We built Sero Listens, a social listening tool that tracks what people are saying about our app across the internet. The idea was simple: instead of manually searching Reddit, Twitter, LinkedIn, YouTube, and Instagram every day, we wanted one tool that does it all automatically and gives us a clean report.
What It Does
The tool searches four platforms — Reddit, Twitter/X, LinkedIn, and YouTube — for mentions of our app using configurable keywords. It collects each mention along with useful metadata: who posted it, when, what they said, engagement numbers (likes, comments, views), and even a basic sentiment reading (positive, negative, or neutral). All of this gets packaged into reports — JSON for raw data, a readable text file, and a polished PDF with charts showing sentiment breakdown, posts per day, and platform-level stats.
There’s also a web dashboard built with Next.js where the team can log in with their Google account (restricted to team emails only) and browse/download past reports. Reports are stored on Vercel Blob and the dashboard pulls them automatically.
How It Works (Tech)
The scraper runs locally via a Python CLI. Reddit is hit through its public JSON API using curl_cffi, which impersonates a real browser’s TLS fingerprint to avoid getting blocked. Twitter, LinkedIn, and YouTube are scraped using Playwright (headless browser automation) since these platforms are JavaScript-heavy and don’t offer easy public APIs. Twitter and LinkedIn require persistent login sessions stored locally.
Once reports are generated, they’re pushed to GitHub. A GitHub Actions workflow picks up new report files, uploads them to Vercel Blob storage, and cleans them from the repo. The web dashboard (Next.js 16, React 19, Tailwind CSS, NextAuth for Google OAuth) then reads from Blob to display reports.