Transparent Social Feed demo

A polished static Caution demo prototype for verifiable social-media ranking algorithms.

The app uses only synthetic example data. It is not connected to a real social network and does not process private user data. Its purpose is to show how a feed can disclose the exact algorithm, formula, weights, inputs, score breakdown, and final ordering behind every item.

What it demonstrates

• Multiple feed algorithms: chronological, popularity, following, topic affinity, and balanced/recommended.
• Realistic synthetic inputs: users, follows, posts, tags, likes, replies, shares, recency, and explicit topic preferences.
• A per-post "why am I seeing this?" explanation with every score component.
• A debug panel showing the active formula, weights, user profile, normalized inputs, and final ranked order.
• Caution deployment conventions: static assets, explicit Procfile, pinned StageX Containerfile, small resource settings, and e2e: true.

Run locally

Quick static preview without containers:

cd /home/tuna/git/demo-transparent-social-feed
python3 -m http.server 8083 --directory public

Then open http://localhost:8083.

Container workflow:

make check
make build
make dev

Stop the container:

make stop

Caution deployment

The Procfile serves /workdir/public with BusyBox httpd on port 8083:

run: /bin/busybox httpd -f -p 0.0.0.0:8083 -h /workdir/public

Containerfile pins the StageX BusyBox base image by digest and copies only static assets into the runtime image. No external paid APIs or runtime network access are required.

Verification notes

The app intentionally computes rankings client-side from inspectable static data so demo visitors can compare the visible formula against the actual code in public/app.js. The dataset is synthetic and labeled as such in the UI.