AgroSage

Inspiration We were inspired by small family farms. They work incredibly hard but often don't have access to the expensive technology that large corporate farms do. We wanted to build a tool that uses the power of AI to help them conserve resources, save money, and grow healthier crops. Our goal was to create a personal "AI advisor" that makes data simple and actionable for any farmer. What it does AgroSage is an AI-powered farming advisor that keeps a 24/7 watch on a farm's health. It pulls in real-time weather data and simulates field sensor data for soil moisture. An AI agent analyzes this information in BigQuery to spot potential problems, like a field becoming too dry before a heatwave. When it finds an issue, another agent uses Google's Gemini AI to write a simple, easy-to-understand alert, like "Your West Field needs watering soon," and shows it on a simple dashboard. How we built it We designed AgroSage as a multi-agent system using Google Cloud. Each core task is handled by its own serverless agent (a Cloud Function). Agents for Weather and Soil Data collect information and store it in Google BigQuery. An Insights Agent runs every few minutes to query BigQuery and look for problems. If a problem is found, it triggers a Recommendation Agent that uses the Vertex AI Gemini API to write the alert. The final results and alerts are displayed on a simple web dashboard built with Streamlit and hosted on Cloud Run. The whole process is automated and deployed using Cloud Build. Challenges we ran into Our biggest challenge was making the different agents work together smoothly. Since they are all separate functions, ensuring one agent correctly triggered the next required careful design. We also spent a lot of time crafting the right prompt for the Gemini AI to make sure its advice was genuinely helpful and not just robotic. Getting the Google Cloud permissions right for all the different services to talk to each other also took some time. Accomplishments that we're proud of We are most proud of building a complete, end-to-end automated system in a short time. Seeing a simulated soil moisture drop in our database, and then watching a clear, helpful alert automatically appear on our dashboard just a minute later was a huge success. We are also proud that we built this with serverless technology, which means it's very efficient and can easily scale to help many farms. What we learned We learned that breaking a complex problem down into small, specialized AI agents is a very powerful approach. It's much easier to build and debug one simple agent at a time than to create one giant, complicated program. We also learned how effective Google's Gemini AI is for the "last mile" problem of converting raw data into natural, human-readable language. What's next for AgroSage The next step is to move beyond simulation and integrate real data sources. We want to connect to Google Earth Engine to analyze actual satellite images of crop health. We also plan to build out an SMS alert system so farmers can get critical warnings directly on their phones, even when they aren't looking at the dashboard. Our ultimate goal is to connect AgroSage to smart irrigation systems to make the entire process of watering a farm fully automated.

Built With

• cloudrun
• google-cloud
• openweathermap
• python
• vertex