Plant ID App

Why I built it

I wanted an AI-assisted tool that connects software to an everyday physical-world task: figuring out what a plant is and how to care for it.

What it does

• Accepts an uploaded plant photo.
• Uses AI vision to identify the plant and shape the result into care guidance.
• Surfaces practical notes such as watering, suitability, safety, and next care steps.

Tech stack

Build / Deployment

App type

Standalone web app

Code location

Separate project repo

Front end

Vite, React, Tailwind

Back end / API

Vercel serverless API route

Engine

OpenAI vision API

Deployment

Vercel

Public URL

https://plants.gavinnesom.com

Notes

Separate from the portfolio site; the public demo includes server-side rate limiting before the OpenAI vision call to reduce the risk of runaway API usage.

Architecture / How it works

• The plant app lives outside this portfolio as a standalone Vite and React application.
• The front end handles upload and report display, while a Vercel serverless API route talks to the OpenAI vision API.
• Uploads are validated by file type and size before processing, and the OpenAI API key is never exposed to the browser.
• The portfolio only describes the project and links out to the public app.

What I built

• A photo-upload flow for plant identification.
• A clean care report surface for identification, watering, suitability, and safety notes.
• A standalone app intended to live at plants.gavinnesom.com.

What I learned

• How to shape AI vision output into a gardener-friendly report.
• Where confidence, caveats, and practical care instructions need to be explicit.
• How much polish matters when an AI result is meant to be used outside the screen.

Current status

The Plant ID App has a public standalone deployment at plants.gavinnesom.com.

Next steps

• Improve the care-report formatting.
• Add stronger uncertainty and safety language where identification confidence is limited.
• Capture real screenshots for the portfolio card and detail page.

Links