Calry

Full-stack calorie tracking app with a FastAPI backend and React frontend.

Stack

• Backend: Python, FastAPI, SQLAlchemy, SQLite
• Frontend: React, Vite
• Containers: Docker Compose

Run

docker compose up --build

The app will be available on:

• Frontend: http://localhost:8081
• Backend API docs: http://localhost:5001/docs

Main Features

• Register and log in with email and password
• Update your own profile data and calorie target estimation
• Manage ingredients with calories and macros per 100 grams
• Build recipes from ingredient quantities with automatic total weight, calorie, protein, carbohydrate, and fat calculation
• Log recipe portions into daily meals
• Edit and delete ingredients, recipes, and meal entries
• See consumed, remaining, grouped daily calories, and consumed macros
• Optional per-user AI mode for Gemini-powered ingredient nutrition and recipe suggestions

Data Model

• User: authenticated account with profile data
• Ingredient: base ingredient with calories, protein, carbohydrates, and fat per 100g
• Recipe: reusable recipe composed of ingredients, with calculated nutrition totals
• MealEntry: eaten grams of a recipe on a given date and meal type, with calculated nutrition values

Notes

• age is optional. If provided, daily calorie estimation uses Mifflin-St Jeor. Otherwise a simpler weight/height-based estimate is used.
• CORS is restricted to local frontend origins by default. Set CORS_ORIGINS as a comma-separated list when the API must be called from other browser origins.
• AI mode is disabled by default per user. Set GEMINI_API_KEY for the backend container, and optionally GEMINI_MODEL (defaults to gemini-3.1-flash-preview).
• If you already ran an older local version, legacy SQLite columns are handled for compatibility, but old anonymous users are not converted into email/password accounts automatically.

Local Admin User

Create a normal account in the UI first, then promote that account locally.

With Docker Compose:

docker compose exec backend python -m app.make_admin user@example.com

Without Docker, from the backend directory:

python -m app.make_admin user@example.com

After the command succeeds, log out and log back in with that user. The Admin button appears in the footer. In the Admin / Users view, admins can also toggle AI integration per user. That toggle updates the user's ai_enabled flag through:

PATCH /admin/users/{user_id}

Request body:

{
  "ai_enabled": true
}

AI Integration

AI integration is per user and is off by default. It works only when both conditions are true:

• the backend has GEMINI_API_KEY configured
• the current user's ai_enabled flag is true

For local Docker Compose, set the key before starting the app:

GEMINI_API_KEY="your-key" docker compose up --build

Optional model override:

GEMINI_MODEL="gemini-3.1-flash-preview"

The backend checks the current authenticated user before every AI request. If ai_enabled is false, the API returns 403.

AI Status

GET /ai/status

Response:

{
  "enabled": true,
  "configured": true
}

enabled means the logged-in user has AI mode enabled. configured means the backend has a Gemini API key.

Ingredient Nutrition Suggestion

POST /ai/ingredient-nutrition

Request model:

{
  "name": "chicken breast"
}

Response model:

{
  "name": "chicken breast",
  "calories_per_100g": 165,
  "protein_per_100g": 31,
  "carbs_per_100g": 0,
  "fat_per_100g": 3.6,
  "note": "Common raw/plain form."
}

UI flow:

• The user types an ingredient name in the ingredient form.
• The AI button sends { "name": ingredientForm.name }.
• The response fills calories_per_100g, protein_per_100g, carbs_per_100g, and fat_per_100g in the form.
• Nothing is stored yet. The ingredient is saved only when the user submits the normal ingredient form, which calls POST /ingredients or PUT /ingredients/{id}.

Recipe Suggestions

POST /ai/recipe-suggestions

Request model:

{
  "only_existing_ingredients": true,
  "prompt": "high protein dinner"
}

Response model:

[
  {
    "name": "Chicken Rice Bowl",
    "instructions": "Cook rice, grill chicken, combine with vegetables.",
    "ingredients": [
      {
        "ingredient_id": 1,
        "ingredient_name": "chicken breast",
        "amount_grams": 180
      },
      {
        "ingredient_id": 2,
        "ingredient_name": "rice",
        "amount_grams": 120
      }
    ]
  }
]

UI flow:

• The user sets only_existing_ingredients and an optional prompt.
• The UI sends those fields to /ai/recipe-suggestions.
• The backend sends available ingredients and the prompt to Gemini, then normalizes the JSON response.
• Suggestions are shown as temporary recipe cards.
• Clicking Use recipe copies the suggestion into the recipe form.
• Nothing is stored yet. The recipe is saved only when the user submits the normal recipe form, which calls POST /recipes or PUT /recipes/{id}.

When only_existing_ingredients is true, the backend tries to return suggestions with existing ingredient_id values. The UI only copies suggestion ingredients that have an ingredient_id, because saved recipes must reference stored ingredients.