together 1.5.28__tar.gz → 1.5.30__tar.gz

{together-1.5.28 → together-1.5.30}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: together
-Version: 1.5.28
+Version: 1.5.30
 Summary: Python client for Together's Cloud Platform!
 License: Apache-2.0
 License-File: LICENSE
@@ -17,6 +17,7 @@ Classifier: Programming Language :: Python :: 3.13
 Classifier: Programming Language :: Python :: 3.14
 Provides-Extra: pyarrow
 Requires-Dist: aiohttp (>=3.9.3,<4.0.0)
+Requires-Dist: black (>=25.9.0,<26.0.0)
 Requires-Dist: click (>=8.1.7,<9.0.0)
 Requires-Dist: eval-type-backport (>=0.1.3,<0.3.0)
 Requires-Dist: filelock (>=3.13.1,<4.0.0)

{together-1.5.28 → together-1.5.30}/pyproject.toml

@@ -12,7 +12,7 @@ build-backend = "poetry.masonry.api"
 
 [tool.poetry]
 name = "together"
-version = "1.5.28"
+version = "1.5.30"
 authors = ["Together AI <support@together.ai>"]
 description = "Python client for Together's Cloud Platform!"
 readme = "README.md"
@@ -43,6 +43,7 @@ numpy = [
     { version = ">=1.26.0", python = ">=3.12" },
 ]
 pillow = "^11.1.0"
+black = "^25.9.0"
 
 [tool.poetry.extras]
 pyarrow = ["pyarrow"]

{together-1.5.28 → together-1.5.30}/src/together/abstract/api_requestor.py

@@ -619,14 +619,29 @@ class APIRequestor:
     ) -> Tuple[TogetherResponse | Iterator[TogetherResponse], bool]:
         """Returns the response(s) and a bool indicating whether it is a stream."""
         content_type = result.headers.get("Content-Type", "")
+
         if stream and "text/event-stream" in content_type:
+            # SSE format streaming
             return (
                 self._interpret_response_line(
                     line, result.status_code, result.headers, stream=True
                 )
                 for line in parse_stream(result.iter_lines())
             ), True
+        elif stream and content_type in [
+            "audio/wav",
+            "audio/mpeg",
+            "application/octet-stream",
+        ]:
+            # Binary audio streaming - return chunks as binary data
+            def binary_stream_generator() -> Iterator[TogetherResponse]:
+                for chunk in result.iter_content(chunk_size=8192):
+                    if chunk:  # Skip empty chunks
+                        yield TogetherResponse(chunk, dict(result.headers))
+
+            return binary_stream_generator(), True
         else:
+            # Non-streaming response
             if content_type in ["application/octet-stream", "audio/wav", "audio/mpeg"]:
                 content = result.content
             else:
@@ -648,23 +663,49 @@ class APIRequestor:
         | tuple[TogetherResponse, bool]
     ):
         """Returns the response(s) and a bool indicating whether it is a stream."""
-        if stream and "text/event-stream" in result.headers.get("Content-Type", ""):
+        content_type = result.headers.get("Content-Type", "")
+
+        if stream and "text/event-stream" in content_type:
+            # SSE format streaming
             return (
                 self._interpret_response_line(
                     line, result.status, result.headers, stream=True
                 )
                 async for line in parse_stream_async(result.content)
             ), True
+        elif stream and content_type in [
+            "audio/wav",
+            "audio/mpeg",
+            "application/octet-stream",
+        ]:
+            # Binary audio streaming - return chunks as binary data
+            async def binary_stream_generator() -> (
+                AsyncGenerator[TogetherResponse, None]
+            ):
+                async for chunk in result.content.iter_chunked(8192):
+                    if chunk:  # Skip empty chunks
+                        yield TogetherResponse(chunk, dict(result.headers))
+
+            return binary_stream_generator(), True
         else:
+            # Non-streaming response
             try:
-                await result.read()
+                content = await result.read()
             except (aiohttp.ServerTimeoutError, asyncio.TimeoutError) as e:
                 raise error.Timeout("Request timed out") from e
             except aiohttp.ClientError as e:
                 utils.log_warn(e, body=result.content)
+
+            if content_type in ["application/octet-stream", "audio/wav", "audio/mpeg"]:
+                # Binary content - keep as bytes
+                response_content: str | bytes = content
+            else:
+                # Text content - decode to string
+                response_content = content.decode("utf-8")
+
             return (
                 self._interpret_response_line(
-                    (await result.read()).decode("utf-8"),
+                    response_content,
                     result.status,
                     result.headers,
                     stream=False,

