fix(ai): no default-positivity opener about data quality (mention only when limited)

Every insight summary opened with the same "Datengrundlage ist sehr
stark" / "Your data foundation is strong" filler — the user does
not see what fields the snapshot carries, so the opener reads as
awkward stock praise.

Add ground-rule 7 to both locales of the system prompt: don't open
with a compliment about the data quantity or quality, mention data
quality only when it materially limits the analysis (n<7 in the
analyzed window, recencyDays>14 since the last entry, or a coverage
gap). When data is fine, dive straight into the analysis.
Banned-opener phrases listed verbatim in both EN and DE so the
model does not paraphrase the same sentiment.

PROMPT_VERSION bumped 4.16.1 → 4.19.0 so feedback aggregation can
distinguish responses generated under the new rule.

New test file pins the rule, the trigger thresholds, the banned
phrases, and the dive-straight-in instruction in both locales.
Existing PROMPT_VERSION pins relaxed to allow any 4.x.x ratchet so
future minor bumps don't sweep up these assertions.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

src/app/page.tsx

@@ -45,7 +45,6 @@ import { summaryToTrend7Delta } from "@/lib/analytics/trend-delta";
 import { GettingStartedChecklist } from "@/components/onboarding/getting-started-checklist";
 import { TourLauncher } from "@/components/onboarding/tour-launcher";
 import { RecentAchievementsCard } from "@/components/gamification/recent-achievements-card";
-import { CompareToggle } from "@/components/comparison/compare-toggle";
 import { InsightsCardPreview } from "@/components/insights/insights-card";
 import { useInsightsAdvisorQuery } from "@/components/insights/use-insights-advisor";
 
@@ -492,11 +491,16 @@ export default function DashboardPage() {
         </DropdownMenu>
       </div>
 
-      {/* v1.4.16 phase D reconcile (CRITICAL C3) — on-surface comparison
-          toggle. The Settings-only Select shipped in B8 was buried 3
-          clicks deep; mounting the toggle here makes the Vormonat /
-          Vorjahr overlay flip a single tap on the page that uses it. */}
-      <CompareToggle />
+      {/* v1.4.19 A3 — comparison toggle removed from dashboard.
+          The tile strip is too tight for an extra control band, and
+          Marc rarely wants Vormonat / Vorjahr context at a glance —
+          when he does, it's on `/insights` where the chart sits next
+          to the toggle. The underlying `compareBaseline` value still
+          flows from `/api/dashboard/widgets` into every chart below
+          so nothing visual changes when the user has the comparison
+          on; only the on-surface affordance moves. Settings →
+          Dashboard remains the power-user backdoor for either
+          surface. */}
 
       {/* v1.4: Getting-started checklist for brand-new users.
        * Self-gates visibility on (onboardingCompletedAt == null

src/lib/ai/__tests__/medical-reference-prompt.test.ts

@@ -27,9 +27,13 @@ describe("PROMPT_VERSION", () => {
   it("is at least 4.16.0 for the v1.4.16 medical-reference grounding update", () => {
     // v1.4.16 phase B5a anchored at 4.16.0; phase B8 bumped the
     // patch component to 4.16.1 when the comparison-mode narrative
-    // ground rule landed. Use a string-prefix check so the test
-    // doesn't bind to a specific patch.
-    expect(PROMPT_VERSION.startsWith("4.16.")).toBe(true);
+    // ground rule landed. v1.4.19 bumped to 4.19.0 when the
+    // no-default-positivity opener ground rule landed. Use a numeric
+    // comparison so future ratchets stay covered without rewriting
+    // this assertion.
+    const [major, minor] = PROMPT_VERSION.split(".").map(Number);
+    expect(major).toBeGreaterThanOrEqual(4);
+    expect(major === 4 ? minor >= 16 : true).toBe(true);
   });
 });
 
@@ -110,7 +114,8 @@ describe("buildSystemPromptWithReferences()", () => {
     const prompt = buildSystemPromptWithReferences("en", ["bp"]);
     expect(prompt).toContain(PROMPT_VERSION);
     // v1.4.16: phase B5a was 4.16.0; phase B8 bumped to 4.16.1.
-    expect(prompt).toMatch(/4\.16\.\d+/);
+    // v1.4.19 bumped to 4.19.0 (no-default-positivity opener).
+    expect(prompt).toMatch(/4\.\d+\.\d+/);
   });
 });
 
@@ -122,6 +127,6 @@ describe("plain getStrictInsightsSystemPrompt() backward compatibility", () => {
 
   it("contains the bumped PROMPT_VERSION", () => {
     const en = getStrictInsightsSystemPrompt("en");
-    expect(en).toMatch(/4\.16\.\d+/);
+    expect(en).toMatch(/4\.\d+\.\d+/);
   });
 });

