case study / Technical challenge for Assimetria
Auto-generated tech blog
Technical challenge: a blog that writes one new technical article every day with an LLM, served by a containerised stack on AWS.
- Role
- Built solo
- Year
- 2025
- Links
- Code
React · Vite · TypeScript · Node.js · Express · SQLite · Docker · AWS ECR · AWS CodeBuild · AWS EC2
The brief
Build a blog that generates a new technical article every day using an AI provider, stores it, exposes it through an API and shows it in a simple UI. Then deploy the whole thing to AWS with containers.
Back end
- Node.js + Express + TypeScript API with two endpoints:
GET /articlesandGET /articles/:id. - SQLite stored on disk, seeded with initial articles on first run.
- A daily
node-cronjob calls an LLM through OpenRouter, generates the article and saves it.
Front end
React + Vite + TypeScript, with a list of the latest articles and a detail page. The API base URL comes from an environment variable, so the same build works locally and in production.
Infrastructure
- Separate Docker images for the front end and the back end, plus Docker Compose for local development.
- AWS CodeBuild builds both images from the repository using
buildspec.ymland pushes them to ECR. - An EC2 instance runs a small deploy script: log in to ECR, pull the latest images and restart both containers.
What I took from it
Splitting the pipeline into build (CodeBuild), registry (ECR) and runtime (EC2) made each step easy to debug on its own. If I took it further, I'd move the database to a managed Postgres and replace the deploy script with a proper CI/CD trigger.