harvest-ai

This is a web application built for farmers, agricultural cooperatives, and agribusinesses. It combines farm and crop record-keeping with two AI features: a chat assistant for agronomic questions and a photo-based plant disease scanner. The project appears focused on the East African context, given that the chatbot supports Kinyarwanda (a language spoken in Rwanda) alongside English, and the roadmap mentions mobile money integrations like MTN and Airtel.

The main features are: managing records for multiple farms and crop cycles, a streaming chat assistant that answers farming questions, a disease scanner where you upload a photo of a plant leaf and receive a diagnosis and suggested treatment, and a dashboard with yield trend charts and market price data. Weather and market data views are present in the interface but listed as needing external API credentials to fully activate.

The underlying AI for both the chat and the image scanner runs through a service called the Lovable AI Gateway, which uses Google's Gemini 2.5 Flash model for chat and Gemini Vision for the image analysis. Authentication supports email and password login as well as Google sign-in. The database is a managed PostgreSQL instance, and every database table is protected so that users can only see and modify their own records.

The tech stack is TypeScript throughout. The frontend uses React 19 with a file-based routing setup (TanStack Start), Tailwind CSS for styling, and component primitives from shadcn/ui. The backend runs server functions in the same process rather than as a separate service. Deployment goes to Cloudflare Workers via the Lovable platform with a single publish step.

The roadmap lists several integrations that are stubbed out but not yet wired to real services, including live weather data, mobile money payments, USSD and SMS support, and voice input and output. The project is licensed under MIT.