Module runtime 

MlRuntime — convenience bundle of ModelRegistry + MlJobQueue so feature code (classifier, symbolic, semantic cache, …) only needs to hold a single handle.

The runtime is detached from crate::runtime::RedDBRuntime: it can be constructed standalone (in-memory) for tests or bound to a shared MlPersistence backend for durable deployments. A future sprint will add a RedDBRuntime::ml() accessor that returns the bound instance — this module provides the pieces it will wire up.

Structs

MlRuntime

Shared entrypoint used by every ML feature.

MlRuntimeConfig

Compile-time defaults for a standalone runtime. Production callers pass an MlRuntimeConfig with their own worker count.