AgentSkillsCN

redis-cache

为活跃的规则版本与合规性清单增设 Redis 缓存层，以满足性能方面的非功能性需求。

SKILL.md
--- frontmatter
name: redis-cache
description: Adds Redis caching layer for active rule versions and compliance lists to meet performance NFRs.

Redis Cache for Evaluation Performance

Overview

This skill governs introduction of Redis caching for high-read paths in transaction evaluation, especially active rule versions and compliance list lookups.

The goal is to reduce database load and support throughput/latency non-functional requirements while preserving deterministic behavior.

When to Use

Use this skill when:

  • •Adding Redis to local runtime and Docker Compose
  • •Configuring NestJS cache integration for rule engine reads
  • •Caching active rule versions and compliance list membership queries
  • •Designing cache invalidation after writes
  • •Tuning TTL and cache keys for predictable behavior

Key Concepts

  • •Active Rule Cache: Cached snapshot of enabled rule versions used during evaluation
  • •List Lookup Cache: Cached membership/entry resolution for compliance lists
  • •TTL: Time-to-live policy controlling cache freshness
  • •Cache Invalidation: Explicit key eviction after changes to source-of-truth data
  • •Read-Through Strategy: Service reads cache first, then database on miss

Guidelines

  • •Add Redis service to Docker Compose and configure env-based connection settings
  • •Integrate NestJS cache manager with Redis store in infrastructure configuration
  • •Cache active rule versions with approximately 60s TTL unless stricter requirement exists
  • •Cache compliance list lookups with approximately 30s TTL unless stricter requirement exists
  • •Implement explicit invalidation on rule version create/deactivate and list entry create/update/delete
  • •Use deterministic cache keys including organization scope and query dimensions
  • •Add tests for cache hit/miss behavior and invalidation paths

Explicitly Forbidden

  • •Using cache as source of truth for compliance decisions
  • •Returning cross-organization cached data
  • •Caching mutable request-scoped data without stable key strategy
  • •Skipping invalidation on write paths that affect evaluation behavior
  • •Introducing Redis-only logic that breaks evaluation when cache is unavailable

Design Rules

  • •PostgreSQL remains the single source of truth
  • •Cache layer MUST degrade gracefully: on Redis failure, fallback to DB reads
  • •Keep caching concerns in application/infrastructure services, not domain entities
  • •Instrument cache hit/miss counters for operational visibility
  • •Keep TTL values centralized and configurable via environment variables
  • •Avoid over-caching low-value paths that add complexity without measurable benefit

Expected Outcome

After applying this skill:

  • •Active rule and list reads are significantly reduced at database level
  • •Evaluation throughput improves under concurrent load
  • •Cache consistency is maintained through explicit invalidation
  • •Redis caching requirement is implemented without violating domain boundaries