Now in Service

Answers Built on Evidence,
Not Guesswork

RAG (Retrieval-Augmented Generation) — it first retrieves the verified sources relevant to your question, then generates an answer only on that evidence.

STEP 1

Understand the Question

Interprets your intent along with farm context such as crop, area, and transplant date.

STEP 2

Search the Knowledge Base

Retrieves relevant evidence from verified consulting materials and cultivation manuals using BGE-M3 embeddings.

STEP 3

Answer with Evidence

Presents the retrieved evidence and a confidence level together — and when confidence is low, it says so.

Verified consulting data BGE-M3 embedding search PLS safe-use standards Expert-reviewed knowledge base
Live AI disease diagnosis screen — automatic detection of tomato brown spot, automatic crop recognition, and differential diagnosis of candidate diseases
Live Screen

AI Disease Diagnosis
in Live Service

  • Automatic image detection — marks lesion locations with bounding boxes and automatically recognizes the crop (tomato)
  • Differential diagnosis — a candidate-disease list presented with similarity scores and the supporting symptoms
  • Certainty indicator — when confidence is low, it says so, and recommends checking photos and samples
  • Reviewed approval — only expert-reviewed data is reflected in the knowledge base
AI farming assistant screen — RAG-based agricultural consultation, questions on VPD, pests, and irrigation by growth stage
Live Screen

An AI Farming Assistant
That Answers with Evidence

A RAG-based consultation that answers from verified sources, not guesswork. It responds based on your farm's growing environment, and flags when confidence is low.

  • Farm context aware — answers grounded in your real farm data, such as crop, area, and transplant date
  • Ready-to-use questions — VPD by growth stage, likely pests and diseases, and today's irrigation and environment tips
  • Clearly advisory — the final decision always rests with the farmer
Try the chatbot demo on the web
How It Works

Three Principles
of Trustworthy AI

We believe that for AI to earn trust in the field, verifiable evidence must come before impressive-sounding answers. FarmPath's RAG system operates on three principles.

  • Evidence first — grounded answers from a RAG engine (BGE-M3 embeddings) built on verified consulting data
  • Safety compliance — automatic integration with PLS (Positive List System) safe-use standards
  • Human-centered — every proposal is executed only after the farmer's approval; the final decision always rests with a person

Try It for Yourself Now

On the FarmStack AI consulting platform (rag.farmstack.kr), sensor monitoring and AI agricultural consultation are running as a live service. We have also prepared a chatbot demo you can try on the web without signing up.

Next : Smart Farm 3.0

The Next Step We Are Researching:
Agentic AI Automated Control

Beyond remote control (1.0) and data-driven automation (2.0) lies Smart Farm 3.0, where AI assesses the situation and plans control actions on its own. FarmPath is researching its core technology — Agentic AI-based automated control.

Smart Farm 1.0

Remote Monitoring & Control

Watching the greenhouse from a smartphone and operating it manually

Smart Farm 2.0 — Today

Data-Driven Automation + RAG AI

Automated integrated environment control plus evidence-based AI farming support — the service FarmPath delivers today

Smart Farm 3.0 — In Research

Agentic AI Automated Control

AI agents collaborating under a shared goal and planning control on their own — R&D in progress

Architecture under research — SENSING → Shared Goal → ACTING
Weather

A sensing role that collects weather and forecast data and shares it with the other agents

Crop Health

A sensing role that detects plant stress and growth stages and shares insights

Irrigation

An acting role that plans irrigation schedules matched to water demand

Nutrient

An acting role that adjusts fertilization plans to the growth stage

Pest Control

An acting role that detects outbreak risk early and proposes control plans

The principle stays the same even in research — every control plan is designed on the premise that it applies only after the farmer's approval.

Live Screens

The AI Diagnoses — and Shows Its Evidence

From differential disease diagnosis to the reasoning and confidence behind each recommendation, shown on real screens. Click to enlarge.

AI disease diagnosis result screen (leaf detection and candidate-disease list)
AI Differential Disease Diagnosis

Leaf object detection + similarity-based disease candidates

AI farming assistant desktop conversation start screen
AI Farming Assistant

Farm context + suggested questions (pests, irrigation, VPD)

AI farming assistant mobile query and comprehensive analysis screen
Mobile Comprehensive Analysis

Overall growing-environment score + improvement tips

AI prescription mobile screen (salt accumulation warning)
AI Prescription

Anomaly detection + recommended actions

AI prescription evidence and confidence screen
Prescription Evidence & Confidence

Best practices and papers + AI confidence