A Health Records result the provider later marks entered in error is
skipped as withdrawn, but a reading already stored from an earlier
version stays until it is deleted in HealthLog. Say so in the changelog,
the import guide and the batch route's contract, and bring the guide's
archive limits and lab notes in line with the current reader.
A urinalysis sent its specific gravity and pH as bare numbers, which
failed as missing_unit and made a normal report a partial success. A
number without a unit is now the unity "1" when its LOINC code names a
specific gravity, a pH or an INR; any other analyte keeps the refusal.
The category check took only FHIR's own "laboratory" and skipped
records a system filed as "LAB" or "Lab Results", although Apple
Health had already classified them as lab results. A code, display or
text containing "lab" now counts; other categories are still skipped.
The import collected every workout-route and Health Records reference
the export's XML named until the parse ended, although the passes read
at most 20,000 routes and 50,000 lab files. Past those ceilings a
reference is now only counted, and the passes report the count as
failed.
The ECG, route and Health Records passes applied the per-member byte,
ratio and count limits to every member of their kind, wanted or not,
so one large Health Records file nobody referenced, or more than 50,000
of them, ended the whole lab pass. A wanted member that broke a limit
also took every other member of the pass with it, and the compressed
bytes were read into memory before any size check.
The limits now apply to the members a pass will read; a breach fails
that member alone, and the compressed size is held to the member limit
before it is allocated. Running totals still end the pass. The ceiling
on directory entries is one shared value of 200,000 for all three
passes, so an export with more than 20,000 files keeps its ECGs.
The import recognised a deleted result by the identity on its own
tombstone, so anything that removed the row brought the result back
with the next Apple Health export or resend: deleting the whole marker,
which hard-deletes its readings, the nightly tombstone purge, and a
portable backup, which leaves deleted readings out.
The identities now move into an import ledger (LabResultImportKey,
migration 0501) in the same transaction as the marker delete and the
purge, and the import checks it beside the stored rows. The backup
carries the ledger, a portable file also the identities of deleted
readings still waiting as tombstones; the restore writes it back
without the identities a restored reading holds. The data wipe clears
it, account deletion cascades.
Both the Health Records import and the CSV import checked the unit
before writing and refused a numeric reading against a marker whose
unit was still empty, although the write itself lets the first number
choose it. An empty marker unit now fits any unit in unitFitsMarker;
the write's conversion stays the authority.
The per-type reads descend measurements_live_covering_idx once per type and
stop at their cap, but the planner chose that descent only when its row
estimates favoured it. The type is a lateral parameter, so one generic
estimate covered every type; an account whose rows arrived after the last
ANALYZE was expected to hold about one row; and statistics taken while the
table held one type made the type condition look free. In those states the
planner walked the type-less (user_id, measured_at) index through every row
of the account once per type, or read every row of each type into a sort.
On a 9,300-row account that was 33,900 row visits for 6,300 rows returned,
1.66 million for the span probe and 845,000 for the dashboard's latest
reading per type.
One shared fragment now holds the descent: the account and the cap reach
the planner through scalar subqueries, and the type is matched as a closed
range it is also ordered on, so only an index leading (user_id, type,
measured_at) can return the rows without a sort. readNewestRowsPerType,
readLiveMeasurementSpan, the dashboard's readLatestEver and the latest-instant
probe of readLatestCanonical use it, and readLatestReadingsPerType reads
through readNewestRowsPerType.
The plan guard moves to its own file, which builds four statistics states on
purpose (none, predating the account over many types and over one type, and
analysed with other accounts) with autovacuum off, and holds every read to
its bound in each. The previous guard depended on statistics another file
had left behind, so it failed in the full suite and passed alone.
Refs #1398
Integration coverage for the nutrient batch under each credential (the
source row, the 403s, the module gate, re-post, backfill, coexistence
with the phone's row and the bridge bucket), kcal and g ingest with kJ
skipped, the day fold on the overview and in the correlation read, and
composition refusing energy, macros and alcohol. Route-level checks run
against Postgres rather than a stubbed handler. The energy card reports
a failed save through the shared error toast.
