Module modality_stubs 

Shared boilerplate for modality-only engines (embedding / transcription / TTS).

These engines wrap a per-modality executor and expose its functionality through the InferenceEngine trait. They do NOT do text generation, so infer / infer_stream return Err and the orchestration accessors (status / metrics / health / shutdown) return inert defaults.

Each modality engine called these with copy-pasted bodies (~80 lines of identical boilerplate per engine). This module hosts the canonical versions so the 3 fake engines can shrink.

Functions

inert_health

Health: always healthy if executor loaded.

inert_metrics

Inert engine metrics — all zeros, defaults on nested types.

inert_status

Inert engine status — modality engines aren’t request-driven, so most fields are zeroed. is_ready: true because the executor is loaded.