Inspiration
Farmer are some of the most importants member of our society but are often forgotten when it comes to technical advance. So we decided to make something to help them .
What it does
It continuously observes plants using environmental sensors and a camera, evaluates their health in real time, and automatically decides when and how much to water them. The system combines plant telemetry, visual cues, and environmental context to assess plant stress and generate actionable recommendations. A live dashboard allows farmers to visualize plant health, camera snapshots, and system decisions, while all control logic runs automatically in the background.
How we built it
-Frontend (React) A modern, minimal dashboard designed for agricultural use. It displays live camera snapshots, sensor readings, plant state, weather context, and system actions. The frontend sends the user’s location to the backend but does not control the system. -Backend (FastAPI) Acts as the central bridge. It ingests sensor data and camera snapshots, fetches weather data based on location, stores the latest plant state, and exposes clean APIs for the frontend. It also forwards final decisions to the actuation layer. -Solace Agent Mesh (SAM) The intelligence layer. Specialized agents evaluate plant telemetry, environmental conditions, and context, while an orchestrator coordinates reasoning and produces a final plant state and recommendation.
Challenges we ran into
hosting a web server on QNX
What we learned
Agent-based architectures make complex decision systems easier to reason about and extend Separating observation, reasoning, and actuation greatly improves reliability
What's next for Untitled
Expand to multi-plant and field-level monitoring Add crop-specific models and seasonal adaptation Incorporate long-term trends and anomaly detection Enable alerts and recommendations across entire farms, not just individual plants
Log in or sign up for Devpost to join the conversation.