PlantWise

Inspiration

We live in cities. Most of us have a windows, a kitchen counter, maybe a small balcony and somewhere in the back of our minds, a quiet wish to grow something. Mint for a morning drink. A chilli plant that actually came from our own home.

But we never start. Not because we don't want to because we don't know where to. Which plant suits this light? How many can fit here? Do I need soil or can I try hydroponics? What do I do in week one?

We wanted to build something that removes every single one of those questions in one go. Not a generic "plants for beginners" blog post a real, personalised answer based on your actual space, your city's climate, and what you already have at home today.

The inspiration was simple, what if a senior farmer could look at your kitchen photo and just tell you exactly what to do?

What it does

PlantWise is an AI-powered plant recommendation advisor that turns a single photo of your space into a complete, actionable grow plan.

Here's how it works: You upload a photo of your kitchen, balcony, living room or any corner of your home. The AI reads the image for light direction and intensity, available surfaces, humidity signals, ventilation, decor, and estimates how many plants the space can hold.

You answer five quick questions, what kind of plants you want (edibles, flowers, leafy, or mixed), how much time you have per week, your experience level, any constraints like pets or children, and your city.

You tell it what you already have at home. old plastic bottles, coconut coir, coffee grounds, a fish tank, seeds you bought but never planted.

Then in one AI call, Space Grow gives you:

• A space analysis like light, humidity, temperature, usable area, max plants
• The right growing method for your space
• Personalised plant cards
• A materials shopping list (home first → buy locally → buy online)
• A 4-week grow plan No database. No backend. No sensors. Just a photo and a conversation.

How we built it

We built PlantWise entirely through MeDo using natural language prompts. no manual coding.

We described the full app concept to MeDo. the four-step user flow, the quiz design with multi-select pill buttons, the AI vision prompt structure, the results layout with space analysis card, growing method banner, plant recommendation grid, materials list, and timeline. MeDo generated the full working application.

The core intelligence is a single structured prompt sent to Gemini 2.5 Flash with the uploaded image attached. The prompt instructs the model to behave as a senior farmer reading the photo for environmental signals, cross-referencing the user's profile and home materials, and returning a complete JSON response covering space analysis, growing method, plant recommendations, shopping list, and grow plan all in one call for better plant specificity, improve the results card layout, and make everything mobile responsive. MeDo deployed the final app to a public URL with one click.

Challenges we ran into

• Getting the AI to be specific, not generic: Early versions returned bland recommendations like "pothos" and "snake plant" for every space. We refined the prompt to force the model to cross-reference the user's city climate, cooking habits, available home materials, and the actual surfaces visible in the photo before recommending.
• Image-based environmental inference : . A photo doesn't come with a thermometer or humidity sensor. We had to design the prompt so the AI reasons from visual clues
• No token burn on publish: Since the app is fully browser-based with no backend, the API key had to be user-provided. We designed the onboarding so users enter their own free Gemini API key on first load. it stays in their browser, never touches a server, and means the app can be shared publicly without burning anyone's tokens.

Accomplishments that we're proud of

We are proud of the "start from home" layer. Every plant card tells you what you can use right now from what you already have your old bottles, your coconut coir, your coffee grounds. This makes the app immediately actionable. You don't need to buy anything to start.

We are proud that the grow plan reads like a person wrote it. Week one: "Wash your old plastic bottle, fill the bottom third with pebbles for drainage, add your coconut coir mix, and push two curry leaf seeds an inch deep. Water until the coir is damp but not soggy. Place it on the sill and leave it. Don't touch it."

What we learned

We learned that people don't need more gardening content they need one specific, personalised answer for their specific space. Generic advice is everywhere. Contextual, actionable guidance is rare. That's the gap plantwise fills.

What's next for space grow

Follow-along mode. After week one, the user comes back and says "my basil is yellowing" or "the curry leaf isn't sprouting." The AI diagnoses the problem from a new photo and adjusts the plan. A true ongoing advisor, not a one-time recommendation. Community grows. Users in the same city growing the same plant share tips, progress photos, and harvest results. A hyperlocal growing community built around real spaces.

Built With

• base64
• gemini
• json
• medo
• txt
• typescript