FARMVILLA

Inspiration

Inspiration ​The inspiration for Farmvilla comes from the urgent need to address fragmented supply chains, food security issues in the agricultural sector and to give the agricultural world a single place to solve all their problems and needs. We envisioned a platform that goes beyond simple information retrieval, serving as a high-impact "AI for Good" tool that empowers farmers and merchants with proactive, intelligent support.
​What it does ​Farmvilla is an AI-native agricultural management platform that acts as a comprehensive ecosystem for the industry:
​Vision Diagnosis: Farmers can upload photos of crops or livestock for instant health assessments and treatment plans.
​Agri-Agent: An autonomous assistant that uses Function Calling to control the UI, such as moving the map to specific locations or normalizing units (e.g., tons to kg) for database searches.
​Community Health Check: Aggregates unstructured chatter to detect regional patterns and proactively warn users about potential disease outbreaks.
​Market Intelligence: Provides real-time market insights and price trends with verifiable source citations via Google Search Grounding.
​Agent Studio(not in the demo to avoid exploitative actions the UI interface was hidden but already coded into the app): Enables farm owners to "export" their own AI by generating a deployable Node.js/Express backend for standalone bots.
​How we built it ​We engineered Farmvilla to showcase the full range of the Gemini ecosystem:
​Gemini 3.0 Flash Preview: Powers the high-performance vision diagnosis and high-speed reasoning.
​Gemini 2.5 Flash Preview TTS: Provides a human-like voice interface for eyes-free operation in the field.
​Thinking Config (thinkingConfig): Handles complex reasoning for community outbreak analysis.
​Function Calling: Bridges the gap between conversational AI and functional application logic.
Grounding: Retrieving real price data and ​Frontend Stack: Built with React, Tailwind CSS, and Leaflet.js for a professional, responsive experience.
​Challenges we ran into ​One of the primary technical challenges was managing API quotas and preventing "429 Too Many Requests" errors, which we addressed by implementing robust error handling, caching logic, and exponential backoff. Additionally, designing for the reality of poor data connectivity in rural areas required the implementation of an offline queue using localStorage for simulated persistence during field diagnoses.
​Accomplishments that we're proud of ​We are particularly proud of the Offline Analysis diagnosis which would really assist farmers who have bad network and the"Agent Studio," which acts as a creative "meta-application" by allowing users to generate their own standalone backends. We also successfully implemented unit normalization reasoning, where the AI understands and converts diverse measurement units on the fly to ensure accurate database queries.
​What we learned ​This project marked a significant transition from "prompt engineering" to "AI systems engineering". We learned how to combine Vision, Grounding, and TTS into a unified agentic workflow that doesn't just talk, but actually does work by controlling the UI and providing tangible deployment paths.
​What's next for FARMVILLA ​The next step for Farmvilla is to expand the Agent Studio to support more platforms, allowing local agricultural cooperatives to further customize and deploy their own specialized AI agents to support their specific regional needs, improve it's offline mode, making, building a website to link to the app and so much more. THE SKY IS NOT THE LIMIT BUT OUR IMAGINATION.

Built With

• germin
• germini2.5
• germini3
• javascript
• node.js
• react