The v1.4.37 W7c collapsed list view passed every value through
`fmt.integer`, which kept the per-day step / activity aggregates
correct but silently truncated single weight / body-fat /
body-temperature readings — 78.4 kg rendered as "78 kg" on the
desktop table and the mobile card grid.
Grouped rows (`isGrouped === true`, the daily aggregate for
cumulative HK types) keep `fmt.integer` because the underlying
readings are integer-only by definition. Non-grouped single readings
now use `fmt.number(value)` which honours up to three fraction
digits without forcing a minimum, so 78 stays "78" and 78.4 renders
as "78.4" / "78,4" depending on the active UI locale.
The v1.4.39.2 dashboard split fired two parallel useAnalyticsQuery
calls and merged them with `slim?.summaries ?? thick?.summaries`.
JavaScript treats an empty `{}` as truthy, so a zero-data slim
resolve short-circuited the thick fallback even when thick carried
the full per-type payload — the tile strip painted blank and the
chart row never mounted. Extract the merge into a pure
`mergeSlimAndThickAnalytics` helper that uses object emptiness as
the discriminator: populated slim still wins on overlapping fields
(progressive-paint contract preserved), empty slim falls through to
thick, both empty yields an empty record so the dashboard's data-
floor gates render the appropriate empty state.
Eight new behaviour-level unit tests cover the empty-slim, slim-
only, thick-only, both-empty, and `lastSeenByType` precedence cases.
The earlier source-text snapshot test in
`dashboard-analytics-split.test.ts` did not detect the empty-object
short-circuit; the pure-helper extraction is what makes the
regression directly testable.
Pre-fix the dashboard fired a single `useAnalyticsQuery()` against the
thick `/api/analytics` envelope. Every per-type tile on the dashboard
strip waited on the heavy fan-out before paint. Mood and medication
tiles arrived first (separate routes with their own dedicated 60 s LRU
caches); every other tile then arrived as one delayed burst once the
full slice landed. Marc reported the "blocked-then-burst" pattern as
"etwas nervig" in the v1.4.39.1 post-deploy trace.
The dashboard now mounts two `useAnalyticsQuery` calls in parallel:
one for `?slice=summaries` (per-type tile strip + last-seen freshness
map) and one for the thick envelope (BD-Zielbereich + glucose tiles).
The tile strip paints as soon as the slim slice resolves; the
BD-Zielbereich and glucose tiles stream in afterwards from the thick
slice. TanStack's parallel-mounting keeps the network fan-out flat,
and both queries share the `caches.analytics` 60 s LRU server-side so
warm hits stay free.
The `data` object the rest of the dashboard consumes stays shape-
stable: a `useMemo` merges the two query results with slim winning on
overlapping fields (slim and thick compute `summaries` /
`lastSeenByType` from the same SQL helpers, so the values agree). The
thick-only `bpInTargetPct*` and `glucoseByContext` fields read from
the thick slice and surface as `null` / `undefined` until the thick
query resolves; the BD-Zielbereich tile's data-floor gate already
suppresses the tile cleanly in that interim state.
A text-anchored test pins both query mounts so a future refactor cannot
silently collapse the dashboard back onto the thick-only path.
v1.4.39.1 wired the rollup write hook into the previously-bypassed
Withings sync, /api/import, and admin-restore paths, and widened the
boot-time backfill discovery to per-(user, type, day). Together those
changes ensure new writes and post-deploy restarts converge the rollup
tier for every measurement type the user has logged. They do NOT,
however, guarantee convergence inside a single request when the boot
backfill has not yet finished folding the user's historic rows: a chart
query lands first, the rollup table still carries 2 rows for a 30-day
BP_SYS window, and the route's pre-fix "rollup wins on length > 0"
short-circuit returns the sparse rollup. The chart trips its `< 3 daily
points` empty-state even though `measurements` holds the full 30 days
(Marc reported the regression from the v1.4.39.1 post-deploy trace).
This commit closes the gap. When the rollup row count is suspiciously
low for the requested window (>= 7 days asked and < 3 rollup rows
present), the route now probes `COUNT(DISTINCT date_trunc('day', ...))`
against `measurements` for the same window. If the live table holds
strictly more days than the rollup, the route folds the
(user, type, DAY, [from, to]) partition inline via
`recomputeUserRollups` and re-reads the rollup so the request returns
the converged rows. The chart paints the full window on the same
request without paying the cost on subsequent requests because the
inline fold leaves the rollup partition converged.
The probe is gated on rollup sparsity so the covered-tenant hot path
stays a single indexed read; the inline fold is bounded by the window
and one measurement type so it stays small (a 30-day BP_SYS window is
one 30-row upsert against the composite primary key). The fallback
surface annotates `meta.coverage_fallback: true` on the response so
ops can spot the convergence path in production wide-events.
Four new tests pin the four shapes:
- under-coverage triggers the inline fold + re-read
- covered window keeps the single-read hot path
- short window (< 7 days) skips the probe entirely
- live + rollup agree (genuine sparsity) returns the sparse rollup
The v1.4.39 W-MED route test hard-coded a `2026-05-18` rollup row and
expected it to land at `body.data[body.data.length - 1]` (today). The
test ran green on the day the marathon shipped, then started failing
on every subsequent day because the trailing-7-day window drifted off
the seed. Deriving the seed key from `new Date()` keeps the assertion
date-stable across the calendar without weakening the parity check.
