=true` removes
the row and its links. BINARY-only for v1.
New: POST/PATCH/DELETE /api/mood/tags/custom[/:key], PUT /api/mood/tags/:key/hidden,
migration 0127 (user_id + label_encrypted columns, mood_tag_hidden table,
`custom` category). Integration setup snapshots + restores the seeded catalogue
because the new mood_tags→users FK makes TRUNCATE users CASCADE wipe it.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
When no AI provider is configured (or before the first AI summary is written),
the period-in-review card stayed empty. Compose a concise, factual, strictly
non-causal summary from the same structured digest the AI prompt is built from
— ranked metric changes vs the prior period, vitals that moved outside their
typical range, and an honest count of statistical associations — and persist it
with a `deterministic` provider marker so a later AI warm replaces it in place.
The read route now warms regardless of provider; the read-never-blocks posture
is unchanged. Fills the no-key demo's retrospective too.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A slot-less "taken now" write (no client scheduledFor) defaulted its
instant to now/takenAt and snapped across the wide ±halfGap window onto
the nearest scheduled slot. For a 07:00 / 19:00 med a midday write
captured the 07:00 morning slot and stamped it taken — a phantom morning
dose. The web fix only threaded the displayed slot on the card; any
client posting takenAt-only (iOS bulk/sync) still tripped it server-side.
Thread an instantIsExplicit flag through resolveSlotInstantForWrite into
the canonical resolver. An explicit scheduledFor keeps the full ±halfGap
snap unchanged. A defaulted-now write snaps only inside the tight
dose-grace window and otherwise resolves to null (PRN), so it records a
standalone taken-now row instead of back-filling a far slot. Both intake
call sites pass scheduledFor !== undefined; takenAt alone is not explicit.
Add resolver unit coverage for midday defaulted-now writes (07:00/19:00 →
null at 11:00/12:00/13:00, explicit 07:00 still snaps) and a DB-backed
regression that a takenAt-only midday write leaves the 07:00 slot untaken.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The retired RATED factors leave no seeded factor with a scale narrower
than the 1..5 rating envelope, so the per-tag out-of-scale "skipped" path
is no longer reachable through the catalog (it stays covered by the
`resolveRatedFactors` unit tests). Update the bulk integration tests to
use surviving factors and assert the envelope rejection, and swap a
deactivated tag for an active one in the tag-key round-trip. Refresh the
stale per-tag-scale comment in the rated-factor schema.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a README to src, src/lib, src/app, src/components, prisma, scripts,
and e2e so a technical visitor browsing the tree gets oriented: what
lives in each directory, the most important files, the local
conventions, and pointers to related dirs and docs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
MOOD is mood-entry backed, not a MeasurementType; it was filtered out of
DISCOVERY_BEHAVIOURS but not DISCOVERY_OUTCOMES, so it leaked into the
measurement.findMany type filter and threw PrismaClientValidationError,
failing the period-narrative warm job.
The BYOK Anthropic client fell back to the retired claude-3-5-sonnet-latest;
match the chain-runner default (claude-sonnet-4-6). The admin ai-settings GET
displayed a gpt-4o-mini fallback that disagreed with the actual provider
default (gpt-4o); align the display string so the surfaced model matches what
the provider would use.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
mapCycle now emits AVERAGE_HEART_RATE and MAX_HEART_RATE from the cycle
score's average_heart_rate / max_heart_rate — both were already on the typed
WhoopCycle interface but dropped on the floor. mapSleep now reads
stage_summary.disturbance_count (added to the WhoopSleep interface) and emits
SLEEP_DISTURBANCE_COUNT. All three guard nulls / unscored records and carry
distinct field tags (avg_hr / max_hr / disturbances) so they don't collide
with the existing strain / energy / stage rows under the shared resource
externalId. Update mapping.md to document the three and note the cycle HR
pair were previously fetched-but-dropped.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add three MeasurementType values for WHOOP fields a coverage audit found
uncaptured: AVERAGE_HEART_RATE and MAX_HEART_RATE (the cycle score's
whole-day aggregates, kept distinct from spot PULSE / RESTING_HEART_RATE /
WALKING_HEART_RATE_AVERAGE) and SLEEP_DISTURBANCE_COUNT (the per-night
disturbance tally). Units: bpm / bpm / count.
Wiring mirrors the existing per-type surfaces, traced from VO2_MAX and the
v1.11.0 WHOOP-native score classes:
- canonical enum + unit + plausibility range in
src/lib/validations/measurement.ts (every ingest / analytics / rollup
path derives from measurementTypeEnum.options, so the mean-based rollup
tier picks the new types up automatically).
- Prisma enum + migration 0125 (ALTER TYPE ADD VALUE IF NOT EXISTS,
non-transactional, idempotent — same pattern as 0111).
- category overlay (cardiovascular for the HR pair, sleep for the count).
- list-view label key / Lucide icon / chart-colour.
- i18n label key in all six locales.
- chart-token allowlist + orphan-enum stripper.
