A truncated or malformed { "summary": "..." } envelope fell through the
status-card parser and rendered raw — braces, the summary property name,
quotes, and literal \n — in the Assessment card (resting pulse and the
other per-metric cards). parseSummaryFromContent returned the raw content
on any JSON parse failure, and a truncated envelope with no closing brace
also fell into that path, so partial structured output reached the reader.
Split the read into extractAssessmentSummary, which classifies a completion
as a parsed summary, genuine prose, or an unparseable structured envelope.
A payload that opened like a JSON object but failed to parse (or parsed
without a usable summary) resolves to the controlled fallback rather than
raw text; plain prose that merely contains a brace is still preserved.
finalizeStatusSummary withholds an unparseable envelope so every status
card serves its deterministic stub, and the two direct callers
(medication compliance, derived-score assessment) get the same guard.
Refs #608
Merge the eight reviewed audit fixes that were cut a week ago and never
landed. Main has moved on since, so each fix was re-checked against the
current tree and each of the five conflicts resolved by understanding
both sides.
Kept, still open on main:
- Proxy early exits now carry the transport security header subset via a
shared helper. Reconciled with main's /api/health worker-only carve-out.
- Batch measurement ingest bounds the entry instant so a future-dated row
can no longer win "latest reading" or poison the rollup buckets.
- Web Push dials through the pinned safeFetch dispatcher with
requirePublicHost, closing the last user-controlled-host egress.
- Mood bulk upsert and its tag/factor links run in one transaction, so a
rejected rating rolls the mood row back instead of leaving a tagless row.
- Full backup carries NutrientIntakeDay with its unit and source. Folded
into main's portable-export / disaster-recovery split.
- Live daily-series fallback restores the min/max spread band the tiered
read already emitted.
- Coach stream aborts its in-flight reader on unmount and resets send state
on a conversation switch.
- Manual water writes share the day-bound helper, nutrient add clamps after
the increment, and the thiamin cap drops to a real supplement ceiling.
- Compliance point window widens symmetrically instead of wrapping overnight.
- Oura, Polar and Fitbit token helpers classify an empty or non-JSON 2xx
body as a handled transient error.
- WebAuthn relying-party fallback to localhost is gated to development.
- Chart-token strip list is derived from the allowlist and pinned by a test.
- Regression tests pin the snapshot-cache tenant key, the GlitchTip
path-secret redaction and the Withings credentials delete.
Superseded by main, tests re-pointed at the current shape:
- The intake CSV import was rewritten on main to run through a background
job with a single-active-job admission gate. The future-date bound is
re-applied at request validation and the per-user limiter is kept for
parity with the bulk twin; the old synchronous-write test is replaced.
The harness landed in v1.30.35, after the slices that built this
release were already cut — so the reaction line's prompt module was
never registered, and its own docblock's warning against the
two-locale collapse was untested. Registering it surfaced a real gap:
the meaning-first instruction lived only in the system prompt's own
words, never in the user prompt in the family's shared phrasing, which
is where the rest of the surfaces carry it and where the tone contract
checks for it. The German body carried no such instruction at all.
Threads locale through the user-prompt builder so the fixed
instruction line matches the reader's own reply language, and wires
the per-user per-day opener-archetype rotation the milestone spec
called for and the first implementation had not connected.
Four suites, each mutation-checked against a reverted protection:
- route-workout-scope: the workout row is read exactly once across a ten-turn
conversation that keeps sending the id, the where clause always names
userId, another user's id resolves to nothing and the chat proceeds, and
the section sits inside the health-data fence.
- workout-evidence: the numbers-only closure. An instruction-shaped
sport_type — reachable through the restore path, which types the column as
free text — folds to "other" and never reaches the block; a free-text leaf
at any depth throws.
- page-launch-params: an unknown ?scope= is ignored silently, and ?workout=
yields to both ?c= and the fenced ?doc= path.
- composite-score-anatomy-coach: every score sheet carries an opener, a
scope narrowed to the composite's inputs, and auto-send.
resolveLaunchState's shape assertion gains the workoutId field.
