Littlecook.io

Accepts a user-selected list of ingredients and uses a large language model (likely GPT-3.5/4 or similar) to generate novel recipe instructions that incorporate those ingredients. The system likely maintains a prompt template that constrains output format (ingredients list, steps, cook time, servings) and may apply post-processing to validate recipe coherence. Generation happens server-side with caching to reduce API costs for popular ingredient combinations.

Allows users to specify dietary constraints (vegetarian, vegan, gluten-free, keto, etc.) and cuisine preferences (Italian, Asian, Mexican, etc.) as filters applied before or during recipe generation. The system likely encodes these as prompt modifiers or post-generation filtering rules to ensure output recipes respect user constraints. Implementation may use keyword matching or semantic understanding to validate generated recipes against specified restrictions.

Provides guidance on ingredient quantities (cups, grams, tablespoons) for each ingredient in the generated recipe and suggests common substitutions if a user lacks a specific ingredient. The system likely uses LLM knowledge of cooking ratios and ingredient chemistry to generate proportions and alternatives, possibly with fallback to heuristic rules for common substitutions (e.g., butter ↔ oil, milk ↔ plant-based alternatives). Substitution suggestions may be ranked by compatibility (flavor, texture, cooking properties).

Analyzes generated recipes to estimate cooking difficulty (beginner, intermediate, advanced) and total cook time (prep + active cooking + passive time). The system likely uses heuristic rules based on ingredient count, cooking techniques mentioned (e.g., 'sauté', 'braise', 'temper'), and equipment required, possibly combined with LLM reasoning to classify difficulty. Cook time may be extracted from generated recipe text or estimated based on cooking method patterns.

Implements a freemium model where free users can generate a limited number of recipes per day/week (likely 3-5 recipes) and access basic features, while premium users get unlimited generation, saved recipe history, and advanced filters. The system uses session/account tracking to enforce rate limits and stores user-generated or favorited recipes in a database (likely with user authentication). Free tier likely has no persistent storage; premium tier stores recipes with metadata (generated date, ingredients used, dietary filters applied).

Allows users to share generated recipes via URL, social media, or email, and potentially discover recipes shared by other users or trending recipes based on popularity. The system likely generates shareable recipe URLs with recipe data encoded in the URL or stored in a database, and may implement a social feed or trending section showing popular recipes. Sharing may include recipe metadata (ingredients, difficulty, cook time) in preview cards for social platforms.

Estimates nutritional content (calories, protein, carbs, fat, fiber, sodium) for generated recipes based on ingredient quantities and cooking methods. The system likely uses a nutrition database (USDA FoodData Central or similar) to look up ingredient nutritional values, applies cooking loss factors (e.g., water evaporation during roasting), and aggregates per serving. May provide macro breakdowns and allow users to track daily nutritional intake against dietary goals (calorie targets, macro ratios).