- PersonalRecord direction (HR pair null — no clean goal axis, like
WALKING_HEART_RATE_AVERAGE; disturbance count MIN, like
BREATHING_DISTURBANCES).
- enum-coverage drift guard (count 60 -> 63, doctor-PDF exclusions for the
device-derived signals) + apple-health HK-counterpart exclusion (these
are WHOOP-only; HealthKit ships no first-class daily-aggregate HR
quantity nor a disturbance tally).
The types render on the measurement list + Insights chart surfaces; like the
v1.11.0 WHOOP-native scores they get no dedicated Insights sub-page (the
sub-page map stays curated) and no LOINC / FHIR mapping (not clinical-grade).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The v1.12.1 diversity block (variety lead, data-strength line, steady-run
repetition signal, FDR-screened cross-metric relations) reached only the
generic archetype cards. The seven hand-tuned headline cards — the ones users
see most — ran at temperature 0.3 with only the previous-context block, so they
stayed prone to repetitive openers and forced filler steps. Thread the same
assessmentContextBlock through the specialized get*UserPrompt builders and
raise their temperature to 0.45 to match the archetypes.
Soften the unconditional "close with one step" across the specialized sections
and the activity / body-composition / sleep archetypes to the base prompt's
conditional form — close with a step only when the finding implies one,
otherwise affirm and name one thing to watch — so a steady metric no longer
manufactures a filler step. EN/DE parity preserved; guard computeSteadyRun
against an absent graded slice.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
general-status and blood-pressure-status read the full medication-intake
history with no time floor, scanning every row for accounts with years of
data. Both only consume the trailing ~365 days (the adherence fold and the
continuity series / 7-30 day compliance gate), so add a scheduledFor floor at
now - 365d.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The reference-grounding builder (buildSystemPromptWithReferences) and its
SOURCES list shipped to no one: both live paths called the plain system prompt,
so the model was asked to cite referenceId values it was never shown — and the
parser rejects invented ids, so the citation machinery the schema and UI
footnote already support stayed dark. Wire the builder into both consumers,
passing the metric sections the snapshot actually carries (a weight-only run
never lists BP guidance). Apply the fenced-JSON parse fallback here too.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Anthropic and the local provider do not send a JSON mode, so a compliant model
still wraps its { "summary": … } reply in a code fence or prefixes a sentence,
which broke JSON.parse and surfaced the raw fenced string as the assessment.
Add a shared stripJsonFences helper (drop the fence, narrow to the first { …
last }) and apply it as a parse fallback in parseSummaryFromContent. No-op on
already-clean JSON.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The insights ground-rules carried changelog tokens ("v1.4.16 phase B8", "W4d")
inside the prompt strings the model reads — token waste and inconsistent with
the project voice. Move the rule content to clean numbered rules; the markers
belong in code comments, not shipped prompt prose. Update the test pins to the
cleaned rule headers.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The OpenAI BYOK + admin-shared fallback defaulted to a small model and
Anthropic to a 2024 Sonnet, so a user on the operator's shared key got the
weakest model for the Insights assessments. Bump the OpenAI default to gpt-4o
and the Anthropic default to claude-sonnet-4-6; both stay overridable by user
config. Sync the provider docs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The snapshot builder probed rollup coverage up front to gate the thick
extras phase, then called computeSummariesSlice which re-ran the
identical probeRollupCoverage query. Thread the already-probed map into
the slice as an optional argument so the snapshot's hot path issues the
probe once. The argument is optional with the internal probe as the
fallback, so every other caller is unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The dashboard snapshot lives in the analytics cache bucket, which had no
stale window, and every measurement write hard-evicted the whole
${userId}| prefix. A high-frequency iOS Apple-Health sync therefore
busted the snapshot into a cold rebuild on the next read all day long
(the same class as the v1.8.7 insight-cache eviction).
Give the analytics bucket a stale-while-revalidate window, read the
snapshot via cachedSwr, and switch the measurement-write invalidation
from a hard deleteByPrefix to markStaleByPrefix. The snapshot now serves
the prior value immediately while a single background recompute warms a
fresh one. Slim / thick / summary cells read via plain cached are
unaffected: a marked-stale entry has expiresAt === now, so their next
read is a clean miss and rebuilds fresh. Widget reorder keeps its hard
point-evict so a layout change paints synchronously; mood and medication
writes keep their hard analytics sweep.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
findCached returned null and the response was only persisted after the
handler completed, so two concurrent requests carrying the same
Idempotency-Key both passed the lookup and both executed the side-effect
(duplicate writes for the very clients idempotency protects).
Claim the key by inserting a pending sentinel row under the existing
(userId, key, method, path) unique constraint before the handler runs.
A racing request either finds the pending row or loses the insert race
and is refused with 409. The claim is promoted to the completed response
on a cachable result, released on a non-cachable result or a
secret-shaped body, and released on a handler throw. The pending row
carries a short claim TTL so a crashed handler self-heals and the
existing expiry-sweep cleanup still applies.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The analytics compliance streak keyed days with host-local
getFullYear/getMonth/getDate against a timeline whose slot.day is minted
in the user's IANA zone, so the streak drifted off by a day whenever the
server clock's zone differed from the user's. Consolidate onto the
detail-page streaksFromTimeline helper threaded with the medication's
timezone so both call sites share one correct day-keyed walk.
