soniox 2.5.0__tar.gz → 2.6.0__tar.gz

{soniox-2.5.0 → soniox-2.6.0}/CHANGELOG.md

@@ -73,6 +73,14 @@ Examples:
 
 ---
 
+## [2.6.0] - 15 jun 2026
+
+### Added
+
+- `endpoint_sensitivity` field on `RealtimeSTTConfig`: adjusts how likely the model is to emit a speech endpoint. Allowed values are between -1.0 and 1.0; the default is 0.0. Introduced in the Soniox v5 model.
+
+---
+
 ## [2.5.0] - 12 jun 2026
 
 ### Added

{soniox-2.5.0 → soniox-2.6.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: soniox
-Version: 2.5.0
+Version: 2.6.0
 Summary: The official Python SDK for the Soniox API (STT, REST)
 Project-URL: Homepage, https://soniox.com
 Project-URL: Documentation, https://soniox.com/docs
@@ -111,7 +111,7 @@ from soniox.utils import render_tokens, throttle_audio, start_audio_thread
 DEMO_FILE = "path_to_your_audio_file"
 
 client = SonioxClient()
-config = RealtimeSTTConfig(model="stt-rt-v4", audio_format="mp3")
+config = RealtimeSTTConfig(model="stt-rt-v5", audio_format="mp3")
 final_tokens: list[Token] = []
 non_final_tokens: list[Token] = []
 

{soniox-2.5.0 → soniox-2.6.0}/README.md

@@ -68,7 +68,7 @@ from soniox.utils import render_tokens, throttle_audio, start_audio_thread
 DEMO_FILE = "path_to_your_audio_file"
 
 client = SonioxClient()
-config = RealtimeSTTConfig(model="stt-rt-v4", audio_format="mp3")
+config = RealtimeSTTConfig(model="stt-rt-v5", audio_format="mp3")
 final_tokens: list[Token] = []
 non_final_tokens: list[Token] = []
 

{soniox-2.5.0 → soniox-2.6.0}/docs/types.md

@@ -992,6 +992,7 @@ Configuration for initiating a realtime transcription session.
 | `enable_language_identification` | `bool \| None` | Enable automatic language detection. |
 | `enable_endpoint_detection` | `bool \| None` | Enable endpoint detection for utterance boundaries. |
 | `max_endpoint_delay_ms` | `int \| None` | Maximum delay between the end of speech and returned endpoint. Allowed values for maximum delay are between 500ms and 3000ms. The default value is 2000ms |
+| `endpoint_sensitivity` | `float \| None` | Adjusts how likely the model is to emit an endpoint. Higher values make endpoints more likely (finalizing sooner); lower values make them less likely. Allowed values are between -1.0 and 1.0; the default is 0.0. Introduced in the Soniox v5 model; earlier models reject it. |
 | `translation` | `TranslationConfigInput \| None` | Translation configuration. |
 | `client_reference_id` | `str \| None` | Optional tracking identifier (max 256 chars). |
 

{soniox-2.5.0 → soniox-2.6.0}/examples/async_soniox_client/realtime_example.py

@@ -10,7 +10,7 @@ DEMO_FILE = Path(__file__).resolve().parents[2] / "assets" / "coffee_shop.mp3"
 
 async def main() -> None:
     client = AsyncSonioxClient()
-    config = RealtimeSTTConfig(model="stt-rt-v4", audio_format="mp3")
+    config = RealtimeSTTConfig(model="stt-rt-v5", audio_format="mp3")
     final_tokens: list[Token] = []
     non_final_tokens: list[Token] = []
     async with client.realtime.stt.connect(config=config) as session:

{soniox-2.5.0 → soniox-2.6.0}/examples/soniox_client/realtime_example.py

@@ -9,7 +9,7 @@ DEMO_FILE = Path(__file__).resolve().parents[2] / "assets" / "coffee_shop.mp3"
 
 def main() -> None:
     client = SonioxClient()
-    config = RealtimeSTTConfig(model="stt-rt-v4", audio_format="mp3")
+    config = RealtimeSTTConfig(model="stt-rt-v5", audio_format="mp3")
     final_tokens: list[Token] = []
     non_final_tokens: list[Token] = []
     with client.realtime.stt.connect(config=config) as session:

{soniox-2.5.0 → soniox-2.6.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "soniox"
-version = "2.5.0"
+version = "2.6.0"
 dependencies = ["httpx>0.25.0", "websockets>11.0", "pydantic>2"]
 requires-python = ">=3.10"
 authors = [{ name = "Soniox", email = "support@soniox.com" }]

{soniox-2.5.0 → soniox-2.6.0}/src/soniox/types/realtime.py

@@ -94,6 +94,14 @@ class RealtimeSTTConfig(BaseModel):
     Allowed values for maximum delay are between 500ms and 3000ms. The default value is 2000ms
     """
 
+    endpoint_sensitivity: float | None = Field(default=None, ge=-1.0, le=1.0)
+    """
+    Adjusts how likely the model is to emit an endpoint. Higher values make
+    endpoints more likely (finalizing sooner); lower values make them less
+    likely. Allowed values are between -1.0 and 1.0; the default is 0.0.
+    Introduced in the Soniox v5 model; earlier models reject it.
+    """
+
     translation: TranslationConfigInput | None = None
     """Translation configuration."""
 

