Module provider_cache 

Provider caching awareness — helps LLM providers cache repeated context.

Many LLM providers (Anthropic, OpenAI, Google) implement prefix caching: if the beginning of a prompt matches a previous request, the provider can skip re-processing those tokens. This module helps lean-ctx structure output to maximize prefix cache hits.

Strategies:

1. Stable prefix ordering: Static context (project structure, types) placed BEFORE dynamic context (current file, recent changes)
2. Hash-based change detection: Only re-emit context sections that changed
3. Cacheable block markers: Mark stable blocks so the LLM host knows they can be cached aggressively

Structs

CacheableSection

A section of context with caching metadata.

ProviderCacheState

Tracks which sections have been sent to the provider.

Enums

SectionPriority

Functions

order_for_caching

Order sections for optimal prefix caching. Stable sections first (system, project structure, types), dynamic sections last (recent changes, current task).

render_with_cache_hints

Render sections with cache boundary markers.