Case 06

Article Summarizer

Handling untrusted web content safely

DESIGNVERIFYOPERATE

An RSS reader that strips remote HTML down to plain text before any of it reaches a language model.

How it fits together

RSS feedfetchersanitizermodel APIReact client

The problem

RSS feeds hand you someone else's HTML. Passing it straight through gives messy summaries and opens an input boundary you did not mean to open.

Where it could go wrong

What is exposed

RSS input
remote HTML
model boundary

Checks in place

HTML parsing
content sanitization
server-side API mediation

Result

clean text extraction
real-time summaries
provider abstraction

How I built it

Spring Boot fetches the feeds server-side, so the API key never reaches the browser. Jsoup strips the markup down to text before it goes to OpenRouter, and React displays the summaries.

JavaSpring BootReactJsoupOpenRouter API

What it does

Jsoup content sanitization
Server-side feed fetching
Clean model input
Java and React architecture

What I would improve

Restrict outbound fetch destinations
Add size, timeout, and content-type limits
Treat remote text as untrusted instructions at the model boundary

Code

View repository