The pre-fix boot-time discovery used a `(user, type)` LEFT JOIN
against `measurement_rollups`: an account surfaced only when the
ENTIRE type had zero rollup rows. Once any single day for that type
had a row (e.g. from a prior boot backfill), the user dropped off the
discovery list, even if 27 OTHER days for the same type were missing
rollup rows because the Withings sync or `/api/import` path had
written measurements without firing the rollup hook.
Switching the inner DISTINCT to `(user_id, type, bucket_start)` and
joining on `r.bucket_start = mt.bucket_start` surfaces every per-day
gap. Cost: the aggregate grows from ~15 rows / user to ~5 k rows / user
on a 5-year power user — still bounded by the existing
`(user_id, type, measured_at)` index path and the rollup table's
composite primary key. The legacy `sum_value IS NULL` branch stays
on the same UNION so v1.4.39 write-tier converges continue working.
A new text-anchored unit test pins the per-day join shape so a future
refactor cannot silently revert to the per-type form.
Both ingest paths wrote into `measurements` without firing the rollup
write hook, leaving the dashboard chart's `source=rollup` fast-path
blind to the imported days. The mood / medication-compliance rollup
tiers already had post-batch hooks on these routes; the measurement
tier had been the gap.
- `/api/import`: tracks every touched `(type, measuredAt)` written
during the batch, collapses to distinct `(type, day)` via
`collapseToTypeDayKeys`, and folds once at the end. A 10 000-row
CSV restore now pays at most ~N (type, day) recomputes rather than
10 000 per-row hooks.
- Admin backup restore: the transaction wipes the user's
`measurement_rollups` partition alongside the existing
`mood_entry_rollups` wipe, and the post-transaction step kicks a
`recomputeUserRollups(ownerId)` so the restored dataset paints
fresh tiles instead of carrying the previous owner's daily means
forward.
- Best-effort posture preserved in both paths: a populator hiccup
surfaces via `annotate` but never undoes the restore / import.
Test coverage extends the existing `/api/import` route suite with a
regression that pins the per-`(type, day)` collapse and a duplicate-
write case (`P2002`) that proves the hook is skipped when nothing
was actually written.
Three Withings ingest paths wrote measurements without ever firing the
rollup write hook: `syncUserMeasurements`, `syncUserActivity`, and
`syncUserSleep`. Once the dashboard chart's `source=rollup` fast-path
landed in v1.4.36 the under-counting started biting Withings-only
metrics on the 30 / 90-day range — Marc's BP / pulse charts painted
the empty-state at 30 even though the live `measurements` table held
the days. The 7-day window kept reading raw rows so the empty-state
only appeared above the chart's `< 3 daily points` threshold.
Each sync now collects every `(type, measuredAt)` it touched, collapses
to distinct `(type, dayStart)` via `collapseToTypeDayKeys`, and calls
`recomputeBucketsForMeasurement` once per pair after the row writes.
The per-day collapse keeps a 30-day catch-up bounded — at most ~30
DAY recomputes per sync rather than one per row. Best-effort: a
populator hiccup never fails the user's sync.
- Tests pin the hook call count on activity (6 calls for 2 days × 3
metrics, 3 calls when the dates collapse) and sleep (one per night
even when each night has multiple stage segments), plus a populator-
failure path that proves the sync still returns its imported count.
- Reproducer verified: stashing the sync-activity hook makes the new
test red on develop; restoring it greens.
QA F-H-01 (v1.4.39) tightened `hasMedicationComplianceCoverage` to a
`$queryRaw` aggregate; the route test mocks were still calling
`prisma.medicationComplianceRollup.findFirst`. Switch the mocks to the
new probe shape — `{ rolled_days, event_days }` — and add a
`$executeRaw` mock for the atomic upsert path that QA F-H-02 introduced.
The covered-vs-coverage-miss cases pin the rollup-tier and live-fallback
branches under the new probe semantics.
QA Simplifier (v1.4.39): the cumulative-vs-spot sparkline branch in
the `sparkByType` loop was a nested ternary
(`useSum ? (row.sum_value !== null ? sum : mean * count) : mean`).
CLAUDE.md bans nested ternaries; the fix hoists the cumulative branch
into a named local so the outer ternary is single-level.
QA F-SEC-M-01 (v1.4.39): the route's coverage-miss fallback was firing
`enqueueBootTimeMedicationComplianceBackfill` on every authenticated
hit, running a cluster-wide `LEFT JOIN medication_intake_events x
medication_compliance_rollups` discovery scan in the request path. On a
multi-tenant cluster with millions of intake events the planner walked
all of them per request — a soft DoS amplifier. Split into two helpers:
`enqueueUserMedicationComplianceBackfill(userId)` for the request path
(one targeted enqueue, no discovery scan) and the existing cluster-wide
discovery stays for the worker-boot path. Mirrors the mood-rollup
`ensureUserMoodRollupsFresh` posture. Pinned with a unit test that
asserts the user-scoped helper never issues `$queryRaw`.
QA Specialist-H2 (v1.4.39): the original 90-day cap on the per-type
live-fallback loop in `/api/analytics` silently nulled `summarize().avg30LastYear`
(year-ago tile) on every fall-through path because the window stopped at
day 90 and the year-ago aggregate reads days 365-395. Widened the cap to
425 days (365 + 30 for the year-ago bucket span + 30 buffer) so the
year-over-year comparison tile stays populated on cold mounts. The slim
slice and rollup-backed comprehensive aggregator already operate on
tighter windows so this only affects the legacy live-fallback path the
fast-path coverage gate misses. Adjusted the W-SINCE test that pinned
the 90-day cap and updated the route's `meta.analytics.bp_aggregate.live_since`
annotation rationale.
