caching-strategy

Caching codebase searches and file reads improves workflow execution time by storing previous results and reusing them on subsequent calls. This eliminates redundant disk I/O and repeated grep or glob scans across the project.

You can cache npm package metadata and web lookups by applying a network caching strategy with defined TTLs. This stores fetched doc and package data locally, preventing duplicate external requests during plan and implement phases.

The best way to avoid recomputing token counts is implementing a computed value cache. By caching token budget calculations across plan, implement, and tasks phases, you prevent repeated processing overhead and reduce execution time.

Mtime-based file invalidation checks the modification time of files like docs and specs against the cached timestamp. If the file changes, the cache automatically invalidates and fetches fresh data to ensure correct results.

Yes, you can monitor cache hit rates and time savings using lightweight monitoring hooks. These hooks measure how often cached data is successfully retrieved versus fetched fresh, allowing you to tune TTLs and invalidation rules effectively.

You should avoid TTL-based network caching when you require real-time data accuracy for web metadata lookups. If the external package documentation changes frequently and stale data impacts correctness, strict TTLs may serve outdated information.