The problem
In 2024, the UK gov released an initative to make fuel pricing more competitive, forcing fuel stations to announce the prices at the pump so that customers could find the best deal. However, the data was hard to find, and even harder to parse, which is useful to nobody. I wanted to help consumers make concious decisions, and learn something new on the side.
Decisions that mattered
Small services, speaking gRPC
The different components of application were built as Go microservices communicating over gRPC, with a thin REST gateway for the browser and rate limiting pushed to the reverse proxy. Every piece stays small enough to hold in your head, with cross-dependencies held to a minimum.
Scrape, normalise, index
A scraper polls the published feeds every 15 minutes, normalises the data into simple shapes, before being indexed in PostGIS, making finding the cheapest fuel near you instant.
Kubernetes, even for one
Containerised in Docker and orchestrated on Kubernetes to ensure it was highly available. Deployments became a breeze and updates were seamless and hassle-free.
The outcome
A live site available to all that always knows the cheapest fuel nearby, refreshed every 15 minutes, with a mobile-first React frontend allowing it to be used on the go with zero effort. The backend service became powered by Go, allowing the services to be lightweight and fast. Every layer, built and run by one person, revisiting and refactaring kept simple.