Also fix the DST step in that helper: it stepped a fixed +24h while its
comment claimed 25h, which could skip a local day on the 23h
spring-forward boundary and run the window short. Snap each cursor to
the user-local midnight and advance +25h then re-snap, mirroring the
cadence expander, so every local calendar day is visited exactly once
across 23/24/25h boundaries.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The per-sample cumulative drain strips already-collapsed stats: rows
from the day bucket, so a trailing watch sync that lands after a day's
total was minted leaves the bucket holding only the late samples. The
upsert update branch then replaced the full day total with the partial
late sum, and the contributing samples were already hard-deleted, making
the lost count unrecoverable.
Probe the existing total inside the write transaction and fold the late
sum into it (existing + late) instead of overwriting. The late samples
are genuine additional readings, unlike the legacy-step pass where the
existing total is HealthKit's own aggregate and folding would double-
count. Per-sample rows are still hard-deleted so a re-run cannot
re-accumulate them.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add an optional Recharts <Brush> to the shared HealthChart, gated behind a
new `selectableDomain` prop. When set, the chart mounts the brush beneath
the plot and reports the Min/Max/Median/Mean over the brushed window per
type through `onDomainStats`. A range-tab change or a series refetch clears
or clamps the selection so the brush never indexes past the data.
The dashboard, trends-row, VO2 row, sleep, and rationale mounts do not pass
the prop, so they render unchanged — the feature is fully isolated behind
the flag.
`computeWindowStats` is a pure helper (no React/Recharts) so the chart can
summarise the array it already holds without a refetch and the maths is
unit-tested directly. The median uses the same linear-interpolated midpoint
as `summarize()` so a window and the full-range summary read on one
algorithm. `useChartDomainStats` lifts the per-type window into shared state
with a stable callback and a shallow guard against repeat windows.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The generic metric-status generator read SLEEP_DURATION row-per-stage and
fed the snapshot a single sleep STAGE (~16 min) as the current value, so the
sleep prose reported a fraction of an hour for a full night. Route the
SLEEP_DURATION status/snapshot through reconstructSleepNights — the same
per-night, source-deduped, midnight-aware time-asleep total the dashboard,
series, and targets surfaces use — so the snapshot's series, summary, and
latest value are all the night total in minutes (the registry unit and normal
range are already minutes). Gated to SLEEP_DURATION; no other metric's status
path changes.
The readiness wellness index could surface a contributor value of 0 for an
account that recorded resting heart rate / HRV / respiratory rate but had no
scorable baseline: a non-finite baseline center or today value fell through
clamp100 to a misleading 0 that reads as a real worst-case sub-score. Return
null instead so the client drops the contributor, and document the null-vs-0
contract: null means no usable signal (no baseline, no recent reading,
non-finite input); a finite value — including a genuine 0 — means the baseline
is established and the latest reading really deviates that far.
(cherry picked from commit 0422ac012765f30069a41f662c2b025197ed9a88)
Make every metric subpage identical by construction in the shared
scaffold rather than per-page bespoke.
- Target tile matches the other cards: dark `bg-card` background and the
`rounded-xl border p-4` spacing the assessment and stat-strip cards
use, instead of the old translucent `bg-card/20` panel.
- Drop the 7T/30T/90T/1J range row and its "+X% vs previous" delta from
every subpage. The chart owns its own window via its in-card range
tabs, so the chart query is unaffected. Retire the now-dead
`MetricRangeControls`, `MetricRangeDelta`, `TimeRangePills`,
`useAnalyticsRange`, the `useInsightsRangePref` persisted-range hook,
and `sentimentFromMetricDirection` along with their tests.
- Fix the spine order to intro -> chart -> target card -> assessment ->
Min/Max/Median/Mittelwert stat strip. The stat strip moves to the foot
of the spine, below the assessment. Remove the "Letzte Messung" card
and the headline primary tile from subpages; retire
`MetricLastMeasurementCard` and `MetricPrimaryTile`.
- Blood pressure gains the stat strip it was missing: two stacked
systolic / diastolic strips, since a single-series strip would drop the
diastolic half. `MetricStatStrip` takes an optional series caption.
Update the spine guard to assert the target card precedes the
assessment and that the retired range row and last-measurement card
appear on no subpage.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
(cherry picked from commit 80b2a810314b4c0951c0d68c04bffc89003005d6)
The per-metric assessment prompt (base-system) and the scope-hardened
insight system prompt (insight-generator) each carried EN and DE as two
parallel template literals kept in sync by hand — a standing drift risk
flagged in the AI-quality audit.
Model each prompt as one section skeleton: a single ordered list where
every section supplies its EN and DE fragment side by side, and a builder
joins the fragments with a blank line. The section order and prose are
unchanged; the DE calibration is preserved verbatim. A section edited in
one locale but not the other now shows up as an asymmetric edit in one
list instead of two literals silently diverging.
