Babysitter Stream Processing Windowing Designer

Designs optimal windowing strategies for stream processing

install

source · Clone the upstream repo

git clone https://github.com/a5c-ai/babysitter

Claude Code · Install into ~/.claude/skills/

T=$(mktemp -d) && git clone --depth=1 https://github.com/a5c-ai/babysitter "$T" && mkdir -p ~/.claude/skills && cp -r "$T/library/specializations/data-engineering-analytics/skills/stream-processing-windowing-designer" ~/.claude/skills/a5c-ai-babysitter-stream-processing-windowing-designer && rm -rf "$T"

manifest: library/specializations/data-engineering-analytics/skills/stream-processing-windowing-designer/SKILL.md

Stream Processing Windowing Designer

Overview

Designs optimal windowing strategies for stream processing. This skill provides expertise in window types, watermarks, and trigger strategies for streaming applications.

Capabilities

Window type selection (tumbling, sliding, session, global)
Watermark strategy design
Late data handling
Trigger configuration
Window aggregation optimization
State management recommendations
Exactly-once semantics configuration

Input Schema

{
  "useCase": "string",
  "eventTimeField": "string",
  "latencyRequirements": {
    "maxLatencyMs": "number",
    "allowedLateMs": "number"
  },
  "aggregations": ["object"]
}

Output Schema

{
  "windowConfig": {
    "type": "string",
    "size": "string",
    "slide": "string"
  },
  "watermarkConfig": "object",
  "triggerConfig": "object",
  "lateDataHandling": "object"
}

Target Processes

Streaming Pipeline
Feature Store Setup

Usage Guidelines

Define use case and event time field
Specify latency requirements
List aggregation operations needed
Consider late data arrival patterns

Best Practices

Choose window type based on business requirements
Configure watermarks based on expected lateness
Use appropriate triggers for latency vs completeness tradeoff
Plan state management for long windows
Test with realistic event time distributions