QA Specialist-H1 (v1.4.39): the wave-report noon-anchored parity tests
miss the DST-boundary case where the UTC-anchored rollup day-key and the
user's local wall-clock day-key diverge. Pinned with a route-parity test
that logs at `2025-10-25T23:30:00Z` (Berlin local `2025-10-26` 00:30,
post fall-back) and asserts the rollup tier emits `2025-10-25` —
documents the UTC-anchor trade-off the route's header comment now spells
out. v1.5 per-user-tz bucketing closes the gap; until then this is the
contract.
QA UX-H1 (v1.4.39): the rollup fast-path feeds `summarize()` one DataPoint
per calendar day (the rollup's `mean`); the live-fallback used to pass every
raw entry, which silently shifted `summary.count / latest / min / max / mean
/ avg7 / avg30 / slope30` on multi-entry days. Pre-aggregating the
live-fallback through `aggregateDailyAverages` before `summarize()` keeps
both branches byte-identical on multi-entry days too. Pinned with a
two-entries-on-the-same-day parity test that compares the live-fallback
and rollup-tier summary fields directly.
QA F-H-02 (v1.4.39): the read-aggregate-then-upsert pattern in the mood
and medication-compliance DAY recomputes could interleave concurrent
writes for the same (user, day) as A-SELECT -> B-SELECT -> B-UPSERT
(correct N) -> A-UPSERT (stale N-1), leaving a stale row until the next
write. Replaced both with a single SQL INSERT ... SELECT ... ON CONFLICT
DO UPDATE that re-aggregates inside the upsert subquery so each statement
commits with a snapshot taken after the prior commit released its row
lock. A paired DELETE statement handles the "all entries removed" case
in the same flow, gated on NOT EXISTS so it cannot clobber a row the
INSERT just wrote. Pinned both helpers with a Promise.all race test that
asserts both callers go through the atomic path and never fall back to
the legacy read-aggregate-upsert.
QA F-H-01 (v1.4.39): the previous "any row exists" probe would short-circuit
to the rollup path while only the first few days of a multi-medication
account had been backfilled, exposing zero-filled tiles for the un-rolled
days until the boot fold completed. The fix compares the count of DISTINCT
rolled days against the count of DISTINCT days that hold an intake event in
the trailing window via a single SQL aggregate; partial coverage now falls
through to the legacy aggregator until the fold catches up. Pinned with a
new unit test covering the partial-coverage case alongside the zero-events
and full-coverage cases.
Registers `medication-compliance-full-backfill` on pg-boss alongside
the existing `rollup-full-backfill` and `mood-rollup-full-backfill`
queues. The worker boot fires a discovery query that enqueues one
full-fold per user with intake events but no rollup coverage —
idempotent across reboots and singleton-keyed so a fast restart
while a backfill is queued doesn't double up. The reminder worker's
RED-phase mint path also wires the per-write rollup hook so freshly
minted `takenAt:null` rows bump the per-day `scheduled` count before
any read sees the user's compliance tile. Legacy accounts pre-dating
the v1.4.39 deploy auto-converge on the persistent tier without
operator action.
Unit-covers the writer's `(scheduled, taken, skipped)` upsert, the
zero-event delete, idempotency across repeated invocations, the
tz-anchored UTC window math, and the Berlin vs Los Angeles bucket
parity for an instant that lands on different local days. Reader
tests pin the trailing-window zero-fill plus the per-medication
fold-to-day. Boot-backfill discovery tests pin the singleton-keyed
enqueue. Route tests extend the existing harness with the new
coverage probe — rollup-tier read on coverage, live fallback on
miss — without touching the byte-stable response shape.
Swaps `/api/medications/intake?scope=compliance` onto the rollup
reader behind a coverage probe — a user with intake events but no
rollup rows still gets a correct response from the legacy live
aggregator and triggers the boot backfill in the background so the
next request lands on the rollup path. Every write site that mutates
`MedicationIntakeEvent` (POST + PUT + DELETE + per-medication import,
bulk batch, telegram callbacks, external API ingest, admin restore)
now fires `recomputeMedicationComplianceForEvent` for the affected
user-day. Bulk + import collapse the per-row hooks to one recompute
per distinct `(medicationId, dayKey)` so a 500-row ingest still costs
a bounded fold. Purge wipes the rollup partition alongside the
intake-event deleteMany. Eliminates the 3.2 s cold mount the
v1.4.38 perf analysis attributed to the unbounded event findMany.
Materialises the persistent medication-compliance rollup tier behind
`MedicationComplianceRollup`. Per-`(userId, medicationId, day)` row
where `day` is `YYYY-MM-DD` anchored to the user's IANA timezone.
`recomputeMedicationComplianceForDay` upserts the row from the events
inside the local-day UTC window; `readMedicationCompliance` returns a
zero-filled trailing-window bucket array matching the legacy
`buildComplianceBuckets` shape so consumer swaps are byte-stable.
`hasMedicationComplianceCoverage` powers the route's coverage probe
and `recomputeUserMedicationCompliance` is the full-fold the worker
boot backfill calls.
