AaharWise

Personal Wellness Forecaster

Comment

My AaharWise Journey: Building a Nutrition Coach

Inspired by real health struggles and endless app iterations, I created AaharWise – a smart nutrition tracker that doesn't just count calories, but predicts your metabolic future.

What Inspired Me

Seeing friends in Bengaluru battle silent killers like diabetes and fatty liver from desk jobs + chaat binges, I wanted more than a calorie counter. AaharWise became my mission: use AI to analyze food photos, track history, and deliver "Metabolic Oracle" verdicts – blunt truths about what today's dosa is doing to your 10-year health trajectory.

Key Learnings

1. Iterative Polish > Perfect First Draft

30+ Gemini iterations transformed a buggy calorie counter into a glassmorphism beauty
From: Syntax errors → Pink overload → Subtle slate gradients + dark mode
Learned: Small visual tweaks (20px cards, colored gauges) create 10x perceived polish

2. Health Communication = Art + Science

  • Dry "1500 calories" → "INSULIN SPIKE DETECTED" grabs attention
  • History analysis reveals patterns: "3rd high-sugar week = pre-diabetic trajectory"
  • Creative severity: "SYSTEM: OPTIMIZED" vs "CELLULAR DEBT ACCUMULATING"

How I Built It

Phase 1: Core AI Nutrition Engine
├─ Image → Calories/Macros (Gemini Vision)
├─ Mandatory ingredient input for juices/milkshakes
└─ Historical pattern analysis for long-term forecasts

Phase 2: Visual Excellence
├─ SVG Fuel Gauge (fixed clipping: viewBox + 85px radius)
├─ Macro distribution pie with dark-mode contrast
├─ Subtle light theme: `linear-gradient(135deg, #f8fafc → #e2e8f0)`
└─ Glass cards: `backdrop-filter: blur(12px) + 20px radius`

Phase 3: Clinical Features
├─ CSV Export (doctor-ready logs)
├─ Timeline charts (calories/water trends)
└─ Biological forecasting based on age + intake patterns

Tech Stack: React/TSX + Gemini AI + SVG visualizations + localStorage persistence

Challenges Conquered

Problem Solution Impact
Fuel gauge clipped edges viewBox="0 0 192 192" + radius 85px Perfect circular progress
Metabolic Oracle text overflow Regex parser for markdown headers Clean "STATUS: INSULIN SPIKE" extraction
Dark theme macro invisibility #10b981 → #f3f4f6 contrast fix Readable protein/fiber labels
"Too whitish" light theme Slate gradient background Premium wellness feel
Juice calorie guesswork Mandatory ingredient modal Precise sugar tracking

The Result

AaharWise now delivers:

Upload plate photo → Instant macro breakdown
3-day history → "Pre-diabetic trajectory warning"
Coach verdict → "SYSTEM: RECOVERY MODE" + 1 fix
Export → Doctor-ready CSV logs

Most proud moment: When the AI caught my 3rd consecutive high-carb day and warned: "Continuing this = 35% diabetes risk by 35." Science + design working together.

Prototype by Vaibhavi Hiremath – Ready for your plate! 🍛📈

Built With

  • gemini-api
  • geminiai
  • google-ai-studio
  • react/tsx
  • tailwindcss
Share this project:

Updates