AgriVision is an intelligent agricultural dashboard designed to help farmers and agronomists make data-driven decisions. By combining real-time weather forecasting with AI-powered agricultural recommendations, it provides actionable insights tailored to specific locations and crop conditions.
Note
This project is currently in active development. Features and interfaces might change as we iterate based on user feedback.
We use Open-Meteo to pull precise, real-time weather data and forecasts. Users can search for specific cities, use their device GPS, or drop a pin on an interactive map to get exact climate conditions for their land.
Instead of just showing temperatures, AgriVision acts as a virtual agronomist. We use Google Genkit and Gemini AI to analyze the current weather data and provide specific farming recommendations.
To ensure the AI advice is accurate and grounded in real agricultural science, we use Pinecone as our vector database. When the AI generates recommendations, it first searches through our embedded agricultural knowledge base in Pinecone. This means the chat and insights are powered by real farming data, not just generic AI responses.
Users can chat with the AI assistant about specific crop concerns. We implemented a seamless history system:
- For guest users, chats are saved locally in the browser.
- For logged-in users, the chat history is automatically synchronized and securely saved to our database, allowing them to pick up right where they left off across different devices.
User accounts and sessions are handled through Better Auth with Google OAuth integration, ensuring a smooth login experience. All user data, including saved locations and chat histories, is safely stored in Supabase (PostgreSQL).
Important
If you are setting this up locally, make sure your OAuth redirect URIs in Google Cloud exactly match your local and production environment URLs.
- Data Collection: The user selects a location. The application fetches current weather and weekly forecasts from Open-Meteo.
- Context Retrieval: The system takes this weather context and queries Pinecone using embeddings to find relevant agricultural best practices.
- AI Generation: Google Gemini processes the weather data alongside the retrieved Pinecone knowledge to generate custom farming advice.
- Delivery: The insights are displayed on the dashboard, and the user can interact further via the persistent chat interface.
- Frontend: Next.js, React, Tailwind CSS, shadcn/ui
- Authentication: Better Auth (with Google OAuth)
- Database: Supabase (PostgreSQL)
- AI Engine: Google Genkit, Gemini Models, Pinecone Vector DB
- External Data: Open-Meteo API