How does the 'cold-start' period work?

It is a warm-up phase where the detector accumulates a minimum number of samples (min_samples) before performing its first training and making predictions.

Which streaming backends are supported by this skill?

The skill supports highly efficient algorithms like ECOD, HBOS, and LODA, which are optimized for low-latency and low-memory performance.

What is the difference between streaming and batch detection?

Streaming detection processes data points individually in milliseconds for immediate alerts, while batch detection analyzes full datasets at once for higher retrospective accuracy.

Can I update models without detection downtime?

Yes, the skill utilizes a 'tick-tock' background retraining strategy to update the model periodically while the current model remains active for detection.

Streaming Anomaly Detection

Name: Streaming Anomaly Detection
Author: llama-farm

byllama-farm

0•

데이터 과학 및 ML

Detects anomalies in real-time data streams using automated model updates and configurable windowing strategies.

The Streaming Anomaly Detection skill empowers Claude to implement and manage continuous monitoring systems within the LlamaFarm ecosystem. It provides specialized guidance for handling high-velocity data streams where immediate alerting is required, offering sophisticated features like cold-start lifecycle management, rolling window configurations, and incremental learning backends. This skill is particularly useful for DevOps monitoring, financial fraud detection, or any application where data arrives continuously and requires instant behavioral analysis without the latency of batch processing.

주요 기능

01Support for efficient backends including ECOD, HBOS, and LODA

02Automatic rolling retraining and background model updates

030 GitHub stars

04Real-time streaming detection with millisecond latency

05Configurable lag features and rolling window parameters

06Comprehensive cold-start and warm-up period management

사용 사례

01Monitoring server health and infrastructure metrics for instant alerting

02Analyzing sensor data from IoT devices to identify operational irregularities

03Detecting fraudulent patterns in real-time transaction streams

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add llama-farm/claude-code-marketplace streaming-anomaly

For use in Claude.ai and ChatGPT

Download Skill