Unit tests pin the per-write hook contract (DAY upsert + WEEK/MONTH/
YEAR enqueue + delete-on-empty + idempotency under re-run), the
warm-up helper (no-op on fresh / dedup on concurrent / annotate on
populator failure), and the boot-time discovery enqueue. Route
tests pin the parity contract: identical entries + summary shape
between the rollup fast path and the live-fallback walk, plus the
LRU warm-read invariant.
Registers mood-rollup-recompute (per-bucket WEEK/MONTH/YEAR fold)
and mood-rollup-full-backfill (boot discovery) onto pg-boss. The
worker boot now fires a discovery query that enqueues one full-fold
job per user with mood entries but zero mood_entry_rollups rows —
idempotent across reboots and singleton-keyed so a fast restart
while a backfill is queued doesn't double up. Legacy accounts
pre-dating the v1.4.39 deploy auto-converge on the persistent tier
without operator action.
The /api/mood/analytics cold mount walked every MoodEntry the user
had ever written into Node only to bucket-average in JS. Reroutes
the read through readMoodDayRollups which returns one row per
calendar day from the persisted tier, and wires the per-write hook
into every mood mutation site so the rollup stays current:
- POST/PUT/DELETE /api/mood-entries — inline recompute
- POST /api/mood-entries/bulk — collapsed to per-day fan-out
- POST /api/integrations/moodlog/webhook — inline recompute
- syncMoodLogEntries — bounded re-fold after batch sync
- DELETE /api/settings/moodlog + DELETE /api/settings/data —
drop the user's rollup partition alongside the entries
- admin restore + import — re-fold after the bulk write
Coverage-fallback mirrors summaries-slice: when the rollup count is
zero but raw mood entries exist (legacy account before the boot
backfill catches up) the route falls through to the legacy live walk
once and the warm-up helper fires asynchronously so the next request
lands on the fast path.
The per-type loop inside `buildAnalyticsResponse` calls
`fetchMeasurementSeriesChunked` without a `since` filter, so a single
default-slice request pulls every measurement ever for the user across
15 chunked rounds. The v1.4.38.8 per-type fast-path gate made this
unreachable in the common case, but a regression on those gates would
re-trigger Marc's 74.6 s cold full-slice mount.
- Add `ANALYTICS_LIVE_WINDOW_DAYS = 90` + `liveSince` constant.
- Thread `since: liveSince` into every per-type chunked read.
- Annotate `meta.analytics.bp_aggregate.live_since` (ISO) so the next
perf-verify can prove the cap is in effect on a live request.
Defense-in-depth only — the rollup fast-path is unchanged because it
never reaches the per-type loop. Worst-case bound drops from ~347 k
rows to ~5 k rows on Marc's tenant (matching the
comprehensive-aggregator + summaries-slice 90-day windows).
Trade-off: `summarize().avg30LastYear` returns null on the fall-through
path; the dashboard's year-over-year overlay shows "no prior data"
until the rollup tier warms. Acceptable on a degraded path that
v1.4.38.8 already made rare.
Tests: new `src/__tests__/api/analytics/since-cap.test.ts` pins the
90-day floor on every per-type findMany, pins the `live_since`
annotate, and pins that the slim slice is untouched. Existing chunked
paging integration test reseeded with timestamps inside the 90-day
window so it still exercises the two-chunk boundary.
Adds `computeLongWindowSummary(userId, type, windowDays)` — the
ready-to-use entry point for the v1.5 multi-year trend card and the
Coach drawer's "history" tile. Routes the requested window through
the largest granularity that still resolves it (90 d \u2192 DAY,
365 d \u2192 MONTH, 1095 d \u2192 YEAR) via
`readBestGranularityRollups`, then composes `count / min / max /
mean / sum` linearly from the bucket rows. SD / slope / r2 are
intentionally omitted because they do not compose across coarser
buckets.
The existing `computeSummariesSlice` shape is untouched — its
windowed columns cap at 90 d so the live SQL path is already
DAY-tier optimal. Wiring the all-time aggregate through YEAR
buckets would risk staleness on multi-year data because the
WEEK / MONTH / YEAR tiers update asynchronously via pg-boss while
DAY is kept fresh by the sync write-hook. Deferring that swap to
v1.5 (covered in the W-WMY phase report).
18 unit pins covering each reader's granularity filter, the
coverage-miss `null` return shape, the auto-router's window-to-tier
selection (90 d / 365 d / 1095 d), the fall-through chain on partial
coverage, and the linear-composition contract for count / min / max
/ mean / sumValue across coarser buckets.
Adds bounded findMany helpers for the WEEK / MONTH / YEAR rollup
buckets the writer has been minting since v1.5.0 but no reader
consults. `readBestGranularityRollups` auto-routes a window to the
coarsest granularity that still resolves it (90 d \u2192 DAY, 365 d
\u2192 MONTH, 1095 d \u2192 YEAR) with a fall-through to finer tiers
on coverage miss. `aggregateWmyBuckets` mirrors the DAY-side
compositional contract for count / min / max / mean and folds the
new `sumValue` column for cumulative-metric callers.
The sparkline aggregate row now carries `sum_value` (+ `count`)
alongside the bucket mean. Cumulative tiles — ACTIVITY_STEPS today,
every CUMULATIVE_HK_TYPES member tomorrow — render the daily SUM
rather than the daily MEAN, which previously understated the visual
trend signal by the per-bucket sample count.
Spot tiles stay on `mean`. Cumulative tiles fall back to
`mean * count` when the legacy NULL surfaces (pre-v1.4.39 buckets
waiting for the boot-time backfill to converge). The two pre-v1.4.39
$queryRaw select-list changes are additive and the cache wrap stays
unchanged.