Pure refactor — every emitted prompt string is byte-identical for both
locales across every metric, surface and reference combination. A new
locale-skeleton-parity test pins the structure (each section carries both
fragments, composed prompt equals the joined fragments); the existing
grounding, filler-ban, forbidden-phrase, refusal and exact-phrasing guards
pass unchanged.
(cherry picked from commit 12cd718d6d602a48cafc14ad06d82282fdaacb27)
Two connect-in-app enhancements for the native client.
return_scheme: GET /api/whoop/connect accepts an optional custom scheme,
threaded through the server-side OAuth-state row so it survives the round
trip without ever riding in the URL or cookie. When a valid scheme is
present, the callback's final redirect targets <scheme>://whoop?whoop=
connected|error&reason=... so an in-app web session auto-completes on its
custom-scheme match; absent or invalid, the web settings redirect is
unchanged. Schemes validate against a strict allowlist (custom app scheme
only, ^[a-z][a-z0-9.+-]*$, http/https/javascript/data/file/vbscript
rejected, pinned to dev.healthlog.app). A rejected scheme falls back to
the web redirect and never reflects an arbitrary value.
Bearer connect ticket: a purely Bearer-authenticated client holds no web
session cookie, so it cannot start the handshake through the cookie-gated
connect route. POST /api/whoop/connect/ticket (Bearer) mints a one-time,
~60s, opaque ticket bound to the user; only its HMAC-SHA256 hash is stored
(the raw value never persists). GET /api/whoop/connect?ticket=<opaque>
accepts the ticket in lieu of a cookie, consumes it atomically (single-use,
reuse rejected), sets the nonce cookie, and 302s to WHOOP exactly as the
cookie path. Expired, consumed, or invalid tickets return a typed 401. The
two combine. The daily WHOOP OAuth-state cleanup also sweeps expired ticket
rows.
The ticket query param is added to the egress redaction denylist so the
opaque value never reaches logs or error reports.
Migration 0124 adds whoop_connect_tickets (id, user_id FK cascade,
token_hash unique, expires_at, consumed_at, created_at) and a nullable
return_scheme column on whoop_oauth_states.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
(cherry picked from commit 221b362011d52dda314139e0be7cc195c47d016e)
The native client SPKI-pins the server's TLS leaf certificate. When the
leaf auto-renews (GTS re-issues it roughly every 90 days) the leaf keypair
— and so the SPKI pin — changes, and a client shipped only the old pin
refuses to connect on the next renewal: a silent outage. Nothing
server-side signalled that a client pins the served leaf.
Add a pg-boss monitor that probes the served leaf every 6 hours, computes
its iOS SPKI pin (base64(sha256(DER subjectPublicKeyInfo)) via a raw TLS
socket + X509Certificate), and compares it against the operator's
known-good set in TLS_LEAF_SPKI_PINS (comma-separated for the dual-pin
renewal window). On a served pin outside the set it fails loud: a
tls.pin.leaf_changed wide-event annotation, a system.tls.pin_changed audit
row, and a high-priority SYSTEM_ALERT fanned out to admins through the
existing dispatcher idiom. Transient TLS/network failures annotate
tls.pin.probe_failed and never alarm.
The baseline is an env var, not a persisted last-seen row — a self-learning
baseline would silently adopt the first rotation and suppress the alarm the
pinned client needs; an unset baseline logs the served pin and warns rather
than auto-adopting. The runbook documents extraction, the re-pin + dual-pin
procedure (≥11 days before expiry, ship to TestFlight), and the assumption
that the probed host is the host the client connects to.
Wires the queue into the worker (allQueues + cron + worker), adds the
optional env to the manifest, compose whitelist, and .env example, and
covers the SPKI computation + change detection with a fixture-cert unit
test plus a queue-registration guard.
(cherry picked from commit cdb88cd77b88aec71ad2f21caf2b15c185dfc01e)
Migration 0122 added MoodEntry.externalId plus the NULL-distinct
@@unique([userId, source, externalId]) and wired the moodLog webhook
and sync importer onto it, but the iOS-facing native routes never used
it: POST /api/mood-entries dropped externalId entirely (the schema did
not declare it) and POST /api/mood-entries/bulk accepted it in its Zod
schema yet parsed-and-ignored it, still upserting on the legacy
(userId, date, moodLoggedAt) key. The dedup index existed but nothing
on the native write path reached it, so the idempotent re-import the
client needs was not functional.
Declare externalId on createMoodEntrySchema (bounded to match the bulk
schema), persist it, and — when present — upsert the single-entry write
on (userId, source, externalId) so a re-post with the same id updates
the existing row in place instead of 409-ing or minting a duplicate.
When absent, keep the legacy first-write create and its 409-on-conflict
behaviour untouched. On the bulk path, branch the per-entry probe and
upsert onto the same compound key when an entry carries an externalId,
refreshing date and moodLoggedAt in the update so a re-zoned re-import
lands the corrected wall-clock on the same row; absent ids keep the
legacy wall-clock key. Both paths resolve source once so the dedup key
and the row write agree. Echo externalId back on the single response
(via the row) and on each bulk per-entry result so the client can map
server ids onto its local rows.
