KrishiMitra AI

Inspiration

I wanted to bridge the gap between advanced AI and Indian farmers. Most agricultural tech tools are English-only and ignore real-time local conditions. I built KrishiMitra to be a multilingual, voice-activated agronomist in the farmer's pocket.

What it does

KrishiMitra AI acts as a 24/7, context-aware agronomist that fits right in a farmer's pocket. It breaks down technical, agricultural science into simple, actionable steps in the user's native language

How we built it

Frontend: A lightweight, mobile-first web app using HTML, CSS, and Vanilla JavaScript. AI Engine: Google's Gemini 2.5 Flash API for multimodal plant disease scanning (Vision) and smart farming chat. Context & Voice: Integrated the Open-Meteo API for live weather data and the native browser Web Speech API for hands-free voice commands in English, Hindi, and Marathi.

Challenges we ran into

My biggest hurdle was getting the AI to return perfect, unbreakable JSON for the frontend to read. Translating complex agricultural terms into Hindi and Marathi often resulted in cut-off text or rogue formatting that crashed the app. I solved this by bumping up the API token limit and implementing Google's strict responseSchema to permanently force the AI into a flawless JSON format

Accomplishments that we're proud of

True Accessibility: Integrating voice-to-text and text-to-speech to break literacy barriers, allowing farmers to interact purely by voice in English, Hindi, and Marathi. Context-Aware Advice: Dynamically injecting live local weather data into the AI prompt so the advice is always hyper-relevant to the farmer's actual environment right now. Lightweight App: Packing advanced multimodal AI and real-time weather tracking into a fast, zero-backend web app that works effortlessly on basic smartphones.

What we learned

Structured AI: How to use JSON schemas to make generative AI outputs predictable, safe, and production-ready. Context is King: Injecting live weather data into the AI prompt makes the generated advice hyper-local and immediately actionable. Accessibility: For rural users, the best user interface isn't always a screen—often, it's just their native voice.

What's next for KrishiMitra AI

More Regional Languages: Expanding the language support to include Telugu, Tamil, Bengali, and Gujarati to reach farmers across all of India. WhatsApp Integration: Building a chatbot version so farmers with low internet bandwidth can simply send a photo of a sick plant on WhatsApp to get instant advice.

Built With

• css3
• geminiapi
• html5
• javascript
• texttospeech
• vanilla
• weatherapi