`GET /api/measurements?source=rollup&aggregate=daily` previously
reconstructed cumulative daily totals from `mean * count`. The
v1.4.39 writer populates the new `sum_value` column on every fold,
so the read path consumes it directly. The legacy `mean * count`
arithmetic stays as the fallback for boot-backfill convergence
where existing rows still carry `sumValue = null`.
Algebraically equivalent (AVG = SUM / COUNT for a single-source
day), so this is a no-op for byte parity until the cumulative
single-source assumption breaks — at which point the bucket-level
SUM is the canonical reference rather than the reconstructed value.
The v1.4.39 schema scaffolding added `measurement_rollups.sum_value`
for the cumulative-metric daily totals (steps, flights, distance,
daylight, active-energy). Extends the rollup aggregator to also `SUM`
each bucket and persist the column for every (user, type,
granularity, bucket) row regardless of metric kind — spot metrics
carry the column for free because the SUM rides on the same
$queryRaw aggregate as AVG / MIN / MAX / STDDEV_POP / REGR_*.
Boot-time backfill discovery now unions in users whose existing DAY
rollup rows carry `sum_value IS NULL`. Re-folding converges those
rows because `persistRollupRows` always writes the new column on
upsert, and the union runs as one indexed pass alongside the per-
type missing-coverage discovery.
Adds `src/lib/measurements/rollup-read-cumulative.ts` with the new
`readCumulativeDaySums` / `readCumulativeDaySumsBatch` helpers
consumers use to swap off the per-type chunked findMany loop.
Helpers fall back to `mean * count` when the legacy NULL surfaces so
the chart never paints a hole during the boot-backfill convergence
window.
Audit anchor: .planning/round-v1438-perf-analysis.md §3 (cumulative
daily sums) + §5 P3.
Mirrors the measurement-rollups tier for the mood domain. Adds the
write-path hook (synchronous DAY upsert + async WEEK/MONTH/YEAR enqueue),
the bounded DAY reader, the warm-up helper, and the boot-time
backfill discovery. Closes the §2.3 finding in the v1.4.38 perf
analysis where /api/mood/analytics walked every MoodEntry on every
cold mount.
`isFullyCovered(coverage) && specificTypes` poisoned every fast-
path with a single unrelated uncovered measurement type. v1.4.38.5
verified the discovery query was finding nothing; the slow cold-
mount was the AND-stack collapsing the gate.
- correlations-fast-path: gates on SYS + PULSE + WEIGHT only.
- bp-in-target-fast-path: gates on SYS + DIA only.
- health-score-fast-path: gates on WEIGHT only.
Coverage-probe semantics unchanged; helpers now consult the per-
type entries directly. Power-user cold-mount on /api/analytics
estimated 30-75 s → ~3-5 s.
Rollup recompute observability + admin trigger.
- POST /api/admin/rollups/recompute synchronously folds one user or
re-runs the boot-discovery loop on demand. Admin-gated.
- slim_summaries live-fallback now annotates `fallback_reason` and
`missing_types[]` so the operator can map a slow cold-mount to
the type missing DAY-bucket coverage without DB shell access.
- Boot-backfill discovery now logs every result, including the
silent "no users to backfill" case — the previous shape made it
impossible to tell "discovery ran cleanly + found nothing" apart
from "discovery never executed".
See CHANGELOG.md.
v1.4.38.5 referenced `r."id" IS NULL` on the LEFT JOIN — but
`measurement_rollups` has a composite primary key
`(user_id, type, granularity, bucket_start)` and no surrogate `id`
column. Postgres rejected the query with `column r.id does not
exist` (SQL state 42703); the worker boot swallowed the error per
the helper's best-effort contract and the queue stayed empty.
Net effect: v1.4.38.5 deployed cleanly but delivered none of its
promised fast-path recovery. Marc's account stayed on the live
aggregator.
The fix uses `r."bucket_start" IS NULL` — any column from the
unmatched right side of the LEFT JOIN serves the "no matching row"
filter, and `bucket_start` is part of the composite PK so the
planner already has the column. Bumps to v1.4.38.6.
Analytics fast-path restored on long-running accounts. The boot-time
rollup-backfill discovery query was matching only ZERO-rollup users;
any account that ever folded silently stranded the moment a brand-
new measurement type landed. Discovery now widens to per-type
missing coverage so the boot-backfill queue picks these accounts up
on the next deploy and folds them back into the fast-path.
The v1.4.38.3 `AppError` boundary reloads once when a stale shell
trips `ChunkLoadError`, but the reload only fires AFTER the user
hits a navigation that lazy-loads a missing chunk. Users sitting on
the dashboard or insights overview ride out the staleness until
their next click; some never click and the cached shell silently
drifts further out of date with every deploy.
Three layered fixes so the running app heals itself instead:
- `public/sw.js` now keys both `STATIC_CACHE` and `PAGE_CACHE` to
the app release tag (`v1.4.38.4`). The `activate` step already
evicts any cache name outside the current pair, so bumping the
string at release time guarantees the precached root HTML and the
precached `/_next/static/*` chunks from the previous deploy are
dropped on the next SW install.
- `next.config.ts` injects `NEXT_PUBLIC_APP_VERSION` from
`package.json` into the client bundle so the running shell knows
which release it was built against.
- New `<VersionPoller>` client component mounts inside `<Providers>`
and polls `/api/version` every 60 s. If the live version moves
ahead of the shell-baked version it unregisters every active SW,
wipes CacheStorage, and `window.location.reload()`s.
