Flora Vision - AI Plant Diagnostics

One photo. Everything your plant needs. Flora Vision identifies your plant, spots what's wrong, and hands you a clear care plan — watering, light, fixes, and propagation steps included. What started as a one-day prototype built to put OpenAI's vision models to the test is now a production SaaS product, live at floravision.co.za.

What It Does

Snap a photo of any plant — leaves, flowers, even a struggling stem — and in around ten seconds Flora Vision returns six care insights drawn from over 1,000 recognised species:

• Species identification — the plant's name and a short profile of how it likes to live.
• Health diagnosis — common problems spotted in your photo, with suggested fixes.
• Care checklist — light, water and feeding routines tailored to the species.
• Propagation guide — step-by-step stem cutting instructions when the plant supports it.

The advice is deliberately practical, not botanical jargon — a care plan you can act on, not a taxonomy lesson.

From Prototype to Product

The 2023 prototype proved the idea; turning it into a product meant rebuilding almost everything around it. The Angular app was upgraded across six major versions (v15 through v21) and redesigned as a Tailwind-based SaaS experience with a proper landing page, brand identity, and a mobile-first capture flow — Capture on your phone, Upload on desktop.

Signed-in users get a full dashboard: every analysis is saved to a personal history with range filters, thumbnail previews, and a modal that reopens the complete care plan and image for any past entry. Authentication supports email registration, verification and password reset flows, plus one-tap Google sign-in via the Cognito Hosted UI.

Under the Hood

Flora Vision runs on a serverless AWS stack:

• AWS Amplify Gen 2 hosting with Cognito authentication reused via referenceAuth during the Gen 1 to Gen 2 migration.
• API Gateway, Lambda and DynamoDB powering the analysis and dashboard APIs.
• S3 storage for uploads, served through short-lived signed URLs with frontend caching.
• OpenAI vision models doing the plant identification and diagnosis.

Small optimisations carry over from the prototype era: images are still resized in the browser before upload, keeping analysis times low even on mobile connections.

Conclusion

Flora Vision reflects how Wezside likes to build: start with a sharp, one-day proof of concept, learn from real use, then invest in the product engineering — auth, persistence, infrastructure, design — that turns a demo into something people can rely on. Got a plant nearby? Ask it how it's doing.