English繁體中文
Back to projects
Contents
Use case Interactive demo Workflow
Generative AI Application Planning · Interactive Prototype

Farm Environment Monitoring & Alerts

Explore how farm sensor time series can become actionable signals. This is an interactive prototype with simulated data and rule-based assessments. It illustrates crop-specific alerts, predefined response advice, and the location of affected plots.

AI application conceptAgent workflow conceptKnowledge base conceptSpatial location (GIS)

Context

A farm has temperature, humidity and soil-moisture sensors. Large volumes of readings across dashboards require continuous manual interpretation.

Problem

Continuous manual monitoring is difficult. Readings do not always translate into immediate action, so anomalies may be noticed late.

Proposed workflow

This prototype compares simulated readings with crop-specific thresholds, selects predefined advice and displays a simulated LINE notification. A future agent could use a knowledge base for this workflow.

Spatial location

Locate alerts on individual plots to help prioritize responses using a GIS view.

Live Prototype

Select a plot or simulate one day of alerts

Click “Simulate one day” to advance from 06:00 to 02:00 the next day. When simulated readings exceed crop thresholds, plot colors and alert symbols update and a notification appears in the LINE-style mock interface.

06:00
Normal Warning (dry / flooded) Alert (cold / heat) * Simulated data, not real sensor readings
Select a plot on the map
to inspect simulated readings and rule-based assessments.
‹ 🌱 Farm Monitor Mockup
No alerts yet.
Click “Simulate one day” to start.
Message mockup➤
Under the Hood

Proposed Architecture

The proposed service follows a sense → assess → advise → notify loop. The diagram describes a possible agent and retrieval architecture; this browser demo uses local rules and predefined text, with no live LLM or RAG service.

IoT sensors Temp · humidity · moisture Time series AI Agent Rules · orchestration Knowledge base concept Retrieval LINE alert concept Locate affected plots (map markers) Farmers / managers
Conceptual architecture: sensor data → assessment with a proposed knowledge base → notifications and plot locations.

Workflow

1

Read sensor data

Read each plot’s simulated temperature, humidity and moisture time series.

2

Knowledge lookup

Illustrate crop-specific reference ranges. A future retrieval service could also use growth stages.

3

Assess anomalies

Apply local thresholds to identify potential stress, such as cold conditions at night.

4

Select response advice

Display predefined example actions and the threshold used for the assessment.

5

Notify and locate

Show a simulated LINE-style alert and mark the affected plot on the map.

Conceptual Knowledge Base Contents

Crop and growth-stage environmental ranges Pest / disease and climate reference tables Ministry of Agriculture / research station cultivation SOPs Cold, heat and heavy-rain response guides Past anomaly cases and response records
About this prototype: This page uses simulated scenario data to present a generative AI application concept: use-case design, an AI-agent workflow, a RAG knowledge-base architecture, and the spatial identification of abnormal risks Spatial location. Locating risks in specific field plots builds on GIS expertise: farm environmental monitoring is inherently spatial, and this is where I see my contribution to smart-agriculture planning.

Interactive prototype with simulated data and rule-based assessments. Readings, plot locations and advice are illustrative. No real farm data, live AI generation, knowledge retrieval or LINE delivery is connected. Basemap © OpenStreetMap, CARTO.