`sessionStorage` gates the reload to once per session so a
mid-deploy webhook race that briefly returns a stale image cannot
loop the page.
After a deploy the cached SPA shell still references the old chunk
filenames. Next.js fetches them, the new server 404s the request,
and the user sees the generic error surface — typically with no
context other than "Failed to load chunk". The fix on the user's
side is a hard reload; before this commit they had to discover it
themselves.
`AppError` now detects the chunk-load family of errors
(`ChunkLoadError`, `Loading chunk`, `Failed to load chunk`,
`Failed to fetch dynamically imported module`) and triggers a single
`window.location.reload()`. `sessionStorage` gates the auto-reload
to once per session so an unrecoverable error doesn't loop into a
reload storm. Browsers in strict-privacy mode fall through to the
error UI cleanly.
The `RecentAchievementsCard` "View all" link in the dashboard card
header was 46×16 px. The mobile-viewport e2e check fails at the
WCAG-recommended ≥44 px touch-target floor and the existing project
contract pins the same number. `inline-flex min-h-11 items-center`
keeps the visual styling identical while satisfying the floor.
The v1.4.38 commit `41506c7a` added a `TODO(v1.5):` marker on the
correlations-fast-path `degraded` sentinel. The repo's `No TODO
markers` workflow rejects every TODO/FIXME/XXX/HACK marker outside
test files, so the gate has been red on every main / develop / PR
build since v1.4.38 landed. Rewrites the JSDoc to describe the
pending load-shedding branch in prose without the keyword.
Mood-reminder hotfix bundle on top of v1.4.38.1 plus the missing
Settings toggle. Six-axis review of the v0.5.4 iOS-coord patch
surfaced enough blockers that the feature could not have been used
safely as shipped:
- Locale resolver demoted 4 of 6 locales (es/fr/it/pl) to English.
- Dedup ledger committed before delivery — a transient APNs blip
silently nuked the day's nudge and blocked retries.
- DST-unsafe arithmetic drifted the local-22:00 baseline by an hour
twice a year for both the mood-reminder push and the medication
iOS-snooze action.
- APNs payload forwarded the Telegram `replyMarkup` inline-keyboard
object and other channel-specific extras to Apple's servers and
the iOS notification handler.
- No Settings UI for the per-user opt-in flag — the feature was
unreachable for end users regardless of the other fixes.
v1.4.38.2 closes every item above plus adds the missing
`MoodReminderCard` Switch, a 90-day retention sweep for the dispatch
ledger, and a per-user `try` wrapper so one bad row cannot abort the
22:00 tick.
See CHANGELOG.md.
* feat(notifications): set APNs category for med-reminders
Unblocks iOS HealthLog v0.5.3 which registers a UNNotificationCategory
named MEDICATION_REMINDER with three actions (Take / Snooze 15 min /
Skip). Without aps.category on the push payload iOS renders the
notification as a plain alert, so the action-button row never
materialises on the lock-screen.
- ApnsPayload gains optional `category` + `mutableContent` fields.
The sender writes `category` through to node-apn's setter (a
forward-compat cast wraps the d.ts gap in node-apn 8.1).
- sendViaApns automatically forwards `payload.eventType` as the
category so every event-type that lands here surfaces as a
matching iOS category identifier without further plumbing. Future
event-types that the iOS app doesn't register fall back to a plain
alert — iOS silently ignores unknown categories.
- mutableContent=1 is set by default so an iOS Notification Service
Extension can hook the payload later without a server-side change.
- New unit tests pin aps.category=MEDICATION_REMINDER on the dispatcher
path, aps.category=MOOD_REMINDER for the new event-type, and the
explicit-category override on the lower-level sendApnsPush entry.
* feat(prisma): add moodReminderEnabled + MoodReminderDispatch ledger
Schema foundation for the v0.5.4 daily mood-reminder push:
- `users.mood_reminder_enabled` (Boolean, default false): per-user
opt-in flag. Default off because mood capture is an emotionally
loaded surface — a fresh-install user must explicitly toggle the
preference in Settings before the server starts nudging them.
- `mood_reminder_dispatches` (new table, unique on userId+date):
idempotency ledger that records exactly one nudge per user per
local date. Doubles as the race-guard for the every-15-min cron
re-ticking inside the same 22:00 window — the unique constraint
makes parallel workers safe even though the worker boots at
localConcurrency=1.
Migration is additive: existing users stay on the opt-out path
byte-identical to v1.4.38 behaviour. Uses IF NOT EXISTS guards
matching the idempotent pattern from 0061 / 0068.
* feat(notifications): add MOOD_REMINDER event type with daily 22:00 cron
Second half of the v0.5.4 ios-coord patch. Now that the schema can
remember "user opted in" and "we already nudged today", wire the
worker side:
- types.ts: extend EVENT_TYPES with MOOD_REMINDER. Default off on
the per-event policy as a defence-in-depth gate behind the
per-user moodReminderEnabled flag.
- mood-reminder.ts: pure dispatcher module with three pinned-down
entry points so the unit tests don't have to spin up pg-boss:
* `evaluateMoodReminderWindow` — boundary contract for
"21:59 → skip, 22:00 → fire, 23:00 → skip" across any IANA tz.
* `buildMoodReminderPayload` — locale-aware title/body
("Stimmung erfassen" / "Wie geht es dir heute?" for DE,
"Log your mood" / "How are you feeling today?" for EN; falls
back to the app default for null/unknown locales).