One test per gate, each asserting the refusal reason AND that the provider
entry was never reached — a gate that refuses but still resolves a provider
would pass a reason-only assertion while costing what the gate exists to
prevent. The cap and the hash are asserted independently of each other, since
the singleton queue key that normally collapses a burst is best-effort.
The backfill fixture walks the real chain: 750 historical sessions plus 40
re-synced ones plus one fresh, through the live classifier and the live spine
worker, yielding exactly one generation.
The peak detector changed under test: on a flat ride the ninetieth percentile
sits a beat above the mean, so a bpm of sampling noise read as two efforts the
rider did not do. Peaks now need an absolute prominence over the session mean
as well.
Status cards and derived-score sheets read one way: they state a verdict and
end there. The metric pages already carried the hand-off; the score sheets
passed neither an opener nor a scope, so they painted no action at all.
- Score sheets (sleep / readiness / recovery / stress / strain) gain the same
auto-sent opener the metric pages use, each scoped to the inputs that drive
the composite rather than to a block it has no equivalent of.
- Workout detail's coach launch carries the workout itself. The chat request
gains an optional workoutId; on the FIRST turn the route narrows by
{ id, userId } and pins ONE bounded evidence section onto the snapshot.
Later turns reuse the pinned snapshot, so per-turn work stays constant in
conversation length.
- /coach?workout=<id> covers the full-page and maximize paths.
The workout block deliberately does not take the fenced transport the stored-
document path takes: a workout row is the user's own sensor record, and the
projection admits no free text. That is enforced, not assumed — sport_type is
a TEXT column, so it is re-asserted against the closed sport union, metadata
is excluded apart from its numeric zone durations read through a Zod slice,
and a recursive guard throws on any non-numeric leaf.
Own-history sport averages move to src/lib/workouts/sport-context.ts so the
narrative and the detail page cannot quote different figures.
The evidence block is a numbers-only server projection: zone seconds, the
workout-window heart-rate shape, route climb and a same-sport 90-day own-history
median. Free-text metadata never enters it, and the sport string is narrowed
against the closed enum first, so the payload carries no text the user did not
generate through a measured field.
The prompt composes the shared assessment base and adds the session contract:
device-attributed, own-history-comparative, effort acknowledged where the
numbers show it, and no training prescriptions of any kind. Registered in the
tone harness beside the assessment cards.
The deterministic fallbacks nine metric surfaces render when no provider is
reachable carried German and English bodies only. `getLocalizedText` returned
`locale === "de" ? de : en`, so French, Spanish, Italian and Polish readers got
English — on the no-key path, an outage, a spent budget or a timeout, which is
precisely when this text is the whole assessment.
Every string now has a body per locale: the verdict leads and value sentences
of the signal-grounded composer, the per-metric labels and pointers, and the
eight no-signal floors. Selection reads the reader's locale; English stays as
the structural floor for a locale that is ever genuinely missing a body.
The bodies are written per language, not translated word-for-word. fr/es/it
replace the comparative adjective ("higher than your usual average") with a
preposition — au-dessus de / por encima de / sopra — because a comparative
would have to agree in gender with the metric noun, and the labels do not share
one. Polish restructures further: the sentence takes `wynosi`, which is
invariant for the subject's gender, and the comparison uses powyżej / poniżej
with the genitive; the labels span all three genders, so nothing that inflects
can sit in a shared template.
The tone harness grades the four locales by the same rules rather than passing
them by default: false cheer, clinical-verdict and imperative-opener vocabulary
extended to all six, the no-assessment sentinel keyed per locale instead of a
de/en binary, and a byte-identity guard that fails if any locale silently falls
back to the English text again. The value-led-opener rule was rewritten as
well — it assumed the copula sits against the figure, which under-matched its
own second locale (German "liegt aktuell bei 24") and would have missed all
four new ones.
The proxy's seven early exits — worker-only 503, two legacy 301s, the demo-mode
403, and the login / onboarding / MFA redirects — returned before the security
header block, so they carried no HSTS, no framing refusal and no nosniff. Factor
the transport subset into a helper and apply it on every exit. The per-route CSP
stays where it is; it is nonce-bound and only meaningful for a response with a
body.
