fix(insights): QA fixes — paired BP targets, stdDev, pulse pressure, personal tone, schema enrichment

- Fix pctInTarget to use paired-reading logic (both sys AND dia must be in range simultaneously)
- Add sdSys30/sdDia30 (BP standard deviation) and pulsePressure30 (arterial stiffness marker)
- Add mood scale descriptor for AI context
- Extend medication data with compliance90, 30-day streak window, and category field
- Rewrite base system prompt: personal advisor tone, second person, positive-first, "One Thing" pattern
- Add conditional correlation instructions to all domain prompts (only mention when |r| > 0.4)
- Enrich InsightResult schema with insightType, classificationLabel, primaryRecommendation

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Marc-André BombeckandClaude Opus 4.6 committed 2026-04-06 12:48:28 +02:00
1 parent 64c72d241c
commit 7b5fabe46c
9 files changed
+89 -28

No files matched your search

+51 -10
View File
@@ -13,6 +13,7 @@ import {
pearsonCorrelation,
type CorrelationResult,
} from "@/lib/analytics/correlations";
import { getMedicationCategories } from "@/lib/medication-category";
interface DataCoverage {
count: number;
@@ -49,6 +50,9 @@ export interface AggregatedFeatures {
allTimeMaxDia: number | null;
slopeSys30: number | null;
slopeDia30: number | null;
sdSys30: number | null;
sdDia30: number | null;
pulsePressure30: number | null;
pctInTarget: number | null;
coverage: DataCoverage;
};
@@ -70,6 +74,7 @@ export interface AggregatedFeatures {
coverage: DataCoverage;
};
mood?: {
scale: string;
avg7: number | null;
avg30: number | null;
latest: number | null;
@@ -109,8 +114,10 @@ export interface AggregatedFeatures {
medications?: Array<{
name: string;
dose: string;
category: string;
compliance7: number;
compliance30: number;
compliance90: number;
streak: number;
missedLast7: number;
}>;
@@ -134,6 +141,13 @@ export interface RawFeatures extends AggregatedFeatures {
}>;
}
function stdDev(values: number[]): number | null {
if (values.length < 2) return null;
const mean = values.reduce((s, v) => s + v, 0) / values.length;
const variance = values.reduce((s, v) => s + (v - mean) ** 2, 0) / (values.length - 1);
return Math.round(Math.sqrt(variance) * 10) / 10;
}
function toDataPoints(
records: Array<{ value: number; measuredAt: Date }>,
): DataPoint[] {
@@ -309,16 +323,19 @@ export async function extractFeatures(
let pctInTarget: number | null = null;
if (bpTargets) {
const sysInRange = sysData.filter(
(m) => m.value >= bpTargets.sysLow && m.value <= bpTargets.sysHigh,
).length;
const diaInRange = diaData.filter(
(m) => m.value >= bpTargets.diaLow && m.value <= bpTargets.diaHigh,
).length;
const total = sysData.length + diaData.length;
if (total > 0) {
pctInTarget = Math.round(((sysInRange + diaInRange) / total) * 100);
const sysByTime = new Map(sysData.map(m => [m.measuredAt.getTime(), m.value]));
let inTargetCount = 0;
let pairedCount = 0;
for (const dia of diaData) {
const sysVal = sysByTime.get(dia.measuredAt.getTime());
if (sysVal === undefined) continue;
pairedCount++;
if (sysVal >= bpTargets.sysLow && sysVal <= bpTargets.sysHigh &&
dia.value >= bpTargets.diaLow && dia.value <= bpTargets.diaHigh) {
inTargetCount++;
}
}
pctInTarget = pairedCount > 0 ? Math.round((inTargetCount / pairedCount) * 100) : null;
}
features.bloodPressure = {
@@ -334,6 +351,22 @@ export async function extractFeatures(
allTimeMaxDia: diaSummary?.count ? diaSummary.max : null,
slopeSys30: sysSummary?.slope30?.slope ?? null,
slopeDia30: diaSummary?.slope30?.slope ?? null,
sdSys30: (() => {
const fromMs = now - 30 * 24 * 60 * 60 * 1000;
const vals = sysData.filter(m => m.measuredAt.getTime() >= fromMs).map(m => m.value);
return stdDev(vals);
})(),
sdDia30: (() => {
const fromMs = now - 30 * 24 * 60 * 60 * 1000;
const vals = diaData.filter(m => m.measuredAt.getTime() >= fromMs).map(m => m.value);
return stdDev(vals);
})(),
pulsePressure30: (() => {
const avgSys = sysSummary?.avg30 ?? null;
const avgDia = diaSummary?.avg30 ?? null;
if (avgSys === null || avgDia === null) return null;
return Math.round((avgSys - avgDia) * 10) / 10;
})(),
pctInTarget,
coverage: computeCoverage(
[...sysData, ...diaData].sort(
@@ -425,6 +458,7 @@ export async function extractFeatures(
: null;
features.mood = {
scale: "1=LAUSIG, 2=SCHLECHT, 3=OKAY, 4=GUT, 5=SUPER_GUT",
avg7: avg(last7),
avg30: avg(last30),
latest: moodEntries[moodEntries.length - 1].score,
@@ -487,6 +521,10 @@ export async function extractFeatures(
});
if (medications.length > 0) {
const categoryMap = await getMedicationCategories(
medications.map((med) => med.id),
);
features.medications = [];
for (const med of medications) {
const events = await prisma.medicationIntakeEvent.findMany({
@@ -500,13 +538,16 @@ export async function extractFeatures(
}));
const c7 = calculateCompliance(mapped, med.schedules, 7, med.createdAt);
const c30 = calculateCompliance(mapped, med.schedules, 30, med.createdAt);
const c90 = calculateCompliance(mapped, med.schedules, 90, med.createdAt);
features.medications.push({
name: med.name,
dose: med.dose,
category: categoryMap[med.id] ?? "OTHER",
compliance7: c7.rate,
compliance30: c30.rate,
streak: c7.streak,
compliance90: c90.rate,
streak: c30.streak,
missedLast7: c7.missed,
});
}