naina-sriv/flash-sale-engine

Flash Sale Engine is a backend system built to handle flash sales, the kind of event where thousands of people try to buy a heavily discounted item at the exact same moment. It is built with FastAPI (a Python web framework), Redis (a fast in-memory data store), and PostgreSQL (a database for long-term storage), and it runs behind Nginx, a tool that spreads incoming traffic across multiple copies of the application. The core idea is that the system never makes a shopper wait on the slower database during checkout. Instead, it uses Redis to instantly check and reduce stock counts the moment someone tries to buy an item, which prevents two people from accidentally buying the last unit at the same time. The actual order record is then saved to PostgreSQL afterward, in the background, so the checkout itself stays fast even under heavy load. To keep things fair and secure, the system includes a duplicate purchase lock so one person cannot submit the same order twice while it is still processing, plus a math challenge shoppers must solve before buying, meant to slow down bots. It also has rate limiting on login and purchase requests, and login is handled with short-lived JWT tokens paired with longer-lived refresh tokens, with password hashing done in a way that does not block other requests. The project ships with Docker Compose so the whole setup, three copies of the API, Nginx, Redis, and Postgres, can be started with one command. It also includes a Locust load test so a developer can simulate a thousand simultaneous buyers and see how the system holds up. The author says the project was inspired by real flash sale crashes, where cheap items caused so much simultaneous demand that ordinary database driven stores fell over. This project is essentially a hands-on demonstration of how to design a backend around that specific problem using fast, atomic operations instead of relying on the database to sort out the traffic under pressure.