Recognize NVIDIA Canary as a transcription model

app/MindWork AI Studio/Models/Kinds/TranscriptionModelsFamily.cs

@@ -16,7 +16,7 @@ public sealed class TranscriptionModelsFamily : ModelFamily
     public override ModelVendor Vendor => ModelVendor.UNKNOWN;
 
     /// <inheritdoc />
-    public override ModelSource Source => new("https://huggingface.co/models?pipeline_tag=automatic-speech-recognition", new DateOnly(2026, 9, 12), "Ported from the transcription markers of Provider/ModelKindExtensions.cs, minus the two which their own families now state.");
+    public override ModelSource Source => new("https://huggingface.co/models?pipeline_tag=automatic-speech-recognition", new DateOnly(2026, 9, 19), "Ported from the transcription markers of Provider/ModelKindExtensions.cs, minus the two which their own families now state. Canary came later: the markers never named it, so it reached the answer meant for everything nobody wrote a rule for.");
 
     /// <inheritdoc />
     protected override void Declare(ModelFamilyBuilder builder)
@@ -27,5 +27,11 @@ public sealed class TranscriptionModelsFamily : ModelFamily
         builder.Modifier("wav2vec").AsSubstring().Inherits();
 
         builder.Modifier("parakeet").AsSubstring().Inherits();
+
+        // NVIDIA's other line of speech models, written plain: canary-1b, canary-1b-flash,
+        // canary-180m-flash. A segment rather than a substring, because canary is an ordinary
+        // English word which would otherwise reach into names it has nothing to do with -- the
+        // same reason the embedding family gives for gte.
+        builder.Modifier("canary").AsSegment().Inherits();
     }
 }