Nutrient targets travel in the account-settings slice of the full
backup and are admitted code by code on restore, so an unknown code or
an out-of-range value is dropped and named while the rest come back. A
data wipe clears them. The targets route joins the OpenAPI registry,
and the regenerated contract carries the wider nutrient enums, the new
sources and the target and authority fields. Refs #1402.
The integration docs said clinical records became measurements; they
never did. They now describe what the export import reads from Health
Records, how lab results are matched and deduplicated, the parser
revision 4 re-upload, and ZIP64 support. CHANGELOG entries for #1400
and the ZIP64 fix.
The export import through the real worker (LOINC matching, unit
conversion, panels, skips, a hand-typed duplicate, re-upload and a
deleted result), and the batch route (resend, an export-written result
recognised, token scope, size and shape refusals, idempotent replay).
Both the disaster-recovery and the portable file now carry
lab_results.external_id, and a restore writes it back, so a restored
account still recognises a Health Records result when it arrives again
and a deleted one stays deleted. Older files restore unchanged.
The iOS app reads lab results from HealthKit's Health Records and sends
them here as FHIR resources, up to 200 per batch. The route maps and
writes them with the same mapper and writer as the export import, so a
result already imported from an export is a duplicate, sending a batch
again is safe, and a deleted result stays deleted. Owner-only (cookie or
full-scope login token), its own 60-a-minute bucket, Idempotency-Key
supported, per-record status with closed reasons, counts-only audit.
The OpenAPI contract documents the route and the source values a lab
result can carry, and the import status result names its lab counts.
A bridge such as a food tracker can now post daily nutrient totals with
the measurement token. The credential picks the source row: a scoped
token writes EXTERNAL, the native client keeps writing APPLE_HEALTH, and
the body has no say. Bridge tokens count in their own rate-limit bucket,
and the source is recorded in the audit row.
The write phase moves into writeNutrientDayTotals so the route keeps
only parsing, validation and bookkeeping. The scope guard now freezes
four doors for measurements:write. Refs #1402.
The catalog gains energy (kcal), protein, carbohydrates, fat, sugar and
fiber (g), sodium and potassium (mg) and alcohol (g of ethanol, no
HealthKit type). Each code now names its group and whether a
supplement's composition may list it; references name their authority
(EFSA, or IOM for protein and carbohydrates, NIH ODS for chromium).
Day totals fold through one rule everywhere: the largest of the
day-total sources (Apple Health, Health Connect, an external bridge)
counts, manual water and supplement doses add on top. This also stops
Apple Health and Health Connect totals for the same day from adding up.
The nutrients page gets an energy and macronutrient card with an inline
daily target, and an alcohol card next to caffeine. The correlation scan
gives the macros their own budget of two channels and leaves logged
alcohol out. Daily series and the MCP tool return the target the person
set. Refs #1402.
A nullable users.nutrient_targets_json column (migration 0499) holds a
target or limit per catalog code in the code's canonical unit. GET and
PATCH /api/nutrients/targets read and merge it behind the nutrients
module; PATCH refuses unknown codes and values above the code's
plausibility cap, and audits the codes set, never the amounts.
A finished import now lists the lab results it found next to the ECG
line, and an archive whose only new data is lab results no longer says
nothing was imported. The status endpoint bounds the lab counts the
same way it bounds the ECG counts.
The export parser counted every <ClinicalRecord> as skipped. Lab-result
records are now handed to a pass after the export.xml rows are written:
each referenced clinical-records/ JSON file is read from the archive
one at a time, mapped with the shared FHIR mapper, given its panel's
name and written through the shared writer. Other clinical types stay
skipped. The pass is fail-soft like the ECG and route passes, and the
job result carries bounded lab counts only.
The ECG pass moves out of the worker into ecg-import.ts, and the parser
revision goes to 4 so an archive imported before can be uploaded again
to pick up its lab results.
