Semantic cache for LLM calls on Kotlin Multiplatform, with guards against false cache hits. Targets JVM, iOS, macOS, Linux, Windows, JS and Wasm.