Chef-GPT is an AI-powered recipe assistant that helps users cook delicious meals with the ingredients they already have. It generates recipe suggestions tailored to dietary preferences (vegan, keto, low-calorie, etc.) and calorie goals. The system uses large language models (LLMs) and incorporates cutting-edge AI techniques such as Prompt Engineering, Function Calling, Structured Output, and Retrieval-Augmented Generation (RAG) to deliver highly personalized and structured cooking guidance.
- 🧾 Ingredient-Based Recipe Suggestions Input what you have in your kitchen and get smart recipe recommendations.
- 🥗 Diet-Friendly Customizations Adapt recipes to be vegan, keto, low-carb, high-protein, etc.
- 🔢 Calorie-Based Adjustments Automatically adjust ingredients to meet calorie goals.
- 📦 Structured Output in JSON Recipes are returned in a structured format with ingredients, steps, calories, and prep time.
- 📚 RAG-powered Retrieval Fetches recipes, substitutions, and cooking tips from a knowledge base or external database.
Chef-GPT aims to solve a common everyday problem: "What can I cook with what I already have?" Instead of searching through dozens of recipe websites, users can simply input available ingredients (e.g., "I have rice, eggs, and carrots"), and Chef-GPT provides a complete, personalized recipe. The system can:
- Customize meals based on dietary preferences.
- Suggest healthy alternatives.
- Return structured and ready-to-use instructions.
This is especially useful for people with specific health needs or limited pantry options.
Prompting helps convert the user's input into something the AI model can understand and respond to meaningfully.
- Zero-shot Prompting: For basic input like "I have oats and milk," the AI generates recipes without prior examples.
- Few-shot Prompting: Providing a few examples of ingredient-recipe pairs so the model learns to follow a pattern.
Example Prompt: "I have tomatoes, spinach, and eggs. Give me a high-protein breakfast recipe."
Function calling ensures more accurate and interactive AI behavior by delegating specific tasks to backend logic.
suggestRecipe(ingredients, diet)→ Fetches a suitable recipe.adjustCalories(target)→ Adjusts portion sizes or swaps ingredients.
This improves efficiency, correctness, and modularity in the codebase.
The model enhances its knowledge by retrieving external documents or database entries before generating output.
Chef-GPT uses RAG to:
- Retrieve cooking tips or substitutions.
- Pull verified recipes from a local or cloud-hosted dataset.
Example: If a user doesn’t have eggs, RAG can fetch vegan alternatives like flaxseed or tofu.
- Recipes match dietary preferences and calorie goals.
- Ingredients are used as requested.
- Instructions are logically sequenced.
- Uses function calling to reduce token usage and speed up response time.
- Caches frequent requests (like popular ingredients).
- Works with a growing recipe dataset (NoSQL or vector DB).
- Modular APIs allow scaling with more users or dietary modules (e.g., diabetic-friendly, gluten-free).
User: I have tofu, spinach, and rice. I want something vegan and under 400 calories. Chef-GPT: Suggests “Vegan Spinach-Tofu Rice Bowl” with ingredients, prep steps, time, and nutrition info.