recipe-generator — Генератор рецептов с учетом диеты и аллергий

--- name: recipe-generator description: Generate and manage AI-powered recipes for NutriProfile. Use this skill when working with recipe generation, ingredients management, cooking instructions, or the Recipes page. Handles multi-model consensus (Mistral, Llama, Mixtral) and user dietary preferences. allowed-tools: Read,Write,Edit,Grep,Glob,Bash --- # NutriProfile Recipe Generator Skill You are a recipe generation expert for the NutriProfile application. This skill helps you work with AI-powered recipe generation that considers user profiles, allergies, and nutritional goals. ## Context NutriProfile uses multi-agent AI (Mistral, Llama, Mixtral) with consensus validation for recipe generation. The system considers: - User dietary preferences (vegetarian, vegan, omnivore, etc.) - Allergies and food restrictions - Nutritional goals (weight loss, muscle gain, maintenance) - Available ingredients ## Architecture ### Backend Files - `backend/app/agents/recipe.py` - Recipe generation agent - `backend/app/models/recipe.py` - Recipe, FavoriteRecipe, RecipeHistory models - `backend/app/api/v1/recipes.py` - Recipe API endpoints - `backend/app/schemas/recipe.py` - Pydantic schemas ### Frontend Files - `frontend/src/pages/RecipesPage.tsx` - Main recipes page - `frontend/src/components/recipes/RecipeGenerator.tsx` - Generation form - `frontend/src/components/recipes/RecipeCard.tsx` - Recipe display card - `frontend/src/services/recipesApi.ts` - API service ## Data Models ### Recipe Model ```python class Recipe(Base): id: int user_id: int name: str description: str ingredients: List[dict] # [{name, quantity, unit}] instructions: List[str] prep_time: int # minutes cook_time: int # minutes servings: int calories_per_serving: float protein_per_serving: float carbs_per_serving: float fat_per_serving: float difficulty: str # easy, medium, hard cuisine_type: str tags: List[str] image_url: Optional[str] confidence_score: float created_at: datetime ``` ### RecipeHistory Model ```python class RecipeHistory(Base): id: int user_id: int recipe_id: int generated_at: datetime ingredients_used: List[str] preferences_applied: dict ``` ## API Endpoints ### Recipe Generation ``` POST /api/v1/recipes/generate { "ingredients": ["poulet", "riz", "brocoli"], "preferences": { "cuisine": "asian", "max_time": 30, "difficulty": "easy" } } Response: { "id": 1, "name": "Bowl Asiatique au Poulet", "description": "...", "ingredients": [...], "instructions": [...], "nutrition_per_serving": {...}, "confidence": 0.85 } ``` ### Other Endpoints - `GET /api/v1/recipes` - List user's recipes - `GET /api/v1/recipes/{id}` - Get specific recipe - `POST /api/v1/recipes/{id}/favorite` - Add to favorites - `DELETE /api/v1/recipes/{id}/favorite` - Remove from favorites - `GET /api/v1/recipes/favorites` - Get favorites ## Multi-Agent Consensus ### Recipe Agent Flow ```python async def generate_recipe(self, ingredients: List[str], profile: UserProfile): # 1. Build context with user profile context = self._build_context(ingredients, profile) # 2. Query multiple models in parallel results = await asyncio.gather( self.query_mistral(context), self.query_llama(context), self.query_mixtral(context) ) # 3. Consensus validation merged_recipe = self.consensus.merge_recipes(results) # 4. Calculate nutrition merged_recipe.nutrition = self.calculate_nutrition(merged_recipe.ingredients) return merged_recipe ``` ### Consensus Rules - Recipe name: Best rated by coherence - Prep/cook time: Average of all models - Ingredients: Union with quantity averaging - Instructions: Merge and order by step logic - Confidence: Minimum of individual confidences ## Freemium Limits | Tier | Recipes/Week | |------|--------------| | Free | 2 | | Premium | 10 | | Pro | Unlimited | Check limits in `backend/app/services/subscription.py`: ```python limits = { "free": {"recipe": 2}, "premium": {"recipe": 10}, "pro": {"recipe": -1} # unlimited } ``` ## Frontend Integration ### React Query Hooks ```typescript // Generate recipe const generateMutation = useMutation({ mutationFn: (data: RecipeRequest) => recipesApi.generate(data), onSuccess: (recipe) => { queryClient.invalidateQueries(['recipes']) toast.success(t('recipeGenerated')) } }) // Fetch recipes const { data: recipes } = useQuery({ queryKey: ['recipes'], queryFn: () => recipesApi.getAll() }) ``` ### i18n Namespace Use `recipes` namespace for translations: - `recipes.title` - Page title - `recipes.generate` - Generate button - `recipes.ingredients` - Ingredients label - `recipes.instructions` - Instructions label - `recipes.nutrition` - Nutrition info ## Best Practices 1. **Respect dietary restrictions** - Always filter recipes based on user allergies 2. **Calculate accurate nutrition** - Use per-ingredient values and sum 3. **Handle missing ingredients** - Suggest substitutions 4. **Support multiple cuisines** - French, Italian, Asian, Mediterranean, etc. 5. **Cache generated recipes** - Save to RecipeHistory for analytics ## Example Tasks ### Add New Cuisine Type 1. Update `CUISINE_TYPES` in recipe agent 2. Add prompt template for cuisine 3. Update frontend dropdown options 4. Add translations for all 7 languages ### Improve Recipe Quality 1. Review agent prompts in `recipe.py` 2. Adjust consensus weights 3. Add more detailed instructions generation 4. Test with various ingredient combinations ### Fix Nutrition Calculation 1. Check `calculate_nutrition()` in recipe agent 2. Verify ingredient quantities are parsed correctly 3. Cross-reference with nutritionReference database 4. Run backend tests: `pytest tests/test_recipes.py`