Merge remote-tracking branch 'upstream/dev' into feat/unified-model-connections

surfsense_backend/tests/unit/podcasts/conftest.py

@@ -38,7 +38,10 @@ def make_spec():
         if speakers is None:
             speakers = [
                 SpeakerSpec(
-                    slot=0, name="Host", role=SpeakerRole.HOST, voice_id="kokoro:am_adam"
+                    slot=0,
+                    name="Host",
+                    role=SpeakerRole.HOST,
+                    voice_id="kokoro:am_adam",
                 ),
                 SpeakerSpec(
                     slot=1,

surfsense_backend/tests/unit/podcasts/test_renderer.py

@@ -71,7 +71,9 @@ def _spec(voice_id: str) -> PodcastSpec:
 
 
 async def test_render_rejects_a_turn_for_an_unknown_speaker(tmp_path):
-    renderer = PodcastRenderer(tts=_UnusedTTS(), catalog=_catalog_with("kokoro:am_adam"))
+    renderer = PodcastRenderer(
+        tts=_UnusedTTS(), catalog=_catalog_with("kokoro:am_adam")
+    )
     transcript = Transcript(turns=[TranscriptTurn(speaker=5, text="Who am I?")])
 
     with pytest.raises(RenderError):
@@ -81,7 +83,9 @@ async def test_render_rejects_a_turn_for_an_unknown_speaker(tmp_path):
 
 
 async def test_render_rejects_a_speaker_whose_voice_is_not_in_the_catalog(tmp_path):
-    renderer = PodcastRenderer(tts=_UnusedTTS(), catalog=_catalog_with("kokoro:am_adam"))
+    renderer = PodcastRenderer(
+        tts=_UnusedTTS(), catalog=_catalog_with("kokoro:am_adam")
+    )
     transcript = Transcript(turns=[TranscriptTurn(speaker=0, text="Hello.")])
 
     with pytest.raises(RenderError):

surfsense_backend/tests/unit/podcasts/test_resolution.py

@@ -64,7 +64,9 @@ def test_a_preferred_voice_invalid_for_the_language_is_replaced():
         speaker_count=1,
         preferred=["kokoro:does-not-exist"],
     )
-    assert voices[0].voice_id in {v.voice_id for v in catalog.for_provider(TtsProvider.KOKORO)}
+    assert voices[0].voice_id in {
+        v.voice_id for v in catalog.for_provider(TtsProvider.KOKORO)
+    }
 
 
 def test_resolution_fails_when_no_voice_speaks_the_language():

surfsense_backend/tests/unit/podcasts/test_voice_catalog.py

@@ -52,7 +52,9 @@ def test_for_provider_returns_only_that_providers_voices():
 def test_for_language_matches_on_the_primary_subtag():
     """A request for 'en' should match an 'en-US' voice (region-insensitive)."""
     catalog = VoiceCatalog([_voice("k1", language="en-US")])
-    assert [v.voice_id for v in catalog.for_language(TtsProvider.KOKORO, "en")] == ["k1"]
+    assert [v.voice_id for v in catalog.for_language(TtsProvider.KOKORO, "en")] == [
+        "k1"
+    ]
 
 
 def test_for_language_excludes_other_languages():
@@ -73,6 +75,59 @@ def test_supports_language_reports_availability():
     assert not catalog.supports_language(TtsProvider.KOKORO, "de")
 
 
+def test_offerable_languages_for_a_concrete_roster_are_its_tags_only():
+    """A provider whose voices are language-bound offers exactly those tags."""
+    catalog = VoiceCatalog(
+        [
+            _voice("k1", language="en-US"),
+            _voice("k2", language="fr"),
+            _voice("k3", language="fr"),
+        ]
+    )
+
+    offering = catalog.offerable_languages(TtsProvider.KOKORO)
+
+    assert offering.languages == ["en-US", "fr"]
+    assert offering.allows_custom is False
+
+
+def test_a_wildcard_roster_offers_the_curated_languages_and_custom_entry():
+    """Voices that speak anything can't enumerate languages themselves, so the
+    catalog offers the curated common list and invites free entry."""
+    catalog = VoiceCatalog(
+        [_voice("o1", provider=TtsProvider.OPENAI, language=ANY_LANGUAGE)]
+    )
+
+    offering = catalog.offerable_languages(TtsProvider.OPENAI)
+
+    assert {"en", "fr", "sw", "hi", "zh"} <= set(offering.languages)
+    assert offering.allows_custom is True
+
+
+def test_a_mixed_roster_offers_the_union_of_concrete_and_curated():
+    catalog = VoiceCatalog(
+        [
+            _voice("v1", provider=TtsProvider.VERTEX_AI, language="en-GB"),
+            _voice("v2", provider=TtsProvider.VERTEX_AI, language=ANY_LANGUAGE),
+        ]
+    )
+
+    offering = catalog.offerable_languages(TtsProvider.VERTEX_AI)
+
+    assert "en-GB" in offering.languages
+    assert "fr" in offering.languages
+    assert offering.allows_custom is True
+
+
+def test_a_provider_with_no_voices_offers_nothing():
+    catalog = VoiceCatalog([_voice("k1")])
+
+    offering = catalog.offerable_languages(TtsProvider.OPENAI)
+
+    assert offering.languages == []
+    assert offering.allows_custom is False
+
+
 def test_get_raises_for_an_unknown_voice():
     catalog = VoiceCatalog([_voice("k1")])
     with pytest.raises(KeyError):