Your plants can’t talk, but your phone can listen - photograph any plant and get quick, practical advice
Live demo: https://bedsense.vercel.app
Video Link BedSense
Just photograph the plant — BedSense names it from the photo. If it names the wrong plant, you can correct it on the result and re-check. You can also add one optional note: plant age or stage, current practices, chemicals already used, recent fertilizer, irrigation or weather, known issues, or anything else.
Farmers on farm beds, and the extension officers who walk those beds with them.
BedSense is an indie entry for the Meta Global AI Developer Hackathon, Indie track, Build for Your World. The screen is in English or Kiswahili.
The photo is read with the plant name you gave. When a symptom is visible, you get likely causes, practical mitigation, and three next checks for that plant. Example Kenyan agrovet shops and KES price bands appear when a spray is the usual next step, or when a fertilizer fits the symptom. When the photo does not separate the likely causes, no chemical is named: you get the safe steps that can start now and a plain statement of the angle or fact still needed. Every check ends with a follow-up question under the three next checks, and a box there lets you ask questions about the feedback without running a new check. Spray rates follow the product label. Confirm with a local agroveterinary officer before you buy or spray.
 |
 |
 |
 |
| Home | A chili check with three next checks | The same check in Kiswahili | Example shops and KES price bands |
Sign in and save a plant, and later checks are read against its history: earlier findings, what you sprayed or fed, and whether things improved. Each check also sets when to look again - the model judges how soon improvement or worsening would show, so a treated plant is re-checked after the spray has had time to work, and a healthy one only weekly.
You need Node.js 20.17 or newer (22 recommended), npm (ships with Node), and git. No API keys are needed: with an empty environment file the app runs the full demo on synthetic data.
macOS: install Node and git with Homebrew (brew install node git), or the LTS installer from nodejs.org, or nvm install 22.
Windows: install Node.js LTS from nodejs.org (or winget install OpenJS.NodeJS.LTS) and Git for Windows (gives you Git Bash). Prefer a version manager? Use nvm-windows, then nvm install 22.
Linux (Ubuntu/Debian): curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash - then sudo apt-get install -y nodejs git (or use nvm). Fedora: sudo dnf install -y nodejs git, checking the version is ≥ 20.17.
Then, in a terminal (Git Bash or PowerShell on Windows):
git clone https://github.com/benweru/BedSense.git
cd BedSense
npm installCopy the repo's example environment file to .env.local, start the dev server, and open http://localhost:3000 — on your phone or in a narrow browser window. The exact commands, the settings table, mock versus live mode, accounts, scripts, and troubleshooting are in docs/development.md.
MIT. See LICENSE.