The auxiliary member reader refused ZIP64 archives, which is what the
Health app writes for a large export, so ECGs and workout routes were
counted as failed exactly for the accounts with the most history. It
now resolves the ZIP64 end record and the per-member extended sizes,
with the same bounds as before.
It also learns the clinical-records/ JSON members, and a stored
(uncompressed) member is no longer returned as zeros: its bytes were
the read buffer the reader wipes after use.
A shared mapper and writer for lab results arriving as FHIR resources
(DSTU2 and R4 Observations): final results only, laboratory category,
quantity / string / coded values, comparator results kept as printed
text, the report's own reference range in the reading's unit, panel
headers naming their members, notes sealed at rest. UCUM units map onto
the spellings the labs feature uses, and a LOINC code hints the catalog
analyte when the account's markers are matched.
Readings carry an identity (lab_results.external_id, a digest of the
issuer and the record identifier) so the same result is recognised
whether it arrives again from an export or from a phone, and a deleted
result stays deleted. A reading the record already holds by value is
skipped as well.
The bulk measurement reads behind the insight features, the Coach snapshot
timelines, the mood cross-metric pairing, the correlation windows and the
weight and blood pressure status cards read newest-first across all their
types under a single row cap. On an account with a wearable heart rate
stream the cap filled within hours, and every weight, blood pressure or
other sparse reading fell out of the read, so blocks reported no data for
metrics with years of it.
All of them now go through one read that descends the live partial index
once per type, each type stopping at its own cap. The feature extraction
and the Coach snapshot cap each type at 6000, so a block built from one
type's read equals the block the full read builds (worst case 72,000 rows
for the twelve feature types). The mood, correlation and status reads keep
their old totals and share them fairly: each type keeps at least its share,
a sparse one all of its window.
The feature context's data span now comes from index probes over every
type in the window rather than from the capped rows, and its measurement
count is the readings the blocks were built from.
Refs #1398
16.4.6 keeps a single header value as a string, so the compression
middleware now sees a route handler's Content-Type and gzips the
response when the client accepts it. Route handlers were never
compressed before. Measured on a production build: JSON and CSV
exports go out gzipped with their Content-Disposition intact and still
flush chunk by chunk, so a streamed download keeps arriving as it is
written. The Coach and document chat streams send
Cache-Control: no-transform, which the middleware honours, so they stay
uncompressed and each frame arrives when it is written. A guard now
holds every text/event-stream surface to that directive.
The notification journey failed twice on the release candidate at the
same step: the e-mail address typed into the card was wiped. The spec
waited for the settings read with waitForResponse, which settles on the
headers, and treated that as the card having copied the server value
into the input. With compression the body follows the headers after a
round through zlib on the thread pool, which a loaded runner can hold
up, and the late copy overwrote the typed value. The spec now also
waits for each card's enable switch, which is disabled until its read
has landed and is enabled in the same render that does the copy.
A marker minted from a text result (OCR, CSV import, inbound document)
stores an empty unit: none has been chosen yet, and its first numeric
reading sets it. The editor read that empty unit as a unit still to be
typed and refused to save, ticking "No unit" sent "1", and the update
route refused any unit change on a marker with readings. Every save of
such a marker failed, a plain rename included, because the form always
sent the unit.
- The form opens an empty unit with "No unit" ticked and sends the unit
only when it changed; the empty unit and the unity count as the same
"no unit" there, so a rename leaves the stored unit as it is.
- PUT /api/biomarkers/{id} locks the unit only on a real change while
numeric readings exist (deleted ones included). Another spelling of
the stored unit is no change, and text results carry no number the
unit could misstate, so a text-only marker may take any unit.
- The empty unit stays distinct from "1": it still adopts the unit of
its first number, which "1" must not. No stored value changes.
