Track what you eat and how your digestion responds — over time, identify which ingredients are correlated with digestive issues.

• AI-powered meal logging — describe your meal in plain text, and GPT-4o-mini automatically extracts individual ingredients (powered by Val Town's built-in OpenAI integration — no API key needed!)
• Bristol Stool Scale — track bowel movements using the medically-recognized 7-type scale (Type 1 = hard lumps/constipation → Type 7 = watery/diarrhea), with visual color indicators and category labels
• Digestion symptom tracking — log bloating, gas, and pain/cramps on a 0–5 scale with color-coded severity buttons
• Date & time backfilling — pick any date/time when logging a meal or symptom, so you can add entries retroactively. On iOS, this uses the native wheel picker.
• Smart correlation — ingredients are automatically linked to symptoms logged within 24 hours, or you can manually link a symptom to a specific meal
• Insights dashboard — see which ingredients correlate with worse-than-average or better-than-average digestion, ranked by deviation from your baseline

Built to look and feel like a native iOS app:

• SF Pro font stack via -apple-system
• iOS system colors (#007AFF blue, #34C759 green, #F2F2F7 grouped background)
• Frosted glass tab bar with backdrop-filter: blur()
• Large title navigation pattern
• Inset grouped table views with 12px rounded corners
• Safe area insets for notch/Dynamic Island devices
• PWA meta tags for "Add to Home Screen"

Rendering mermaid diagram...

index.ts                       ← Hono server (API routes + frontend serving)
lib/
  db.ts                        ← SQLite schema, queries, correlation analysis
  ai.ts                        ← OpenAI ingredient extraction via std/openai
frontend/
  index.html                   ← HTML shell (iOS fonts, Twind, safe areas)
  index.tsx                    ← React entrypoint
  favicon.svg                  ← App icon
  components/
    App.tsx                    ← Root layout + iOS frosted tab bar
    LogMeal.tsx                ← Meal entry + AI extraction + date picker
    LogSymptom.tsx             ← Symptom ratings + Bristol Stool Scale + date picker
    History.tsx                ← iOS grouped list with date section headers
    Insights.tsx               ← Ingredient correlation cards with delta bars

• Frontend: React 18 + Twind (Tailwind at runtime, no build step)
• Backend: Hono on Val Town
• Database: Val Town built-in SQLite (std/sqlite)
• AI: Val Town's built-in OpenAI proxy (std/openai) — free tier covers gpt-4o-mini, no API key required

1. Open the app on your phone (it's mobile-first!)
2. Meal tab → type what you ate → tap "Extract Ingredients with AI" → review/edit the ingredient chips → pick a date if backfilling → save
3. Symptoms tab → rate bloating/gas/pain → select Bristol Stool Scale type → optionally link to a meal → pick a date if backfilling → save
4. History tab → see all entries grouped by date with Bristol color indicators
5. Insights tab → see which ingredients correlate with bad digestion

The more meals and symptoms you log, the more reliable the correlations become. At least 2 meals per ingredient are needed before it appears in insights.