Automated Crop Clustering and Anomaly Detection Workflow

This n8n workflow is designed to set up and manage crop data clusters for anomaly detection using Qdrant vector database and AI models. It modifies dataset clusters, calculates medoids (central points of clusters), and establishes threshold scores for detecting anomalies in crop datasets, facilitating better agricultural monitoring and analysis.

The workflow begins when a manual trigger initiates the process and fetches the total points in a specified collection from Qdrant. It computes cluster distances using a distance matrix API, and processes the results to find medoids — the most representative data points — via Python scripting with SciPy. These medoids are used to mark core crop samples.

Subsequently, the workflow retrieves medoid vectors and payloads, prepares for similarity searches, and calculates thresholds based on the furthest points from the medoids, setting these as anomaly boundaries. It also embeds crop descriptions into vector space for multimodal similarity analysis.

Additional steps involve fetching crop cluster details, counting points, and classifying crops based on descriptive data. Sticky notes are used to document key insights and configurations.

This workflow is particularly useful for farmers, agronomists, and data scientists aiming to detect abnormal crop growth patterns, optimize crop management, and leverage AI-powered analytics to enhance decision-making in agriculture.