Cover the present-vs-absent upsert key, cross-user isolation, and the
echo with unit tests on both routes plus a real-Postgres integration
test proving a re-posted externalId resolves to one row updated in
place, a null externalId keeps wall-clock behaviour, and the same id
under two users stays isolated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
(cherry picked from commit 6a65de22bc436daeeb73dd2a1cb612336c74a928)
Settings → Integrations fired five status round-trips on every mount —
the consolidated /api/integrations/status alongside the four per-provider
/api/<provider>/status queries. Extend the consolidated endpoint to carry
every field the four cards render (Withings activity scope, WHOOP/Fitbit
backfill state, moodLog webhook secret + entry count) and drive all four
cards off the single envelope. The legacy per-provider routes stay for the
iOS/test callers; the web cards no longer hit them.
Mood-insights cache was a plain 60 s TTL that was hard-evicted on every
mood write, so an active logger re-paid the multi-second cold compute on
every entry. Add stale-while-revalidate to ServerCache (wrapSwr / cachedSwr
+ a markStaleByPrefix that collapses the fresh TTL without dropping the
value), give the moodInsights bucket a 10-minute stale window, route the
read through cachedSwr, and switch the mood-write invalidation from a hard
evict to a mark-stale so the prior aggregate serves immediately while a
single background recompute warms a fresh one.
Defer the eager Recharts on the mood-insights surface. The trajectory
forecast card pulled Recharts into the initial chunk of every
trajectory-eligible metric page though the main chart is already deferred;
split its fan band into trajectory-fan.tsx behind next/dynamic. The three
below-fold mood mini-charts (distribution / weekday / time-of-day) were
static imports under an already-dynamic hero line chart; defer them with
next/dynamic. Each loader paints a skeleton sized to the chart's own band
so the chunk arrives without a layout shift; the charts stay Recharts and
visually identical.
(cherry picked from commit 2b627893ab6b4d7bc457fc18e6e5585b048190e1)
The per-metric assessment cards are tightly grounded but their form had
converged: every card opened the same way, forced a "one doable step"
even when nothing was actionable, and a metric that stayed steady for
weeks produced near-identical paragraphs. Squeeze more value from data
already computed server-side, without touching the grounding floor or any
safety contract.
- Surface FDR-surviving cross-metric correlations on the per-metric card.
The discovery engine already screens behaviour×outcome pairs with a real
Pearson + exact p-value + Benjamini-Hochberg FDR control; that
intelligence only reached the period narrative. A new read-only consumer
runs the same full-matrix discovery and filters to the pairs that
involve the current metric, then feeds the engine's own conservative,
descriptive, never-causal interpretation strings into the prompt as
grounded relations. Best-effort: a correlation hiccup can never block the
generation it only decorates.
- Streak-aware repetition. Derive a steady-run length from the graded
series (consecutive recent weeks within a band of the user's own
baseline) and tell the model when it has already reported "no material
change" N times running, so it acknowledges the continuity in one clause
and pivots to a different facet instead of restating the same level.
- Stop forcing a step; add controlled variety. The closing step is now
conditional — when nothing is genuinely actionable the model affirms and
names one thing to watch rather than manufacturing the filler that lets
platitudes back in. A deterministic per-render variety token (seeded from
user + metric + day, never Math.random / Date.now) rotates the opening
angle so consecutive cards and days don't read identically, and the
phrasing temperature lifts 0.3 → 0.45 while the facts stay pinned by the
snapshot and the forbidden-phrase guards.
- Surface the computed data strength (n + recency) into the prose prompt so
it hedges on thin data instead of guessing what "few" means, matching the
UI confidence badge.
- Add locale-matched few-shot examples (a grounded assessment + a labelled
banned-filler counter-example) to the assessment system prompt, which
most helps weaker local-provider models follow the contract.
The own-baseline grounding, computed-not-hallucinated stats, schema-
enforced output, filler-phrase ban, and correlation-as-association
discipline are all preserved and re-pinned by new guard tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
(cherry picked from commit c9d04cab1523e0e57a3f50c87c9f88684b01b26f)
The tag × metric crosstab summed cumulative metrics (steps, active
energy, sleep duration) per day across every source with no canonical
pick, so once two sources reported the same day — Fitbit + Apple steps,
Fitbit + WHOOP sleep — the per-day total double-counted and inflated the
with/without averages and the Welch delta the surface ranks on. Run the
metric rows through the same per-day source-priority + device-type picker
the analytics steps/sleep path uses before bucketing, keyed by the
metric's priority ladder and the user's Berlin calendar day. Same-source
per-stage rows still sum into the night total; only cross-source twins
collapse. Thread the user's source-priority blob through the read so the
pick honours the configured ladder.