* `runMoodReminderTick` — orchestrator that pulls opted-in
users, applies the window + already-logged + already-
dispatched filters, reserves the dedup row, and only then
calls the dispatcher. The P2002 unique-constraint violation
is caught and reclassified as "already dispatched" so a
lost race doesn't pollute the wide-event error stream.
- reminder-worker.ts: registers the `mood-reminder-check` queue
with a 15-min cron, mirroring the medication-reminder cadence so
any user's 22:00 boundary is caught within 15 minutes of wall-
clock without one-cron-per-zone. localConcurrency=1 keeps a
worker from racing itself. Also enriches the existing med-
reminder dispatch with a `scheduledAt` ISO string so the iOS
notification handler can pin "snooze 15 min" against the actual
schedule slot rather than the wall-clock moment APNs delivered.
- messages/{de,en,es,fr,it,pl}.json: new `moodReminders.dailyTitle`
+ `moodReminders.dailyBody` keys for all six locales.
- mood-reminder.test.ts: 13 unit tests covering opt-in gating,
window boundaries (21:59 / 22:00 / 23:00 + non-Berlin DST),
logged-today skip, ledger-already-exists skip, lost-race P2002
handling, and the multi-user multi-locale fan-out.
Robustness sweep, perf hotspots, full localization, coach gate
cascade closure.
- /api/dashboard/summary cold-mount 4.6 s → ~500 ms via DAY-bucket
sparkline + 60 s response cache.
- Coach API gate inventory test surfaced and closed two orphan
routes (chat[id] GET+DELETE, chat-message feedback POST).
- Cross-tz fast-path runtime guard falls back to live for users
more than ±3 h from UTC; proper bucket minting stays v1.5.
- es / fr / it / pl coverage 27 % → 63 % (~3 400 new strings) plus
three medication-cluster placeholder restorations.
- IANA timezone validation closes the User.timezone self-DoS
regression the new dashboard SQL would have surfaced.
- Unit suite 4520 → 4524.
See CHANGELOG.md for the full set.
The v1.4.37.1 fire-and-forget hotfix unblocked the event loop
but live HAR still showed /api/analytics?slice=summaries at
~3.1 s per cache miss on a power-user account. Root cause: the
v1.4.35 implementation read EVERY DAY rollup bucket for the
user without a bucketStart window and composed count/min/max/
mean in JavaScript. On the live account that materialised as
a 306k-row transfer plus a JS loop.
Replace the unwindowed prisma.measurementRollup.findMany with
a per-type $queryRaw GROUP BY. Returns 8 rows
(one per type with count/min/max/mean already shaped server-
side) instead of 306k bucket rows. The downstream JS aggregation
is bypassed because the per-type aggregate arrives pre-composed.
Cache-miss budget on Marc-sized accounts: ~3.1 s expected to
drop into the < 100 ms range.
The narrow aggregate query continues to carry the windowed
avg7/avg30 and regression slope/r2 columns that the GROUP BY
cannot reconstruct. Response shape is unchanged.
Hotfix on top of v1.4.37. The post-deploy verify caught
/api/analytics cold-mount hits of ~62 s and parallel /api/health
checks of ~58 s on the live instance, far above the v1.4.37
1.5-3 s perf claim. Root cause: the synchronous
ensureUserRollupsFresh(userId) call at the top of each heavy
read path folds the trailing 90-day DAY window when the rollup
watermark trails the newest measurement. On a power-user account
whose iOS step samples push the watermark every few minutes,
each cache miss synchronously paid the full refresh cost on a
Node worker that also serves the rest of the request fan-out
(including /api/health, /api/version, and concurrent iOS calls),
and the event loop was blocked for the duration of the
recompute.
Convert the three call sites (analytics route, comprehensive
aggregator, summaries slice) to fire-and-forget. The downstream
probeRollupCoverage gate falls back to live SQL on partial
coverage so correctness is preserved; the only user-observable
change is one cache window of staleness on the very first
request after a new measurement landed.
Carries the v1.4.36 perf carry-over for the full /api/analytics
route (correlations + healthScore + bp_in_target on the rollup-
coverage probe, 111s cold worst-case down to ~1.5-3s on power-user
accounts), consolidates Apple Health step samples into a one-row-
per-day list view with chevron drill-down and a nightly drain
(~50x table compression once it converges), restores
IntakeHistoryListV2 to the pre-v1.4.28 contract, converges every
Coach affordance behind a single feature gate with an invariant
test pin, lifts the medication-detail Mounjaro card to byte-
equivalent parity with Ramipril, brings the Insights overview
hero row to height parity, hardens getClientIp behind a new
TRUST_CF_CONNECTING_IP env flag, ships the Arztbericht hero card
on Settings -> Export, adds the Medikamenteneinnahme quick-add
overlay to the dashboard, drops the timezone override toggle for
a silent browser-zone auto-seed, and fixes the v1.4.36 red CI
workflows. Documentation refreshed across all three repos with
five new architecture diagrams in the Dracula palette. Unit suite
4354 -> 4486 passing; typecheck + lint clean.
