/sticky-notes
Semantic Caching
A short working note from the thinking workspace.
Tags
Semantic caching maps natural language queries to vector spaces. If a query falls within the similarity threshold, we bypass the LLM. But loose boundaries lead to cache collisions and serve stale data. Optimize for safety by raising the similarity gate.