Refs #1393
4.0.0 moves the package to ESM and class syntax and drops minimist from
the CLI. The sender and the VAPID route load it with a dynamic import
and call generateVAPIDKeys, setVapidDetails and generateRequestDetails,
which keep their names, arguments and module-level VAPID state; the
request is still dialled through safeFetch, so the library's own
transport and error shapes are not on our path.
Every sender test mocks the library, so a contract test now runs the
real one: a stored VAPID pair is accepted, the request carries the same
headers (TTL of four weeks, Urgency, aes128gcm), the VAPID JWT verifies
against the public key with the endpoint origin as audience, and the
subscriber's keys decrypt the body back to the payload. It passes on
3.6.7 and on 4.0.0.
@aws-sdk/client-s3 3.1169 and lib-storage 3.1167, @napi-rs/canvas 1.1.1,
@simplewebauthn/server 14.1.0, @tanstack/react-query 5.107.1,
dotenv 18.0.10, jose 6.2.17, lucide-react 1.73, mmdb-lib 3.0.5,
nodemailer 10.5.0 and pg-boss 12.38.2.
The pins that hold these move after reading the installed sources:
- pg-boss 12.38.2 keeps the failed-row ledger the 72-hour reader relies
on: deletion still runs on completed_on + deletion_seconds against
job_now(), with the 14-day retention, 7-day deletion, 15-minute expiry
and two retries unchanged. The release only moves keep_until along
with start_after on retry and resume.
- @simplewebauthn/server 14.1.0 adds passkeyProviderMetadata to the
registration result. It is filled only by an initialised FIDO metadata
service, which downloads from the FIDO Alliance; HealthLog initialises
none, so the contract now pins the field as absent. The stored key
bytes are unchanged. The new origin and RP ID checks reject a trailing
slash or non-ASCII form, and both values already come from URL.origin
and URL.hostname.
- dotenv 18.0.10 moves its CLI to a separate entry; the library API is
unchanged. The Prisma CLI stage installs the same version.
postgres:18.6-trixie now resolves to e3d24f6, a multi-arch index with
linux/amd64 and linux/arm64 on the current Debian trixie-slim base. The
PostgreSQL package is unchanged (18.6-1.pgdg13+2).
node:22-trixie-slim now resolves to 68c76cf, a multi-arch index with
linux/amd64 and linux/arm64 that carries Node 22.23.5, a security
release with HTTP, HTTP/2 and TLS fixes. All four Node stages and the
release guard move to the new digest together.
A night on which every account was skipped, for instance because every
provider was paused for background work, wrote all_failed false. The
repeated-failure alert read that as a successful run: it ended the
streak and announced a recovery while the provider was still down, and
nights then alternated between failed and paused so the streak never
reached the alert again.
The run now also records run_attempted. A completed run with
run_attempted 0 is neither a failure nor a recovery: the alert and
readConsecutiveRunFailures step over it.
Refs #1384
The footer reserve under a day-linked chart's loading shell still held the
caption line that was dropped from those charts, so every such shell stood
20 px taller than its chart (40 px on a phone). The dashboard's mood chart
loaded behind the generic shell, which is about 30 px shorter than the
mood card. When the shared chart chunk arrived late the dashboard's chart
row changed height under a reader already scrolled past it; the tablet
shell journey measured <main> right after such a swap and found it no
longer at its end.
The reserve now holds only the day rug, and the mood chart gets a shell
built from the same card primitives and boxes. The tablet journey also waits
for every chart cell to reveal and every loading shell to go before it
scrolls to the end, instead of trusting two equal heights 250 ms apart.
Raising the record-session hold clears the query cache, which remounts the
account query, so the switching tab has a GET /api/auth/me in flight beside
its own POST /api/account/switch. When the server answered that read from
the record being left and the answer was applied after the commit, fetchMe
rewrote the scope mirror back to the old record. A peer tab that had already
settled read that write as a foreign switch and published a hold expecting
the old record, which no /me would ever confirm, so every tab sat behind the
hydration gate until the stored transition aged out.
The transition store now counts the changes a tab applies, and fetchMe
repeats a read whose request straddled one instead of applying it.