Caching — Overview

The UDIOT caching system provides transparent, multi-layer caching with pluggable backends, tag-based invalidation, and an AI-powered analyzer that suggests optimal TTL values from observed access patterns.

Key Concepts

Transparent caching. The @AICache decorator wraps any async method so callers see no difference between a cache hit and a freshly computed value.

Multiple strategies. CacheStrategy controls which backend stores the entry: 'memory' (in-process LRU), 'redis', 'cdn', 'tiered', or 'auto'. The 'auto' value defers selection to the AICacheAnalyzer.

Three integration layers.

Layer	API surface	Use case
Decorator	`@AICache` + `@CacheInvalidate` + `@CacheKey`	Method-level caching in service classes
Manager	`CacheManager.getOrSet()`	Explicit programmatic caching
Middleware	`cacheMiddleware()`	HTTP response caching in a middleware pipeline

AI-powered TTL selection. When ttl: 'auto' is passed, the manager scales the TTL relative to the factory's compute duration. The standalone AICacheAnalyzer tracks access patterns and can produce optimization reports and pre-warming predictions.

Tag-based invalidation. Every cached entry can carry string tags. Calling invalidate({ tags: ['user'] }) removes every entry tagged 'user' across all registered providers.

Event-driven invalidation. CacheManager.setupEventInvalidation() maps domain event names to keys, tags, or patterns to remove when handleEvent() is called.

How the Layers Fit Together

Domain event / method call
        │
        ▼
  @CacheInvalidate  ──► CacheManager.invalidate()
                                │
  @AICache          ──► CacheManager.getOrSet()
                                │
  cacheMiddleware   ──► CacheManager.get() / set()
                                │
                         ICacheProvider (memory / redis / ...)

Caching — Overview

Key Concepts

How the Layers Fit Together

See Also