The cross-metric read was a flat `take: 5000` over the 365-day window
ordered `measuredAt desc`, unioned across per-stage sleep + per-sample
pulse/HRV; a multi-source, high-frequency user blew past it and silently
dropped the oldest months, biasing the crosstab and the long-window
correlations toward recent data. Lift the cap to cover the realistic
worst case for the window and annotate a wide event when the cap is hit
so a truncated read is observable rather than silently wrong. The read
stays raw rather than moving to the DAY rollup tier because that tier
buckets on UTC midnight while these aggregates key on the user's Berlin
day, which would skew the day pairing at the boundary.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
(cherry picked from commit aa72ffba6e3bcac9f97be4adc26d689d4f352e98)
Migration 0121 replaces the MedicationIntakeEvent unique constraint with a
partial unique index `WHERE deleted_at IS NULL`. A slot the user previously
deleted (tombstoned) no longer occupies the unique slot, so a re-take
re-creates the row cleanly instead of P2002-ing against the tombstone and
500-ing. The migration dedups any pre-existing duplicate live rows first
(keep newest, tombstone the rest) so the unique build can't fail on real
data.
Measurement and Workout deliberately stay full uniques: Measurement's
compound-key writes use prisma.upsert, which Prisma 7 compiles to native
INSERT ... ON CONFLICT and Postgres cannot arbiter against a partial unique
(and consolidate-daily-mean intentionally resurrects a tombstoned canonical
row, which needs the full unique); Workout has no deleted_at column. The
rationale is documented in the migration header and the schema comments.
Migration 0122 adds MoodEntry.externalId plus a NULL-distinct
`@@unique([userId, source, externalId])`. The moodLog webhook, the pull sync,
and the JSON import now carry the source's stable id into externalId and
upsert on the new key when present, so a re-import is idempotent even when
moodLoggedAt re-rounds or re-zones. Native/MANUAL entries keep their NULL
externalId and stay on the legacy (userId, date, moodLoggedAt) path.
Migration 0123 adds AuditLog(userId, action, createdAt desc) and
MoodEntry(userId, moodLoggedAt) composite indexes for the status-cache and
mood-insights read paths.
Also corrects the stale "v1.6.0 drops this column" note on
MedicationSchedule.daysOfWeek (the column is still dual-written and read by
live paths, so it is retained) and documents the UTC-only timezone storage
invariant in a schema header comment.
Integration tests cover the intake re-take after delete (no resurrection /
no 500) and mood re-import idempotency (with and without an external id) on
real Postgres.
(cherry picked from commit 9a6723f469bdec063a99cd08df132e07ec3c95a5)
Close the launch-blocking gap where the ConsentReceipt infrastructure was
built but never enforced as a precondition: a direct API caller with a valid
token could forward a health snapshot to a server-managed external LLM with no
receipt on file. Consent was enforced client-side only.
Add a consent guard (src/lib/ai/consent-guard.ts) and call it before the first
external-provider call on the server-managed path. The gate fires only when the
resolved chain could egress via the operator's global key (admin-openai); a
user's own BYOK key (openai/anthropic), their own ChatGPT OAuth account, and
the self-hosted local provider stay ungated — that egress is the user's own act
and the settings flow is its consent. The check fails closed: a BYOK-primary
chain with an admin-openai fallback still requires a receipt, since the runner
may cascade to the operator key. The surface maps to the consent kind (coach ->
ai_coach, insights -> ai_insights_only; master ai_full satisfies either).
- Interactive routes (insights/chat, insights/generate POST) throw
ConsentRequiredError; api-handler renders a 403 with
meta.errorCode = "consent.ai.required", mirroring assistant.disabled.* so
the client can render an inline grant-consent notice.
- Off-request pipelines fail closed quietly: comprehensive generation returns a
typed skipped:no-consent outcome, and the per-metric status family
(runStatusCompletion, gating every status card + the period-narrative warm +
the off-budget Coach memory workers) surfaces the no-key fallback instead of
egressing.
Also rate-limit POST /api/fitbit/sync (the one Fitbit route without a limiter):
a 5/60s baseline plus a tighter 1/hour bucket on the expensive fullSync walk,
keyed by user id, matching the sibling Fitbit routes.
Harden the Fitbit OAuth redirect_uri: assert it is an absolute https origin (or
http on localhost), lands on /api/fitbit/callback, and stays same-origin with
NEXT_PUBLIC_APP_URL — defence-in-depth behind Google's registered-URI check so
a Host-coerced or misconfigured app URL cannot redirect the authorization code
off-origin.
(cherry picked from commit 46b5fc2cea1b6e80bee0445bfdacf5790442c67d)
Three hardening + performance fixes to the Fitbit/Google Health sync.
Stop soft-delete resurrection. The measurement write probed the existing
row by the `(userId, type, source, externalId)` unique key without excluding
soft-deleted rows, so a reading the user deleted came back on the next hourly
sync (the upsert matched the tombstone and took the update branch — silently
undoing the delete). The write now probes only live rows (`deletedAt: null`)
and treats a tombstone as absent, minting a fresh insert instead. The fresh
insert relies on the partial unique index over `deleted_at IS NULL` to keep
live-row uniqueness; the tombstone sits outside it, and `skipDuplicates` guards
a live-row race.
