deeppakhare/farmgpt-ai-foundation

FarmGPT AI is a chat based assistant built to help smallholder farmers in India, where most farms are under two hectares. The README explains that these farmers usually face the same recurring questions each season: what disease is damaging their crop, whether to irrigate today, what the market price is nearby, whether they qualify for government support programs, and how much fertilizer to apply. Existing options like local advisory offices, input dealers, generic weather apps, and government portals each fall short in some way, so this project tries to combine everything into one tool. The core idea is a multi agent system. When a farmer types or asks a question, an Intent Router figures out what kind of question it is and passes it to one of six specialist assistants covering crop disease, weather, market prices, government schemes, fertilizer planning, or general questions. Each specialist uses Google's Gemini 3 Flash model to generate answers, and results are shown as structured cards rather than plain text, such as weather panels or price tables. The app includes a dashboard, a camera based disease scanner that gives a diagnosis and treatment steps after a farmer uploads a leaf photo, a farm planner that produces a season long crop plan, a market intelligence section for selling decisions, and a weather section tailored to farming tasks like spraying or harvesting. On the technical side, the project uses TypeScript, React, and TanStack Start, is deployed on Cloudflare Workers for fast loading even on weaker connections, and stores data in Supabase with row level security so each farmer only sees their own information. Authentication supports Google and email sign in. The interface is built with shadcn and Tailwind and supports both light and dark themes. This is a student or hackathon style capstone project rather than a polished commercial product, built for a Kaggle AI agents competition, and would mainly interest developers curious about multi agent AI systems or anyone exploring agricultural technology. The full README is longer than what was shown.