Heavy-aggregate skip on the rollup-fresh + fully-covered path,
daily chart fetches opt into rollup buckets via source=rollup,
Insights extractFeatures swap drops the prompt payload from
double-digit MB to the low hundreds of kB on power-user accounts,
page-level isLoading gate replaced with per-section early-skeleton
paint, trends-row chart slot pinned for CLS, IntakeHistoryListV2
restored on the medication detail page, About card folded into the
Admin Console with an auto-checking update badge, cumulative metric
tiles read the day's running sum, IP-whois fallback surfaces city
+ country when the ASN lookup misses. Migration 0068 adds an
additive users.insights_exclude_metrics column. Unit suite 4354
passing; typecheck + lint clean.
Removes the operator step that was implicit in the v1.4.35 foundation
release. Self-hosters no longer need to run scripts/backfill-rollups
after upgrade; the worker boot discovers any account with measurements
but no rollup coverage and enqueues a one-shot full fold via pg-boss.
New rollup-full-backfill queue + handler folds one user per job at
serial concurrency. enqueueBootTimeRollupBackfill runs once at
startReminderWorker boot, idempotent across reboots via the
discovery query (matches only zero-rollup accounts) and a
singleton-key per user so fast restarts coalesce.
Best-effort: discovery errors surface through the helper's return
value and are logged via workerLog; the worker boot never fails
because the backfill missed.
Tests: 4285 unit + 230 integration. tsc + lint clean.
Foundation release for the persistent measurement_rollups cache tier.
Adds the additive table + enum, populates it from every measurement
write hook, lands the Apple-Health-import end-of-job backfill, and
ships the parity-checked read-swap on two reader surfaces.
The comprehensive insights aggregator and the slim analytics
summaries slice now source count / min / max / mean per type from the
DAY rollup buckets via aggregateBuckets, plus the comprehensive
dailyByType correlation feed straight from the bucket means. Slope /
R² / standard deviation / anomaly counts stay on live SQL — they do
not compose across DAY buckets. A defensive parity check against the
live aggregate's COUNT(*) falls back to live SQL when divergence is
detected.
Includes the new scripts/backfill-rollups.ts CLI for the one-shot
fold of historical measurements into all four granularities.
Tests: 4280 unit + 228 integration. tsc + lint clean. Migration
0067_v1434_measurement_rollups is idempotent under re-run.
Consolidated hotfix landing the findings from a six-axis audit
covering mobile security, UX + accessibility, code quality + docs,
README + discoverability, web performance, and mobile-deep responsive
behaviour. None of the changes break the public API or the existing
data shape; every fix is additive and SAFE under the v1.4.34.x web
freeze. Full audit reports live in `.planning/round-audit-*.md`.
Security
- Password rotation now wipes API tokens + refresh tokens in the
same `$transaction` as session revocation. A stolen iOS Bearer no
longer survives a user-initiated password change or an admin reset.
- `/.well-known/` proxy bypass tightened to an explicit-allowlist
`Set`. Stale `/api/auth/codex/callback` PUBLIC_PATHS entry dropped.
- HSTS gains `preload`. Withings host removed from the global CSP
and gated to the integration's own routes.
- `getClientIp` warns once per process on `TRUST_PROXY_HOPS`/XFF
mismatch + exposes a tagged helper for tighter universal buckets.
- Service worker rejects off-origin notification-click URLs. Offline
fallback body switched to a locale-neutral icon. Manifest gains
`scope`, `id`, `display_override`.
UX + accessibility
- i18n leaks closed on Dialog/Sheet close-X, ChartSkeleton's loading
announcement, and the notifications-section breadcrumb.
- Focus rings on bottom-nav, the More overflow button, and the
mobile top-bar user menu.
- Tap-target floor: Input / NativeSelect / Select / DateTimeInput
switch to `h-11 sm:h-10`.
- Admin chip-strip auto-scroll-into-view (matches v1.4.33 settings).
- List-page subtitles visible on mobile.
- Insights empty-state CTA vocabulary normalised so the link opens
the right measurement form (was silently failing for
BODY_TEMPERATURE / BLOOD_GLUCOSE / PULSE).
Mobile-deep
- `viewport.themeColor` reacts to light/dark mode.
- TopBar + Coach FAB reserve safe-area-inset-top / -bottom for
notched iPhones.
- Sonner toaster switches to `theme="system"`.
Code quality
- `SESSION_SECRET` dropped — the env var was documented as required
but never read by code. Three secrets, not four. Self-hosters
upgrading can delete the existing line.
- `toJson<T>()` helper consolidates seven Prisma JSON-cast sites.
- `db-compat.ts` gains a header docstring naming it the
schema-bootstrap path.
- Passkey response types adopted from `@simplewebauthn/server`;
`any` + eslint-disable pragmas drop out.
- Zero-TODO CI gate added at `.github/workflows/no-todo-markers.yml`.
Documentation + discoverability
- README rework: How-it-compares matrix, Apple Health + AI Coach
promoted to standalone Key-Features bullets, status block, latest
release + GHCR badges, OpenAPI pointer, three-secret Quick Start.
- openGraph + twitter cards now carry `images` and a
`summary_large_image` Twitter card.
- GitHub repo metadata: description tightened from 290 to 156
chars + keyword-front, `homepageUrl` flipped from a personal
tenant to the demo site, six generic / PII-adjacent topics
swapped for six high-intent ones.
- Five new user-facing docs: `docs/self-hosting/getting-started.md`,
`docs/self-hosting/reverse-proxy.md`,
`docs/integrations/apple-health.md`,
`docs/integrations/withings.md`,
`docs/integrations/ai-providers.md`.
- `docs/README.md` clarifies the docs tree as internal playbooks
vs the user-facing `docs.healthlog.dev` site.
Marc-André Bombeck