The WebAuthn relying-party candidate list always carried http://localhost:3000,
including in production. The signed rpIdHash still pinned the real domain, so
this was never a reachable bypass, but a configured production origin has no
reason to carry it. Gate the fallback on development, keeping it when nothing is
configured so getRpId() still resolves.
ORPHAN_ENUMS was a hand-maintained literal that drifted every time
ALLOWED_CHART_TOKENS grew — eighteen allowlisted metrics had no strip entry, so
a model writing 'Your PHQ9_SCORE is elevated' leaked the raw enum into rendered
prose. Derive it from the allowlist and pin the coupling with a test.
The no-signal floors were the coldest text the product renders. They run when
there is nothing to ground against, and they opened on a bare order — "Measure
blood pressure at rest and under comparable conditions", "Evaluate mood trends
over several weeks" — so the one moment the surface has nothing to say was also
the one moment it told the reader what to do.
They now open on what the surface can honestly say: no assessment is available
right now. That is the meaning-first opening for this case, and it is the only
one available, because a floor with no signal may not imply a read. What
follows describes how the measure behaves rather than ordering the reader
around. The signal-grounded lines above them are unchanged.
The prompt version bump re-fingerprints the assessments cached under the old
biomarker prompt; the per-card input hash folds the constant in, so the
regeneration is one-time and paced by the nightly warm.
A deterministic grader over the assembled assessment prompts and the
deterministic fallback text. No provider, no network, no database — it runs in
the normal suite in under half a second, which is the point: the live-judge
eval needs a key and can only advise, so the regressions that actually shipped
need a check that can block.
Three classes are blocking. A surface stops leading with meaning (it drops the
shared opening contract, or its user prompt loses the meaning-first
instruction). A locale collapses (fr/es/it/pl fall back to the German body or
lose the reply-language directive). A prompt reintroduces a value-first
instruction — matched with negation handling, since the shared prompts quote
that failure mode in order to ban it.
The deterministic fallbacks are graded on the same axis, including the
asymmetry that matters: a signal with no baseline supports no confident read,
so that one line honestly opens on the value. Pinning it stops a future
"always lead with meaning" edit from pushing the surface into claiming
something the data does not support.
A coverage guard fails when an assessment prompt module has no harness entry,
so the next surface cannot stay uncovered the way the biomarker card did.
The per-biomarker assessment was the last surface in the family building its
prompt inline: a five-line scaffold that skipped the base system prompt, took
no opener hint, and instructed the model to "state the current value" — so the
one card where a bare number reads most like a verdict was value-first by
instruction while every sibling had moved to meaning-first.
It now composes exactly as the seven bespoke cards do. The new prompt module
carries a METRIC section that states the one thing this card needs and the
others do not: its snapshot has no signal block, so the read comes from the
pre-computed range placement plus the direction across draws.
Also point both archetype-driven cards at a registered budget. The generic
metric card and the biomarker card each carried temperature/maxTokens inline
at 0.45/1000, so the registry that documents every surface's ceiling was
silently wrong about the two widest assessment surfaces.
The single-marker no-key fallback replaces the multi-metric overview tip the
card used to render — text about watching several metrics at once, on a card
showing one lab value.
Three surfaces where the words no longer matched the code, plus the missing
web control for a preference the server already honours.
Delete confirmations. Four dialogs promised permanence on surfaces that
render an Undo: the measurements and mood bulk bodies, the lab-result body,
and the illness-episode body (which said "cannot be undone" sixty lines
below the code that fires the Undo toast). All four now use the same
sentence the single-row measurement and mood dialogs already used. A sweep
of every other confirmation in the tree found the rest accurate — the
biomarker cascade, the mood-tag purge, and the account purge keep their
warning because those deletes really are unrecoverable.
Sleep debt. v1.19.0 replaced the summed shortfall with a running balance
that credits catch-up sleep; the labels still described the old sum, so the
word contradicted the number for anyone who slept in after a short night.
Corrected in all six locales, plus the places the old semantics had
propagated: the Coach system prompt, the OpenAPI field descriptions the
native client reads, and four module docblocks.
