Imported from deonmenezes/robotics-for-social-good (AGENTS.md). Install upstream with npx skills add deonmenezes/robotics-for-social-good. Copyright stays with the author.

Robotics for Social Good

A browse-first library of robotics videos labeled for social good use cases, deployed as a static site with Python serverless API functions on Vercel.

Tech Stack

• Frontend: Plain HTML, CSS, vanilla JavaScript (no build step)
• Backend: Python (Flask), serverless via @vercel/python
• Database: Supabase (supabase-py)
• ML: NomadicML SDK (nomadicml)
• Deployment: Vercel (static + Python functions)

Setup

pip install -r requirements.txt

Create a .env file with:

SUPABASE_URL=...
SUPABASE_KEY=...
NOMADICML_API_KEY=...

Build / Run / Test

This is a static site with serverless API routes — no local server command is defined in the repo. To run locally:

# Install Vercel CLI and run locally
npx vercel dev

Or open index.html directly in a browser for the frontend (API calls will not work without a server).

Project Structure

robotics-for-social-good/
├── index.html          # Main landing/browse page
├── explore.html        # Dataset exploration page
├── upload.html         # Video upload page
├── impact.html         # Impact/about page
├── pricing.html        # Pricing page
├── api/
│   └── upload.py       # Serverless Python handler for video uploads
├── js/
│   ├── main.js         # Core site JS
│   ├── datasets.js     # Dataset listing logic
│   ├── explore.js      # Explore page logic
│   ├── home.js         # Home page logic
│   ├── upload.js       # Upload form handling
│   └── shared.js       # Shared utilities
├── css/                # Stylesheets
├── assets/             # Images and media
├── data/               # Static data files
├── labels.json         # Video label taxonomy
├── label_videos.py     # Script for labeling videos via NomadicML
├── supabase_setup.sql  # Database schema
├── requirements.txt    # Python dependencies
└── vercel.json         # Vercel deployment config

Architecture & Key Files

• Static HTML pages are served directly; no client-side routing framework.
• api/upload.py handles video file uploads as a Vercel Python serverless function.
• label_videos.py is a one-off script to batch-label video content using the NomadicML API.
• supabase_setup.sql sets up the Supabase tables; run it in your Supabase project before deploying.
• labels.json defines the social good categories used throughout the app.

Conventions & Notes for Agents

• No Node.js or build toolchain — the frontend is pure HTML/CSS/JS; edits take effect immediately.
• Python dependencies are for the serverless API only (api/*.py).
• SUPABASE_URL, SUPABASE_KEY, and NOMADICML_API_KEY must be set as environment variables (.env locally, Vercel env vars in production).
• improve.sh and improve.log are development iteration scripts; they are not part of the production app.
• The vercel.json routes /api/upload to api/upload.py; add new API routes there if extending the backend.