Batch the backfill writes + collapse the rollup recompute. The write loop
was one upsert per reading (N+1) followed by a per-(type,day) DAY-rollup
recompute, so a multi-year backfill paid thousands of serial round-trips on a
concurrency-1 worker. The existence probe is now a single `findMany` over the
batch; fresh rows go through chunked `createMany`; only already-live rows take a
per-row update for their differing values. On a `fullSync` the per-write rollup
hook is deferred and the touched type-days collapse into one
`recomputeUserRollups(from, to)` pass at the end of the cycle. The incremental
path keeps the inline per-day hook. The `stats:` overwrite contract, the
single-watermark `markSynced`, and the all-403 soft-skip guard are unchanged.
Bound the hourly poll concurrency. The poll synced every connection in one
serial loop, so a single slow Google response stalled the whole cohort. The
cohort now fans out through a bounded `p-limit` pool with per-user error
isolation, extracted into `runFitbitPollCohort` so the contract is unit testable
without exporting worker internals. The accumulators are read after the await to
avoid a compound-assignment lost-update across overlapping pool tasks.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
(cherry picked from commit 8369af2471fb8da04a938b5a784433318d81a248)
Harden the native auth-token surface against four confirmed findings.
- /api/auth/logout was a no-op for a bearer credential: destroySession only
clears the cookie, leaving an Authorization: Bearer hlk_ token live until
expiry. When the request carries a bearer, also flip the matching
ApiToken.revoked and revoke its paired refresh-token sibling. Cookie path
unchanged.
- A browser spoofing X-Client-Type: native was handed a 60-day refresh token
in the JSON body. Gate refresh-token issuance on a new
isCookielessNativeCaller check (no Mozilla UA, no inbound session cookie),
so the secret reaches genuine cookie-less native callers only and never
alongside a browser session. The iOS native path is unaffected; a spoofing
browser falls back to the short-lived access token with no refresh.
- Refresh-reuse revoke trusted the stored row's deviceId, so a stolen token
replayed under a fabricated X-Device-Id confined the revoke to the
attacker's own id and left the victim's family live. On a present-but-
mismatched deviceId, escalate to the user-wide family revoke (a real device
never changes its id mid-family). The matching-id and null-id cases keep
their existing scope.
- Tighten the auth:refresh rate limit from 60/15min to 10/15min, aligning the
high-value rotation endpoint with the passkey-verify tier.
Add coverage for each: logout revokes bearer + sibling, a cookie/Mozilla
caller never receives a refresh token, a device-mismatch replay triggers the
wide revoke, and the cookie-less native path still gets its refresh token.
(cherry picked from commit 14dcdefc5f40855305d8aaf573d073faebc61476)
Use FITBIT_ACTIVITY_PAGE_SIZE for the sleep and exercise reads instead of a
magic 25, remove the unused FitbitDataTypeKey type, and declare the crosstab
day floors independently so they aren't a duplicate re-export of the influence
floors.
Snapshot the incremental watermark once per sync cycle and thread the same
lower bound to every resource (metrics, activity, sleep, workout), stamping
lastSyncedAt once at the end. Previously each resource read lastSyncedAt then
stamped now() independently, so the first resource advanced the watermark and
later resources only fetched the last overlap window — silently dropping the
gap after an outage longer than the overlap. The full/backfill path is
unaffected (it passes no lower bound).
Reject an OAuth callback whose token response carries no refresh_token instead
of persisting an empty string. An empty refresh token bricks every future
token refresh; the callback now errors out cleanly so the user re-consents
rather than landing a dead connection.
Model the INTERVAL data types (steps/distance/calories/floors daily totals,
sleep and exercise sessions) on Google Health's interval.start_time /
interval.civil_start_time anchor instead of a sample_time / bare date. The
incremental filter for sleep/exercise previously targeted
sample_time.physical_time, which 400s/empties for interval types and stalls
incremental sync. Daily-summary metrics and spot readings keep their existing
date / sample anchors. Shape mismatches still fail safe (return []/null).
Mark the Google Health (Fitbit & Pixel) settings card as experimental with a
badge and a short note that data coverage is still being verified, in all six
locales.
(cherry picked from commit 3f116802149eebe9144729e2779f7f4405f248f2)
Thread the acting user id through createTagLinks, replaceTagLinks, and
replaceRatedFactorLinks, and assert the target MoodEntry belongs to that
user before writing any link. The check runs against the same client (tx
or singleton) as the write, so it shares one snapshot. Behaviour is
unchanged for the existing POST and bulk paths, where the entry is freshly
created and owned; the guard closes a latent cross-user link hazard for
the planned PATCH mood-edit route. A mismatched or missing entry throws
MoodEntryOwnershipError before any catalog or join touch.
Add IF NOT EXISTS to every ADD COLUMN in migration 0119 so a clean replay
is idempotent, matching the 0117 convention.