Glucose reference band. `hasDiabetes` selects the tighter reference band and
grounds the Coach, but no web surface set it — a web-only account was pinned
to the non-diabetic scale with no way out. The control goes on Targets,
above the per-metric editor, because it answers the same question that page
answers and the two are ordered: it picks the default band, an override
below beats it. The copy says what it is — a reference band, not a
diagnosis, never inferred from a reading, no clinical action taken on it.
The passkey list was reported as always reading "never used". It does not
reproduce: the column is selected by the list route and stamped on every
verified assertion, both since v1.23.0. Pinned with tests on both halves so
the contract cannot quietly regress.
The Today hero stayed on its all-clear line while a dose was actually due:
the dose-window item only fired once the slot was already overdue, so a
dose the user could take right now produced no rail item at all. Emit the
item from the moment the dose becomes takeable — the on-time window's lead
plus the early grace, the same instant logging it starts crediting the
slot — and give the two faces distinct wording and status: a due dose reads
calm and informational, an overdue one keeps the warning. A settled day
still emits nothing, and a cached block whose anchor has since passed keeps
rendering neither face. Push reminders are untouched; they still fire only
on overdue.
The briefing grounding gate discarded the whole block over one number it
could not verify, and it could not verify plenty of legitimate ones: the
allow-set was a hand-maintained list of feature blocks that had already
drifted from the prompt three times, and still omitted medication
compliance, the context counters, the correlation coefficients, every
coverage object and the comparison snapshot — all of which the prompt hands
the model verbatim. Derive the allow-set by walking the prompt payload
instead, so grounded means exactly what it says: a figure the server gave
the model. A fabricated number still matches nothing.
Fix the disposal to match. A content-failed retry left a hole while a
transport-failed one fell back to the previous cached briefing, though in
both cases the rejected text is discarded and the previous briefing passed
this same gate when it was written. Fall back on both paths, and only
report the briefing as withheld when there is genuinely nothing to show.
The on-demand route had no fallback at all; it now mirrors the background
generator that writes the same cached payload.
generateInsight() parsed against the strict aiInsightResponseSchema,
cross-checked that every recommendation's metricSource appeared in
citations[], and threw InsightSchemaError on a second failure. Its only
caller was runWithFallback(), which had no production caller — every
live surface runs runRawCompletionWithFallback. So none of it executed.
Wiring it into the live path was the alternative and it is not viable:
the live insightResultSchema has no citations[] field at all and treats
metricSource as optional, so the cross-check would have rejected
essentially every generation rather than tightening anything.
Removed: runWithFallback and its param/result types, generateInsight
and its private helpers, findUncitedRecommendations,
findRecommendationsMissingRationale, the never-thrown InsightSchemaError
and the isHardProviderFailure branch that excluded it, and
computeConfidence — a deterministic override that only ran inside the
dead wrapper.
Kept: buildRetryCorrectionMessage, which comprehensive-generate sends on
its own retry leg; the file it lived in is renamed to match what it now
holds. aiInsightResponseSchema stays as the shape that message
describes. Note the retry prompt asks for a stricter payload than the
live validator requires — that mismatch predates this and is left alone.
computeCitationCoverage is untouched and still live: it is a metric, not
a gate, and v1.30.24 wired it into the comprehensive post-parse step.
Chain-runner coverage the deleted tests carried (last-working cache,
empty-input guard) is ported onto runRawCompletionWithFallback, which
shares the machinery.
Pin the four guarantees at the new chokepoint and at both call sites: spend
lands on the ledger, over-cap is refused before any egress, a BYOK chain is
measured against the user-plan ceiling rather than the operator's, and a
failed call refunds in full. Also pin the ordering the consent gate depends
on — a consent-blocked user is never charged.
The fixture described a clean resting series but supplied rows for only
twenty of the thirty days it fed PULSE for. Now that the resolver merges
per day, the ten resting-free days are estimated from their own readings
— and on a day whose only readings are a 150 bpm workout the honest
estimate falls outside the band. Match the spans so the case tests what
its name claims; gap-day behaviour is pinned in resting-pulse.test.ts.
The status tier funnels every generation through runStatusCompletion, which
reserves against the day's ledger before egress and reconciles after. The
comprehensive briefing did not: the shared generator and the on-demand
POST /api/insights/generate route both called the provider chain directly, so
none of that spend reached coach_usage and no daily ceiling applied to it.
