AgroVision is a digital farming assistant and AI-driven decision support system that takes the guesswork out of crop health management.
A platform that combines a Multi-Agent System, Agentic Retrieval-Augmented Generation, vector databases, and GIS-based visualisations to support farmers and agricultural officers throughout the crop lifecycle.
Crop losses caused by extreme weather and pests, inefficient use of water and inputs, and the lack of accessible agricultural expertise in rural communities.
Specialised agents handle one concern each. An orchestrator coordinates them, and a judge agent checks the final response before it reaches the user.
Architecture diagram in design
A detailed diagram will replace this placeholder once the system design is finalised.
Identifies crop diseases from images using Visual Question Answering.
Interprets soil context to tailor treatment to the field.
Pulls live climate variables to ground recommendations in current conditions.
Coordinates the agents and runs the reasoning loop end to end.
Final-stage gatekeeper that blocks harmful or unverified output.
The Orchestrator receives a question and decides which specialised agents to engage.
Specialised agents gather evidence via Agentic RAG and live climate variables.
The Judge Agent checks the final response for safety before it reaches the user.
The exact technology choices are not yet finalised. Brand logos and names will appear in each slot below once they are.
Cross-platform mobile experience.
Asynchronous request handling.
Multi-agent coordination.
Reasoning and response generation.
RAG retrieval over verified data.
Structured data and spatial mapping.
Reference image storage.
Hosting and scaling.
Build and deploy automation.
Live metrics and alerting.
LLM responses are restricted to verified agricultural data stored in the vector database.
A final-stage agent checks every response for harmful content before it reaches the user.
Performance and accuracy are quantified against expert-verified answers.
API health and load behaviour are monitored continuously.
Built so users with varying levels of technical literacy can use the system without friction.
Sensitive data and public surfaces are separated by role-based controls.
