RecipeGPT

Easily find recipes based on what ingredients you have at home! Inspired by JainRecipes!

Table of Contents

1. About The Project
   • Built With
2. Getting Started
   • Installation
3. License

About The Project

RecipeGPT is a project that can create recipe recommendations. Using Flask and an available Hugging Face GPT-2 Model that was trained on the RecipeNLG dataset, RecipeGPT will both provide the recommended JainRecipes along with an assortment of GPT-2 generated recipes. Users will be able to select what ingredients they have at home from a comprehensive list scraped from JainRecipes and will be provided with a personalized culinary recommendation.

(back to top)

Built With

• 
• 
• 
• 

Relevant libraries include:

• flask
• json
• Levenshtein/fuzz (for similar string identification)
• transformers/hugingface (for the gpt-2 model)
• re
• BeautifulSoup
• markdown
• datetime
• random
• os/dotenv
• requests
• time

(back to top)

Getting Started

Installation

To setup on your local machine

1. Generate a GitHub Access Key (needed to scrape JainRecipes) & a Spoonacular API key if you wish to use spoonacular for image/recipe link generation
2. Clone the repo

   git clone https://github.com/JainRecipes/RecipeGPT.git

3. Install all prerequisites (requirements.txt has all pip libraries and extras from other projects I have worked on)

   pip install -r requirements.txt

4. Enter your API in .env

   GITHUB_ACCESS_TOKEN = 
   SPOONACULAR_API_KEY = 

5. Run the app

python app.py

(back to top)

License

Distributed under the MIT License. See LICENSE.txt for more information.

(back to top)

Credit to Pratul Tandon for training the GPT-2 model for recipe generation based on the RecipeNLG dataset. More info here https://huggingface.co/pratultandon/recipe-nlg-gpt2