Route every briefing provider call — first pass, JSON-shape retry,
grounding-correction retry, and the daily paragraph re-roll — through one
chokepoint that reuses reserveBudget / reconcileSpend and resolveDailyCap. The
reservation lands after the consent gate, so a consent-blocked user is never
charged, and a failed or timed-out call is refunded in full.
Over-cap is a typed skipped outcome for the off-request generator (no upstream
fault to retry against, so no failure marker and no retry enqueue) and a 429
for the on-demand route.
The red-team corpus probed only user messages and model replies. Neither is
the channel that reaches the prompt unguarded: a lab analyte, panel or unit
name, or a medication label, none of which the account holder authors when
the row was committed from an uploaded document.
The free-text guard's file list was hardcoded and had rotted — three of its
four entries were no longer reachable from the snapshot, and it covered one
of the dozen files that actually emit free text, which is why the lab fields
went unnoticed. Derive it from the import graph instead; it now covers 75
files, with a floor assertion so a resolver break cannot silently empty it.
Deriving the list immediately surfaced a second leak: features.ts shipped a
raw medication name and dose into the snapshot medications block.
Also fix sanitizeForPrompt itself. It deleted control characters, which
welded neighbouring tokens together and destroyed the word boundary its own
injection patterns anchor on, so a payload preceded by a newline survived
because of the newline. Replace with a space and collapse before matching.
The Coach SNAPSHOT was spliced into the prompt bare. Its JSON leaves are not
all server-computed: lab analyte, panel and unit names are transcribed by a
model out of an uploaded PDF and stored verbatim as the biomarker name, so a
hostile document chooses those strings. The about-me block has been fenced
since v1.16.1; extend the same pattern rather than adding per-field escapes.
- Move the fence markers and the wrapping primitive into a shared data-fence
module. Content is scrubbed against every known marker, not just its own
pair, so text in one block cannot forge another block's boundary.
- Fence the SNAPSHOT with an explicit data/instruction frame.
- Fence the OCR text handed to the document extractor; the previous
'OCR TEXT:' label was a label, not a boundary.
- Sanitise analyte, panel and unit at the labs source, which feeds both the
snapshot prompt and the labs_read tool result, and the medication label in
the adherence storyline.
- Capture a ---REMEMBER--- sentinel as proposed rather than active. It was
the only model-driven write with no confirm step, and the reminder recall
block fed it back into every later turn. Proposed reminders are already
excluded from that block and already carry a confirm control.
The status-provider suite drives the real budget module against an
in-memory stand-in for the day's usage row, so it exercises the atomic
reserve/reconcile path rather than asserting a mock was called: spend
lands on the ledger, an exhausted day is refused before any egress, a
failed generation is refunded, and a chain on the user's own key is
measured against the user-plan ceiling rather than the operator's.
The snapshot suite walks a 60-turn conversation and pins the full
snapshot to exactly two turns — the first and the elision crossing —
including the case where a lost reply shifts the turn parity.
The provider chokepoint every status and reference generator funnels
through ran with no budget accounting at all. Nine generator families
egressed behind it — the specialised status cards, the generic metric
cards, biomarker cards, the batched assessment, the derived scores, the
period narrative, and the off-request Coach memory workers (rolling
summary, fact extraction, plan proposals) — and none of that spend landed
on the daily ledger or met any ceiling. On an operator-key account it was
unmetered operator spend.
Reserve and reconcile around the provider call, using the same atomic
upsert the Coach and document paths use. The ceiling follows the cost
owner: a chain whose primary is the operator's own credential draws the
operator ceiling, everything else draws the user-plan ceiling, so a
self-hoster on their own key is never rationed by the operator's bill.
The status enqueue keyed its de-dupe singleton on (user, metric, locale).
The locale on a status read is client-chosen, so each of the six
supported locales opened its own slot and one legitimate generation pass
could be multiplied into six. Key on (user, metric) and carry the locale
in the payload instead.
