YouTube Video to Recipe Generator

This application transcribes a YouTube cooking video and uses OpenAI's GPT-3 model to summarize the transcript into a recipe.

Try it at https://videorecipegen.com/

Features

• Transcribes YouTube cooking videos into text.
• Summarizes the transcription into a recipe using OpenAI's GPT model.
• Formats the response into a recipe card.
• Limits the number of requests per user to prevent abuse.

Technologies Used

Backend

• Flask: Used to serve the application.
• Redis: Used to store the number of requests per user.

Frontend

• Pico CSS: A minimalistic CSS framework.
• HTMX: Used to provide dynamic HTML content.

Infrastructure

• Kubernetes (EKS): Application and Redis are currently run in EKS, installed via the helm chart

Local installation

1. Clone the repository.
2. Install the dependencies with pip install -r requirements.txt.
3. Run a redis docker container with docker run --name my-redis -p 6379:6379 -d redis.
4. Create a .env file with your OPENAI_API_KEY and YOUTUBE_API_KEY values, you may also optionally set environment variables for the Redis, request limit and max video length settings.
5. Run the application from the /flask directory with flask --app main run.

Usage

1. Submit a YouTube cooking video URL.
2. The application will transcribe the video and generate a recipe.
3. If the video is not a cooking video or is greater than the MAX_VIDEO_DURATION, an error message will be displayed.

Limitations

• The application limits the number of requests to 5 per user per day. This is done using the IP address of the user.
• The application limits the duration of the video to 45 minutes. This is to prevent inputs greater than the models max tokens.
• The accuracy of the recipe depends on the quality of the video transcription and the performance of the GPT-3 model.

CI/CD

CI Actions can be found at ./github/workflows/. Container images are automatically built on pushes to main. Linting/Formatting is automatically ran on every push, using ruff

CD automation is WIP, currently deployed via helm chart into EKS. Migrating to ArgoCD GitOps running on a homelab K8s cluster.

Tests

Building out the testing coverage. Tests can be run with pytest from the root directory using the command pytest --envfile flask/.env to set up the env.