💡 Inspiration
Agriculture is the backbone of Bangladesh, yet our marginal farmers are often deprived of modern tech, proper crop advisory, and financial tracking. Initially, we thought of building a B2C marketplace. However, after analyzing the root problems, we realized that before selling, farmers struggle with production, unexpected weather changes, and lack of affordable machinery. This inspired us to pivot and build Krishi-Hub—a hyper-localized, AI-powered Super App designed strictly to empower farmers with shared resources and smart AI assistance in their native Bengali language.
⚙️ What it does
Krishi-Hub is an all-in-one digital ecosystem for farmers featuring: Community & Resource Hub: A P2P platform where farmers can rent out idle tractors, share cold storage, or split logistics costs (e.g., sharing a pickup truck to the market). Smart AI Plant Doctor: Powered by Gemini 1.5 Flash, farmers can simply snap a picture of a diseased crop leaf. The AI instantly analyzes the image and provides the disease name and exact chemical remedies in clear Bengali. Dynamic Early Warning System: The app checks the user's location and local weather, generating a dynamic, AI-crafted agricultural warning (e.g., advising to clear drainage if heavy rain is expected). Upcoming Modules: A Voice-based Financial Ledger and live Market Prices integration.
🛠️ How we built it
We built the MVP with a mobile-first approach, prioritizing accessibility for non-tech-savvy users:
Frontend: Developed using React and Tailwind CSS to ensure a clean, fast, and highly responsive UI.
AI Integration: Integrated the Google Gemini API (gemini-1.5-flash) for Vision capabilities (leaf disease detection) and natural language generation (localized weather warnings).
Data Formatting: Used react-markdown to parse Gemini's markdown responses into a beautiful, readable UI.
Location/Weather: Integrated navigator.geolocation and weather data to feed contextual prompts to the AI.
Challenges we ran into
API Fallbacks & Edge Cases: Dealing with live weather APIs and Geolocation meant handling scenarios where users deny permissions or network fails. We overcame this by building robust mock-data fallbacks so the app never crashes during crucial moments. Markdown Parsing: Gemini returns responses in raw markdown. Rendering this cleanly in React without breaking the UI required careful integration of markdown parsers. Scope Management: Pivoting from a complex two-sided marketplace to a focused SaaS-style app for farmers within the hackathon timeframe was a tough but necessary product management decision.
🏆 Accomplishments that we're proud of
Successfully integrating Gemini Vision to accurately diagnose local crop diseases and output solutions in fluent, accessible Bengali. Designing an intuitive, visually pleasing UI that feels like a native app despite being a web-based MVP. Completing a functional prototype that actually solves real-world agricultural problems rather than just being a theoretical idea.
📖 What we learned
How to effectively prompt engineering large language models (LLMs) to act as domain experts (agronomists). The importance of UI/UX when building tech for the grassroots level (larger fonts, clear icons, native language). MVP scoping—learning to prioritize core features (AI Doctor & Community) over secondary ones (Ledger) to meet tight deadlines.
🚀 What's next for Krishi-Hub
Digital Voice Ledger: Allowing farmers to track daily expenses/income just by speaking into the app, eventually using this data to help them secure micro-loans. Market Price API: Real-time updates of local wholesale markets to prevent farmers from being underpaid. SMS Integration: Sending AI warnings via SMS for offline farmers.
Built With
- api
- gemini
- google-cloud
- javascript
- react
- tailwind-css