The Coach re-shipped the full snapshot on every turn past the history
cap, not just the turn that crosses it, so a long conversation paid the
per-turn snapshot cost the snapshot-once model exists to remove. Re-ground
on the crossing only, tested as a narrow window rather than an equality
because a turn advances the count by one or two depending on whether the
reply persisted.
Give /api/ai/test a daily ceiling and a ledger entry. Its burst bucket
bounded rate but nothing cumulative, and with an empty body it resolves
the same chain generation uses — which can be the operator's credential.
Correct the memory-refresh worker comment, which asserted a budget check
inside the chokepoint that did not exist.
The coverage check hung off generateInsight, whose only caller runWithFallback
has no production caller left -- every live surface routes through
runRawCompletionWithFallback instead. The admin quality dashboard was reading a
metric nothing emitted. Widen the input to the structural shape both payloads
share and call it from the comprehensive post-parse step, so the annotation
reflects real generations.
Writing the locale tests found three gaps the English-only banks hid: Polish
verb suffixes fell outside the ASCII word class, a trailing word boundary after
an accented character never closed, and German subordinate clauses put the verb
last so the natural phrasing of a dose instruction never matched. The payload
collector also skipped legacy bare-string recommendations entirely.
The briefing gate grades numbers only, and only the paragraph, signals and
key-findings blocks -- so a digit-free risk assertion with a dose suggestion
attached passed it untouched, and recommendations and summary were not graded
at all. Screen every readable prose field against the insights contracts.
A safety violation cannot be surgically stripped the way an ungrounded number
can: the offending sentence may sit in any field, and deleting one would leave
the rest referring to it. So the background generator fails the whole run and
persists nothing, leaving the previous cached payload in place, while the
user-initiated route returns an explicit error for the same reason.
The causal bank lived in the narrative validator as an English and German list,
so the no-causation rule was enforced in code on one surface and for two of the
six locales it ships in. It now comes from the shared screen, which carries all
six, and the narrative also gets the dose and risk-score contracts it never
checked -- persisted prose with a dose imperative in it is as unsafe as a chat
reply with one.
Red-team cases carry their language. The German outbound case was probed as
English, which would have reported a pass the guard never earned.
The outbound safety screen ran on the Coach only. A dose-change imperative or
a fabricated clinical risk score written onto a per-metric, biomarker or
derived-score status card persisted as that day's cached assessment, while the
identical sentence in the Coach was caught and replaced -- the only transform
between provider and persistence was whitespace normalisation.
Move the pattern banks into a shared screen that takes the reader's locale.
The old signature took no locale, so the dose-refusal and risk-score contracts
had no enforcement for fr/es/it/pl even though the ground rules ship in all six
contract files. Banks now cover all six locales and each screen runs the
reader's bank plus English, since a provider often answers in English
regardless of the language directive.
Status cards withhold on a trip: serve the deterministic stub and persist no
model text. Nobody waits on a background-generated card, and persisting a
refusal would pin it to the card for the day.
getBaseSystemPrompt appended the directive, then each metric module
appended its own section after it, so the directive was no longer the
last instruction the model reads - the one property that makes it
authoritative. The base now exposes its body, and every composing module
wraps the finished prompt so the directive ends it.
Also drops a duplicate import that survived the merge of the two prompt
branches.
Three layers. The German and English narrative prompts are frozen by content
hash, normalised for the prompt-version string that a prompt change is
required to move — any other drift fails. The four riding locales must name
their language, carry the verbatim reply-language directive from their own
safety-contract file as the last instruction, and contain no German-body
sentinel. A generator-level test drives a French run end to end and asserts
the row is keyed fr, the cache action is locale-scoped, and the system prompt
reaches the provider in English with the French directive.
Alongside it a source-level ban on the exact shape the bug took
(`=== "en" ? "en" : "de"`, `?? "de"`) across the coach and narrative
surfaces, with a self-check so a typo in the pattern cannot turn the guard
into a silent pass, and coach-memory recall cases for all four locales.
Covers the assertion that would have caught the narrowing: normalizeLocale
passes all six locales through and defaults an unknown value to English, the
de/en cache keys stay byte-identical while the four widened locales get their
own, and a fr locale entering a status generator reaches the prompt builder
as fr and writes the fr cache row. The forced-warm case pins the same for the
comprehensive and per-status warms.