{soniox-2.5.0 → soniox-2.6.0}/tests/unit/_sdk_bindings.py

@@ -86,9 +86,9 @@ SDK_BINDINGS: dict[str, SdkBinding] = {
         async_call=lambda c: c.stt.count(),
     ),
     "create_transcription": SdkBinding(
-        sync_call=lambda c: c.stt.create(model="stt-async-v4", audio_url=_AUDIO_URL),
-        async_call=lambda c: c.stt.create(model="stt-async-v4", audio_url=_AUDIO_URL),
-        expect_json={"model": "stt-async-v4", "audio_url": _AUDIO_URL},
+        sync_call=lambda c: c.stt.create(model="stt-async-v5", audio_url=_AUDIO_URL),
+        async_call=lambda c: c.stt.create(model="stt-async-v5", audio_url=_AUDIO_URL),
+        expect_json={"model": "stt-async-v5", "audio_url": _AUDIO_URL},
     ),
     "get_transcription": SdkBinding(
         sync_call=lambda c: c.stt.get(_ID),
@@ -168,19 +168,19 @@ SDK_BINDINGS_FULL: dict[str, SdkBinding] = {
     ),
     "create_transcription": SdkBinding(
         sync_call=lambda c: c.stt.create(
-            model="stt-async-v4",
+            model="stt-async-v5",
             audio_url=_AUDIO_URL,
             client_reference_id="ref-123",
             config=_FULL_CONFIG,
         ),
         async_call=lambda c: c.stt.create(
-            model="stt-async-v4",
+            model="stt-async-v5",
             audio_url=_AUDIO_URL,
             client_reference_id="ref-123",
             config=_FULL_CONFIG,
         ),
         expect_json={
-            "model": "stt-async-v4",
+            "model": "stt-async-v5",
             "audio_url": _AUDIO_URL,
             "client_reference_id": "ref-123",
             "language_hints": ["en", "de"],

{soniox-2.5.0 → soniox-2.6.0}/tests/unit/test_api.py

@@ -208,7 +208,7 @@ def test_error_validation_errors_appear_in_message(client: SonioxClient) -> None
     )
 
     with pytest.raises(SonioxInvalidRequestError) as exc_info:
-        client.stt.create(model="stt-async-v4", audio_url="https://example.com/audio.mp3")
+        client.stt.create(model="stt-async-v5", audio_url="https://example.com/audio.mp3")
 
     message = str(exc_info.value)
     assert "Invalid payload" in message

{soniox-2.5.0 → soniox-2.6.0}/tests/unit/test_coverage_gaps.py

@@ -163,7 +163,7 @@ def test_translation_config_two_way_clears_target_language() -> None:
 def test_create_transcription_payload_rejects_both_sources() -> None:
     with pytest.raises(ValidationError, match="Only one of audio_url or file_id"):
         CreateTranscriptionPayload(
-            model="stt-async-v4",
+            model="stt-async-v5",
             audio_url="https://a/b",
             file_id="f1",
         )
@@ -171,4 +171,4 @@ def test_create_transcription_payload_rejects_both_sources() -> None:
 
 def test_create_transcription_payload_requires_a_source() -> None:
     with pytest.raises(ValidationError, match="Either audio_url or file_id"):
-        CreateTranscriptionPayload(model="stt-async-v4")
+        CreateTranscriptionPayload(model="stt-async-v5")

{soniox-2.5.0 → soniox-2.6.0}/tests/unit/test_types.py

@@ -45,7 +45,7 @@ def test_structured_context_general_keeps_typed_list_unchanged() -> None:
 
 def test_raw_format_requires_sample_rate_and_channels() -> None:
     with pytest.raises(ValueError) as exc:
-        RealtimeSTTConfig(model="stt-rt-v4", audio_format="pcm_s16le")
+        RealtimeSTTConfig(model="stt-rt-v5", audio_format="pcm_s16le")
     msg = str(exc.value)
     assert "sample_rate" in msg
     assert "num_channels" in msg
@@ -53,18 +53,18 @@ def test_raw_format_requires_sample_rate_and_channels() -> None:
 
 def test_raw_format_requires_sample_rate_only() -> None:
     with pytest.raises(ValueError) as exc:
-        RealtimeSTTConfig(model="stt-rt-v4", audio_format="pcm_s16le", num_channels=1)
+        RealtimeSTTConfig(model="stt-rt-v5", audio_format="pcm_s16le", num_channels=1)
     assert "sample_rate" in str(exc.value)
 
 
 def test_raw_format_requires_num_channels_only() -> None:
     with pytest.raises(ValueError) as exc:
-        RealtimeSTTConfig(model="stt-rt-v4", audio_format="pcm_s16le", sample_rate=16000)
+        RealtimeSTTConfig(model="stt-rt-v5", audio_format="pcm_s16le", sample_rate=16000)
     assert "num_channels" in str(exc.value)
 
 
 def test_raw_format_with_both_succeeds() -> None:
-    cfg = RealtimeSTTConfig(model="stt-rt-v4", audio_format="pcm_s16le", sample_rate=16000, num_channels=1)
+    cfg = RealtimeSTTConfig(model="stt-rt-v5", audio_format="pcm_s16le", sample_rate=16000, num_channels=1)
     assert cfg.audio_format == "pcm_s16le"
     assert cfg.sample_rate == 16000
     assert cfg.num_channels == 1

{soniox-2.5.0 → soniox-2.6.0}/uv.lock

@@ -559,7 +559,7 @@ wheels = [
 
 [[package]]
 name = "soniox"
-version = "2.5.0"
+version = "2.6.0"
 source = { editable = "." }
 dependencies = [
     { name = "httpx" },