ABATE AI

Reduce to increase. Customers often hesitate when placing orders online. ABATE AI encourages them to purchase by generating a persuasive message upon attempted item removal from an online cart.

Comment

Inspiration

Customers often hesitate when buying goods online. During checkout, they might remove items from their cart because they don't feel ready to purchase yet. Or perhaps they may abandon the cart altogether, which occurs at a rate of 70.19% according to the Baymard Institute. Businesses lose revenue to the tune of $18 billion annually when these events occur and shoppers miss out on discovering new products that exceed their expectations. This problem inspired the birth of ABATE AI.

What it does

When a customer attempts to remove an item from an online cart, ABATE AI generates a persuasive message explaining why he or she should keep the product and purchase it. By reducing the number of removal events, ABATE AI increases sales for businesses and helps connect customers to their new favorite products.

How we built it

Next.js frontend was paired with a Python FastAPI backend. I utilized Docker Compose to network and run servers during development. I deployed services to a Google Cloud Run production environment using GitHub Actions. All pizzas are stored as Catalog Objects in a Square sandbox account and are fetched when the website is loaded. The tailored, persuasive message is created by integrating item data from Square with the chat-bison-001 text prompt model from the MakerSuite/PaLM API.

Challenges we ran into

Getting the JavaScript frontend to communicate with the Python backend took a lot of research. The Docker build cache used up so much memory that my laptop had less than 200 MB remaining on the hard drive several times. Development with Docker Compose was very slow due to my laptop's limited processing power and RAM. Last-minute debugging with React hooks and state was another challenge that I encountered.

Accomplishments that we're proud of

  • Connected a JavaScript frontend to a Python backend in both development and production environments
  • Successfully deployed containerized services to Google Cloud Run, which I have not used before
  • First time using Square APIs (Catalog) and Google MakerSuite/PaLM API
  • Wrote a simple API using FastAPI
  • Built ABATE AI in less than four weeks (I became aware of the hackathon two weeks after it began)

What we learned

  • Learned the basics of Next.js, FastAPI, and Docker Compose within a few weeks
  • Discovered the power and simplicity of Square and Google APIs

What's next for ABATE AI

  • Implement ordering functionality with Square APIs

Built With

  • docker
  • docker-compose
  • fastapi
  • github-actions
  • google-cloud-run
  • makersuite
  • next.js
  • square
Share this project:

Updates