Holds every sibling prompt module to one invariant: strip the directive
and the interpolated language name out of a French, Spanish, Italian or
Polish prompt and what remains is the English prompt, byte for byte. A
German instruction body leaking back into a non-German prompt, an
English body drifting away from what the four locales receive, and a
module appending a second directive on top of the one the base prompt
already carries all fail here.
Also drops the now-unused narrow locale alias from the two modules whose
prompt builders take the full locale.
The narrative read route narrowed every non-English locale to German, so a
French, Spanish, Italian or Polish account read the German narrative row and
warmed it from the German prompt body. Carry the resolved locale through
instead: the row is keyed (user, period, locale) across the full UI locale
union, and the four locales now compose the English instruction body with
their language named and the locale's own reply-language directive appended
last. The German and English prompts are unchanged apart from the version
string.
The nightly warm keeps the user's stored locale for the same reason — it has
to write the key the read route reads — and the coach and dashboard narrative
recalls read that same key. Coach memory stays de/en: it is model-facing
context, not user-facing prose, so the four locales compose and key English
rather than German.
NARRATIVE_PROMPT_VERSION 1.13.0 -> 1.14.0.
The per-metric assessment pipeline typed every locale as `de | en` and
narrowed on the way in, so a French, Spanish, Italian or Polish account
never reached a prompt with its own locale — the output-language directive
could not fire and the reader got English, or German where the binary
defaulted that way.
Widen `SupportedLocale` to the six locales the UI ships and replace the
narrowing binaries with one validating `normalizeLocale` that passes a
recognised locale through unchanged and defaults an unknown value to
English. The status-generate payload, status batch, status cache,
comprehensive generation, the pre-generate warm passes and the forced-warm
handlers all carry the full locale now.
Fix the collapse sites that discarded it: the warm enqueue on the
comprehensive route and the pregenerate route, the nightly batch cron, and
the forced-warm job handler — the last two defaulted to German outright.
The deterministic derived-score templates and the previous-context block
ship de/en bodies only, so they route through `instructionLocale` (German
for German readers, English otherwise) while the prompt names the reader's
own language.
The derived-score archetype and the Coach preference overrides branched
on the raw locale, so fr/es/it/pl composed German instruction text. Both
now route through `instructionLocale`. The score archetype rides on
`getBaseSystemPrompt` and therefore appends no directive of its own —
the base prompt already carries it.
The biomarker assessment is the one prompt in the family that does not
compose the base scaffold, so it appends the reader's directive itself;
the append collapses into the shared path once that prompt moves onto
the base scaffold.
The two prompt builders in each of those modules now take the full
`Locale` rather than the narrowed de/en alias, so the routing is right
the moment the pipeline stops collapsing the four locales on the way in.
Also targets the output-language contract test's regex at the compile
target instead of the dotAll flag, which does not compile.
German and English prompts are unchanged, byte for byte.
The generated-insight narrative layer built its user-facing sentences by
string interpolation and never routed through the i18n bundle, so a
non-English UI rendered English text spliced into localised copy — e.g.
"Higher activity steps tends to go with lower next-day weight" surfacing
mid-German on the Weight insight. Because those strings bypassed t(), the
locale-integrity guard could not catch them.
Route every correlation narrative through the server translator:
- Correlation discovery (daily lagged pairs + labs-vs-outcome) now takes
the reader's locale, resolved from the correlations route the same way
other insight generators resolve it. Sentence templates, the direction
word, and every channel label live under a new insights.correlation.*
namespace, translated in de/en/es/fr/it/pl. The three fixed hypotheses
(BP-compliance, mood-pulse, weight-weekday) and their weekday labels are
localised the same way; the analytics cache key now includes the locale
so a language switch cannot serve another locale's cached text.
- Metric labels resolve to stable localised channel labels; the never-
causal framing ("not a cause" / "an association, never a cause") and the
faint/moderate/high hedging tiers are preserved in every language.
English output is byte-for-byte unchanged. Adds a regression test pinning
German narration and asserting it is not the English string.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018V8V9qapgHi7fhgtMJ8SkM
Marc-André Bombeck