Get-Cooking-

Cooking Recommendations

Overview

Cooking is both a passion and a big concern for some people. Cooking, on the other hand, can always use a helping hand. Being a student, deciding what to eat for lunch or dinner is always a challenge. When you only have a few ingredients in the kitchen, deciding what to make for dinner might be difficult. This prompted us to develop a system that can suggest recipes based on suggested ingredients. Recipe provider that uses web-scraped recipe data to provide users with the recipes based on their preferences and the ingredients they have.

Goals

1. Create a search algorithm that utilizes similarity scoring to rank recipes according to the greatest similarity to the search query.
2. To provide recipes when the user has limited items in the kitchen, or cannot decide what to cook for his/meal.
3. Recommend recipes based on the dish name
4. Recommend recipes based on the ingredients the user has with him/her.

Code Structure

main.py readME.md Food_search.py

Dataset and Pre-processing

Scraped data from : https://www.jamieoliver.com/

Content extracted:

• Recipe_urls
• Recipe name
• Ingredients
• Serves
• Cooking_time
• Difficulty

Pre-processed dataset:

• Recipe_urls
• Recipe
• Ingredients
• Ingredients_parsed (stopwords, measure, unnecessary words like fresh, etc. removed)

Modules

• Data Extraction and Pre-Processing
• API for Fetching Recipes - FastAPI
• Streamlit UI
• User Authentication and User Authorization - JWT
• Airflow workflows
• RecipeCache - For storing cache if the user has no ingredients and wants recipes to be suggested
• SendingEmails - when the user registers for subscriptions
• Data Validation and Unit testing
• Deployment on Cloud (GCP)
• Hosting Streamlit
• Visualizations
• Github Issues

Folder Structure

Streamlit UI

• Functionality: If the user provides ingredients, search through the dataset and recommend recipes based on the ingredients that match. If the user has no ingredients, provide recipes from cache

Includes:

• User authentication and authorization using JWT
• App design elements
• Import data validations using regex
• Storing the registered users’ info and uploading it into GCP storage buckets

Files:

• Streamlit.py - has the design of the UI with error handling
• FoodSearch.py - takes the ingredients and gives dataframe as output based on the user input

Airflow

Functionality:

• Automates the workflows like caching
• Deployment on cloud using Google Cloud Composer

Includes:

• Two automation tasks deployed on GCP
• Cloud Composer syntax of defining DAGs
• Storing the cache results on cloud

Files:

• Food_search_dag.py - for randomly updating cache daily to store 5 recipes
• Send_email_dag.py - to send emails to the registered users

Data Validations and Pytesting

Contains sample pytesting for the user input on Streamlit. The sample test cases and the status of the tests ran are uploaded in the same folder.

Data Visualization

Details:

• Once a new user registers, the user registry gets updated The registry automatically gets uploaded to Storage bucket.
• Using BigQuery and Data Studio, the visualization are obtained

Test our aplication

To run our application, click on this link : https://share.streamlit.io/niramay-kelkar/get_cooking/main/Streamlit/streamlit.py

References

1. Fast API Authentication : https://www.youtube.com/watch?v=6hTRw_HK3Ts
2. Apache Airflow : https://insaid.medium.com/setting-up-apache-airflow-in-macos-2b5e86eeaf1
3. Cloud Composer : https://towardsdatascience.com/airflow-by-google-a-first-impression-of-cloud-composer-f4d9f86358cf
4. Send Emails Using Code : https://www.youtube.com/watch?v=Y_tnWTjTfzY
5. Uploading to GCP : https://stackoverflow.com/questions/40683702/upload-csv-file-to-google-cloud-storage-using-python
6. Unit Testing : https://www.youtube.com/watch?v=byaxg00Gf9I
7. BigQuerying using files from GCP storage : https://cloud.google.com/bigquery/docs/loading-data-cloud-storage-csv