src/lib/ai/__tests__/no-default-positivity-opener.test.ts

@@ -0,0 +1,84 @@
+import { describe, it, expect } from "vitest";
+import {
+  getStrictInsightsSystemPrompt,
+  PROMPT_VERSION,
+} from "../prompts/insight-generator";
+
+/**
+ * v1.4.19 phase A4 — strip the default-positivity opener.
+ *
+ * Marc reported every insight summary started with the same
+ * "Datengrundlage ist sehr stark" / "Your data foundation is strong"
+ * filler sentence. The user does not see what fields the snapshot
+ * carries, so this opener reads as awkward filler.
+ *
+ * The new ground rule:
+ *   - Mention data quality ONLY when it materially limits the
+ *     analysis (n<7 in the analyzed window, recencyDays>14, or
+ *     coverage gap that biases the comparison).
+ *   - When data is fine, dive straight into the analysis.
+ *   - Banned opener phrases are listed verbatim in the prompt.
+ *
+ * These tests pin the rule's presence in BOTH locales so a future
+ * prompt rewrite cannot silently drop it.
+ */
+
+describe("v1.4.19 no-default-positivity opener rule", () => {
+  it("PROMPT_VERSION is at least 4.19.0", () => {
+    const [major, minor] = PROMPT_VERSION.split(".").map(Number);
+    // 4.19.x or any later major.
+    if (major === 4) {
+      expect(minor).toBeGreaterThanOrEqual(19);
+    } else {
+      expect(major).toBeGreaterThan(4);
+    }
+  });
+
+  it("English prompt forbids opening with a data-quality compliment", () => {
+    const prompt = getStrictInsightsSystemPrompt("en");
+    expect(prompt).toMatch(/do not open with a compliment about the data/i);
+  });
+
+  it("German prompt forbids opening with a data-quality compliment", () => {
+    const prompt = getStrictInsightsSystemPrompt("de");
+    expect(prompt).toMatch(/beginne nicht mit einem kompliment/i);
+  });
+
+  it("English prompt names the trigger thresholds for the data-quality caveat", () => {
+    const prompt = getStrictInsightsSystemPrompt("en");
+    expect(prompt).toContain("n<7");
+    expect(prompt).toContain("recencyDays>14");
+    expect(prompt).toMatch(/coverage gap/i);
+  });
+
+  it("German prompt names the trigger thresholds for the data-quality caveat", () => {
+    const prompt = getStrictInsightsSystemPrompt("de");
+    expect(prompt).toContain("n<7");
+    expect(prompt).toContain("recencyDays>14");
+    expect(prompt).toMatch(/coverage-lücke/i);
+  });
+
+  it("English prompt lists 'Your data foundation is strong' as a banned opener", () => {
+    const prompt = getStrictInsightsSystemPrompt("en");
+    // Allow whitespace incl. newline between phrase tokens so a future
+    // re-flow of the prompt doesn't break the assertion.
+    expect(prompt).toMatch(/Your data foundation\s+is strong/);
+  });
+
+  it("German prompt lists 'Datengrundlage ist sehr stark' as a banned opener", () => {
+    const prompt = getStrictInsightsSystemPrompt("de");
+    // Allow whitespace incl. newline at any token boundary so the
+    // assertion survives a future re-flow of the prompt.
+    expect(prompt).toMatch(/Datengrundlage\s+ist\s+sehr\s+stark/);
+  });
+
+  it("English prompt instructs to dive straight into analysis when data is fine", () => {
+    const prompt = getStrictInsightsSystemPrompt("en");
+    expect(prompt).toMatch(/dive straight\s+into the analysis/i);
+  });
+
+  it("German prompt instructs to dive straight into analysis when data is fine", () => {
+    const prompt = getStrictInsightsSystemPrompt("de");
+    expect(prompt).toMatch(/steige\s+sofort in die analyse\s+ein/i);
+  });
+});

src/lib/ai/prompts/insight-generator.ts

@@ -27,7 +27,7 @@ import {
 } from "../medical-references";
 
 /** Stable identifier for the active system prompt revision. */
-export const PROMPT_VERSION = "4.16.1" as const;
+export const PROMPT_VERSION = "4.19.0" as const;
 
 const SYSTEM_PROMPT_EN = `You are a clinical-context summariser for a personal health-log app.
 Prompt version: ${PROMPT_VERSION}.
@@ -83,6 +83,16 @@ GROUND RULES — ZERO HALLUCINATIONS
    comparison numbers, and do NOT extrapolate beyond the metrics
    listed. When the block reports "no prior-period data available"
    for every metric, state that explicitly and skip the narration.
+7. v1.4.19 — Do NOT open with a compliment about the data quantity
+   or data quality. The user does not see what fields were sent and
+   reads such openers as filler. Mention data quality ONLY when it
+   materially limits the analysis: n<7 readings in the analyzed
+   window, recencyDays>14 since the last entry, or a coverage gap
+   that biases the comparison. When data is fine, dive straight
+   into the analysis without commenting on it. Banned opener
+   patterns include "Your data foundation is strong", "Datengrundlage
+   ist sehr stark", "You have a solid baseline", "Great dataset" and
+   any rephrasing of the same sentiment.
 
 GUIDELINE TARGETS — generic, do NOT compute precise risk scores
 - Adult resting blood pressure (ESH/ESC 2024 generic): aim < 140/90
@@ -206,6 +216,17 @@ GRUNDREGELN — NULL HALLUZINATIONEN
    nicht über die gelisteten Metriken hinaus. Wenn der Block für
    alle Metriken "no prior-period data available" meldet, sag das
    explizit und lass die Narration weg.
+7. v1.4.19 — Beginne NICHT mit einem Kompliment über Datenmenge
+   oder Datenqualität. Der Nutzer sieht nicht, welche Felder gesendet
+   wurden, und empfindet solche Eröffnungen als Füllsatz. Erwähne
+   Datenqualität AUSSCHLIEßLICH dann, wenn sie die Analyse
+   substanziell einschränkt: n<7 Messwerte im analysierten Fenster,
+   recencyDays>14 seit dem letzten Eintrag, oder eine Coverage-Lücke,
+   die den Vergleich verzerrt. Bei ausreichender Datenlage steige
+   sofort in die Analyse ein, ohne die Datenlage zu kommentieren.
+   Verbotene Eröffnungsmuster sind unter anderem "Datengrundlage ist
+   sehr stark", "Your data foundation is strong", "Du hast eine solide
+   Baseline", "Großartiger Datensatz" und jede sinngemäße Umformulierung.
 
 LEITLINIEN-ZIELWERTE — generisch, KEINE genauen Risiko-Scores berechnen
 - Erwachsenen-Ruheblutdruck (ESH/ESC 2024 generisch): Ziel < 140/90