Promote the per-turn magnitude bag to a per-conversation typed Grounding
Ledger built from every source the model was actually shown: this turn's
tool payloads, prior-turn tool figures, coach-memory goals, the medication
schedule, the reference-grounding block, and guided blocks. Assistant prose
is never a registration source, so a figure that slipped a rung cannot
launder into an authoritative value on the next turn.
Complete the action ladder: a reconciled number passes, a reference or
education sentence is exempt across all six locales, a positively-typed
benign number (date, time, year, ordinal, list marker) passes while an
untyped unreconciled one stays on the strip path, and an unreconciled
health figure is soft-stripped.
Gate the dose continuation exemption on the user's actual schedule: a
"keep taking your N mg" is trusted only when N is a dose the user is on;
an off-schedule maintenance dose is caught. Extend the risk and percent
classifiers to fr/es/it/pl to match the EN/DE depth.
No schema migration: the ledger is rebuilt per turn from persisted tool
traces and re-rendered prompt blocks.
Replace the untyped magnitude bag in the Coach numeric verifier with a typed
tokenizer and a normal-form reconciler, and narrow the outbound risk and dose
banks so correct figures and correct refusals stop being blocked while every
fabrication shape still blocks.
Grounding verifier (coach-prose-grounding.ts):
- Classify each prose number before grading — ISO/written dates, clock times,
bare years, ordinals, list markers, hyphenated time spans, ranges, percents,
thousands-separated integers, unit-suffixed measurements. Benign types are
never graded or stripped; ranges decompose into per-endpoint magnitudes.
- Grade against typed authoritative entries (value + kind + aggregation).
Sign-insensitive, canonical rounding, ratio to percent in percent space, and
minutes to hours widen the match only for an aggregate, kinded entry; kind
scoping stops a weight value grounding a blood-pressure claim.
- Thousands/decimal separators follow the reply language.
- Boundary-safe strip so a flagged token never clips a larger grounded number.
Outbound screen (outbound-screen.ts):
- Risk bank requires an asserted figure or a categorical engine result; a bare
engine or horizon mention (education/refusal) passes. Spelled-out percents
count as figures. No refusal-context exemption. EN + DE.
- Dose bank keeps a maintenance-continuation restatement, voided by any change
stem. Taper suffix fixed.
Route + fenced-chat thread the reply locale into the verifier.
The named-risk-engine pattern in the outbound safety screen used
score2? (the 2 optional), so it matched bare "score" too — any mention
of one of this app's own computed, documented scores (Sleep Score,
Readiness Score, Health Score, Recovery Score, Stress Score, Strain
Score) tripped the bank meant only to catch a fabricated named
clinical-risk calculator (SCORE2, Framingham, ASCVD, QRISK).
Require the literal "SCORE2" instead. The other risk patterns —
percentage-shaped risk claims and the other three named engines — are
untouched, so a genuinely fabricated risk figure or a named
risk-calculator claim still blocks.
Closes#587.
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 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.
Marc-André Bombeck