evals: state the transcriber constraints as constraints

Both notes explain a choice the code can't show — why Moonshine is the wrong
transcriber for a non-English bot, and why a runtime language switch reports
instead of raising. Phrase each as the constraint itself.

Author: markbackman

src/pipecat/evals/services.py

@@ -198,10 +198,10 @@ def moonshine_service(config: dict) -> STTService:
             - ``language``: Optional language code (e.g. ``es``) or ``Language``.
               When omitted, Moonshine keeps its own default (English).
 
-    Prefer :func:`whisper_service` for a non-English bot. Moonshine's non-English
-    models handle synthesized speech unevenly — returning an empty transcript on
-    audio Whisper reads correctly, and dropping the tail of a Mandarin utterance —
-    and an empty transcript reads as a bot that said nothing.
+    Prefer :func:`whisper_service` for a non-English bot: Moonshine's non-English
+    models transcribe synthesized speech unreliably, returning an empty transcript
+    or dropping the tail of an utterance, and an empty transcript is
+    indistinguishable from a bot that said nothing.
     """
     from pipecat.services.moonshine.stt import Model, MoonshineSTTService
 

src/pipecat/services/whisper/stt.py

@@ -365,8 +365,8 @@ async def _update_settings(self, delta: STTSettings) -> dict[str, Any]:
         """Apply a settings delta, reporting a language the model can't handle.
 
         A mid-call switch to an unsupported language is reported rather than
-        raised: the model stays loaded and keeps transcribing as English, which
-        beats tearing down a live pipeline.
+        raised, so a settings change can't tear down a live pipeline; the model
+        stays loaded and keeps transcribing as English.
 
         Args:
             delta: An STT settings delta.