(cherry picked from commit 1bf983f6c2c5e44796c99983d57c1b1d90cd7099)
Extend the mood-relations surface from mood to a health metric. For each
structured mood tag present on enough days, compare a metric's value on
tag-present vs tag-absent days: active energy on workout days (same day),
sleep length on nights tagged sleep (same day), and next-day recovery
after a tag like alcohol or food (D to D+1 lag). The comparison reuses the
existing statistics engine wholesale — the Welch two-sample t-test, the
per-group day floors, the confidence band, and the Benjamini-Hochberg FDR
control already used by the influence and discovery boards — so no new
statistics code lands. Only pairs that clear p < 0.05 and the FDR control
surface, ranked by adjusted q then absolute delta and capped at eight rows.
The result rides on GET /api/mood/insights as a new additive
`tagMetricCrosstab` field and renders as a confidence-gated, observational
card on the mood insights page, framed as an association in your own data
rather than a cause. The endpoint is server-rendered and not in the OpenAPI
contract, so no contract change is required.
Deprecate the standalone moodLog integration. Mood is now tracked fully
inside HealthLog through entries, structured tags, and rated factors, so
the external webhook and reverse-sync bridge no longer add anything. The
webhook route, the push and pull helpers, and the validation schemas carry
an @deprecated note pointing at the native path; the surface stays
functional this release and is slated for removal in a future major. The
settings card shows a deprecation hint, localised across all six locales.
Extract the Coach chat surface (header actions, message thread,
composer, provenance rails, settings, mobile trays) into a shared
<CoachConversation> component so the drawer and a new full-page route
render the exact same implementation — no forked chat logic. The drawer
keeps only its <Sheet> chrome (close button + maximize control); the
page supplies a minimize-back-to-drawer control.
The maximize control in the drawer header navigates to /insights/coach
and closes the drawer (aborting any in-flight stream); the page's
minimize control routes back to /insights and reopens the drawer through
the shared launch context. Both surfaces preserve incremental streaming,
collapsible history + provenance, the single composer disclaimer, the
auto-grow composer, refusal / budget handling, and the aiMode gating —
the full page redirects to /insights when the operator flag is off or
the user has opted out.
Lift useResettableValue / nextResettableValue into their own module so
the shared surface seeds its composer from the same controlled-prefill
contract without importing the drawer shell. Register the new route in
the coach feature-flag and per-user-disable gate-site invariants, and
add the maximize / minimize labels across all six locales.
A weekly injectable (e.g. a GLP-1 dose) with real recorded intakes
reported 0% compliance on the medication card across every window
(7 / 30 / 90 days). The web reads the server-computed rate verbatim,
so the server value itself was wrong.
The intake-to-slot matcher used a fixed +/-12h pairing radius. That is
correct for a daily cadence (24h inter-slot gap) but far too tight for a
once-weekly dose: a real intake is rarely logged within 12h of the
schedule's configured time-of-day, because the user takes the shot on
whichever day and time of the dosing week suits them. Those intakes fell
outside the radius, every weekly slot read "missed", and the rate
collapsed to 0% while the matching intakes were orphaned.
Scale the pairing radius with the cadence gap. `pairDoses` now derives a
per-slot radius from half the distance to its nearer neighbouring slot
(so two adjacent slots never double-claim one intake), floored at the
12h base so daily and multi-dose-daily cadences keep their exact prior
behaviour. `buildCadenceTimeline` additionally probes the schedule's
intrinsic cadence over a padded window and passes half the minimum
inter-slot gap as a radius floor, so a window holding a single expected
slot (a weekly med over a 7-day window) widens correctly too.
A weekly cadence now pairs an intake logged anywhere in the dosing week
to that week's slot, reporting the true rate. Daily-cadence matching is
unchanged.
Adds unit coverage for the widened radius and a real-Postgres integration
test mirroring the weekly-injectable row + off-time intakes.
Extend the structured mood taxonomy so a tag can carry a magnitude. A
rated factor is a MoodTag of kind RATED that the user scores per entry;
the score persists on MoodEntryTagLink.rating. Binary tags keep
kind BINARY and a null rating, so every existing row and read path stays
byte-identical.
Migration 0119 is additive: four marker columns on mood_tags (kind,
scale_min, scale_max, inverse), a nullable rating on mood_entry_tag_links,
and a seeded factors category with five MVP factors (work, social, sleep
quality, stress, conflict). Stress and conflict are inverse-scaled
(higher means worse); conflict is a 1..2 Yes/No factor.
Write contract: POST /api/mood-entries and the bulk route accept an
optional ratedFactors: [{ key, rating }] array parallel to the binary
tagKeys. The resolver builds the join rows field-by-field, drops unknown
and non-RATED keys silently, and rejects a rating outside the factor's
own scaleMin..scaleMax (422 on the single-entry path, per-entry skipped
on bulk). Create and list responses surface the persisted factors split
from binary keys so a client hydrates without a refetch. GET
/api/mood/tags now emits kind, scaleMin, scaleMax, and inverse so a
client can render factors versus tags.
This is the ingestion half; analytics over rated factors follow later.
Factor labels ship in all six locales.
