nutrition-analyzer

name	nutrition-analyzer
description	Analyze food items and calculate nutritional values for NutriProfile. Use this skill when working with food detection, nutritional calculations, meal analysis, or the Vision page. Covers calories, proteins, carbs, fats, and portion estimation.
allowed-tools	Read,Write,Edit,Grep,Glob,Bash

NutriProfile Nutrition Analyzer Skill

You are a nutrition analysis expert for the NutriProfile application. This skill helps you work with food detection, nutritional calculations, and meal analysis features.

Context

NutriProfile uses multi-agent AI (BLIP-2, LLaVA) for food detection from photos. The nutrition data is stored in:

• Backend: backend/app/models/food_log.py - FoodLog, FoodItem, DailyNutrition models
• Frontend: frontend/src/data/nutritionReference.ts - Local nutrition database with 30+ foods
• API: backend/app/api/v1/vision.py - Vision analysis endpoints

Nutrition Reference Database

The local nutrition database (nutritionReference.ts) contains per-100g values:

• calories: kcal
• protein: grams
• carbs: grams
• fat: grams

Supported Foods (30+)

Common foods: riz, pâtes, pain, poulet, boeuf, saumon, oeuf, lait, fromage, yaourt, pomme, banane, orange, carotte, brocoli, tomate, salade, pomme de terre, haricots, lentilles, huile d'olive, beurre, miel, chocolat, café, thé, jus d'orange, eau, vin, bière

Calculations

Portion Calculation

function calculateNutrition(food: string, quantity: number, unit: string): NutritionValues {
  const baseNutrition = NUTRITION_DATABASE[normalizeFood(food)]
  const multiplier = getUnitMultiplier(unit) * (quantity / 100)

  return {
    calories: Math.round(baseNutrition.calories * multiplier),
    protein: Math.round(baseNutrition.protein * multiplier * 10) / 10,
    carbs: Math.round(baseNutrition.carbs * multiplier * 10) / 10,
    fat: Math.round(baseNutrition.fat * multiplier * 10) / 10
  }
}

Unit Multipliers

• g (gram): 1.0
• ml (milliliter): 1.0
• portion: 150g equivalent
• piece/pièce: 100g equivalent
• cup/tasse: 240g equivalent
• tablespoon/cuillère: 15g equivalent

Backend Integration

FoodLog Model Structure

class FoodLog(Base):
    id: int
    user_id: int
    meal_type: str  # breakfast, lunch, dinner, snack
    image_url: Optional[str]
    detected_items: List[dict]  # AI detected
    user_corrections: List[dict]  # User edits
    total_calories: float
    total_protein: float
    total_carbs: float
    total_fat: float
    confidence_score: float
    created_at: datetime

Vision API Endpoints

• POST /api/v1/vision/analyze - Analyze food photo
• GET /api/v1/vision/logs - Get user's food logs
• GET /api/v1/vision/logs/{id} - Get specific log
• PATCH /api/v1/vision/logs/{id} - Update log (user corrections)
• DELETE /api/v1/vision/logs/{id} - Delete log

Frontend Components

Key Components

• VisionPage.tsx - Main page with camera/upload
• AnalysisResult.tsx - Display detected foods (pre-save editing)
• FoodLogCard.tsx - Display saved food logs (post-save editing)
• EditFoodItemModal.tsx - Modal for editing food items
• EditFoodItemModalV2.tsx - Enhanced modal with USDA integration

State Management

Uses React Query for:

• useVisionLogs() - Fetch food logs
• useSaveFoodLog() - Save new analysis
• useUpdateFoodLog() - Update existing log
• useDeleteFoodLog() - Delete log

Best Practices

1. Always validate nutrition values - Use the local database for estimation
2. Handle unknown foods - Return default values (100 kcal, 5g protein per 100g)
3. Preserve user corrections - Never overwrite user edits with AI values
4. Calculate totals correctly - Sum all items in a meal
5. Support i18n - Use translation keys from vision namespace

Example Tasks

Add New Food to Database

1. Add entry to NUTRITION_DATABASE in nutritionReference.ts
2. Add food name aliases if needed
3. Add tests in nutritionReference.test.ts
4. Update food suggestions in autocomplete

Fix Nutrition Calculation Bug

1. Check calculateNutrition() function
2. Verify unit multipliers
3. Test with edge cases (0 quantity, unknown food)
4. Run tests: npm test

Enhance Food Detection

1. Review Vision Agent in backend/app/agents/vision.py
2. Check consensus validation in backend/app/agents/consensus.py
3. Add new model or improve prompts
4. Test with diverse food images