{together-1.5.28 → together-1.5.30}/src/together/cli/api/endpoints.py

@@ -132,6 +132,10 @@ def endpoints(ctx: click.Context) -> None:
     type=int,
     help="Number of minutes of inactivity after which the endpoint will be automatically stopped. Set to 0 to disable.",
 )
+@click.option(
+    "--availability-zone",
+    help="Start endpoint in specified availability zone (e.g., us-central-4b)",
+)
 @click.option(
     "--wait",
     is_flag=True,
@@ -152,6 +156,7 @@ def create(
     no_speculative_decoding: bool,
     no_auto_start: bool,
     inactive_timeout: int | None,
+    availability_zone: str | None,
     wait: bool,
 ) -> None:
     """Create a new dedicated inference endpoint."""
@@ -177,6 +182,7 @@ def create(
             disable_speculative_decoding=no_speculative_decoding,
             state="STOPPED" if no_auto_start else "STARTED",
             inactive_timeout=inactive_timeout,
+            availability_zone=availability_zone,
         )
     except InvalidRequestError as e:
         print_api_error(e)
@@ -203,6 +209,8 @@ def create(
         click.echo("  Auto-start: disabled", err=True)
     if inactive_timeout is not None:
         click.echo(f"  Inactive timeout: {inactive_timeout} minutes", err=True)
+    if availability_zone:
+        click.echo(f"  Availability zone: {availability_zone}", err=True)
 
     click.echo(f"Endpoint created successfully, id: {response.id}", err=True)
 
@@ -337,13 +345,30 @@ def delete(client: Together, endpoint_id: str) -> None:
     type=click.Choice(["dedicated", "serverless"]),
     help="Filter by endpoint type",
 )
+@click.option(
+    "--mine",
+    type=click.BOOL,
+    default=None,
+    help="true (only mine), default=all",
+)
+@click.option(
+    "--usage-type",
+    type=click.Choice(["on-demand", "reserved"]),
+    help="Filter by endpoint usage type",
+)
 @click.pass_obj
 @handle_api_errors
 def list(
-    client: Together, json: bool, type: Literal["dedicated", "serverless"] | None
+    client: Together,
+    json: bool,
+    type: Literal["dedicated", "serverless"] | None,
+    usage_type: Literal["on-demand", "reserved"] | None,
+    mine: bool | None,
 ) -> None:
     """List all inference endpoints (includes both dedicated and serverless endpoints)."""
-    endpoints: List[ListEndpoint] = client.endpoints.list(type=type)
+    endpoints: List[ListEndpoint] = client.endpoints.list(
+        type=type, usage_type=usage_type, mine=mine
+    )
 
     if not endpoints:
         click.echo("No dedicated endpoints found", err=True)
@@ -432,3 +457,25 @@ def update(
 
     click.echo("Successfully updated endpoint", err=True)
     click.echo(endpoint_id)
+
+
+@endpoints.command()
+@click.option("--json", is_flag=True, help="Print output in JSON format")
+@click.pass_obj
+@handle_api_errors
+def availability_zones(client: Together, json: bool) -> None:
+    """List all availability zones."""
+    avzones = client.endpoints.list_avzones()
+
+    if not avzones:
+        click.echo("No availability zones found", err=True)
+        return
+
+    if json:
+        import json as json_lib
+
+        click.echo(json_lib.dumps({"avzones": avzones}, indent=2))
+    else:
+        click.echo("Available zones:", err=True)
+        for availability_zone in sorted(avzones):
+            click.echo(f"  {availability_zone}")

{together-1.5.28 → together-1.5.30}/src/together/resources/audio/__init__.py

@@ -3,6 +3,7 @@ from functools import cached_property
 from together.resources.audio.speech import AsyncSpeech, Speech
 from together.resources.audio.transcriptions import AsyncTranscriptions, Transcriptions
 from together.resources.audio.translations import AsyncTranslations, Translations
+from together.resources.audio.voices import AsyncVoices, Voices
 from together.types import (
     TogetherClient,
 )
@@ -24,6 +25,10 @@ class Audio:
     def translations(self) -> Translations:
         return Translations(self._client)
 
+    @cached_property
+    def voices(self) -> Voices:
+        return Voices(self._client)
+
 
 class AsyncAudio:
     def __init__(self, client: TogetherClient) -> None:
@@ -40,3 +45,7 @@ class AsyncAudio:
     @cached_property
     def translations(self) -> AsyncTranslations:
         return AsyncTranslations(self._client)
+
+    @cached_property
+    def voices(self) -> AsyncVoices:
+        return AsyncVoices(self._client)

{together-1.5.28 → together-1.5.30}/src/together/resources/audio/speech.py

@@ -30,7 +30,7 @@ class Speech:
         response_format: str = "wav",
         language: str = "en",
         response_encoding: str = "pcm_f32le",
-        sample_rate: int = 44100,
+        sample_rate: int | None = None,
         stream: bool = False,
         **kwargs: Any,
     ) -> AudioSpeechStreamResponse:
@@ -49,7 +49,7 @@ class Speech:
             response_encoding (str, optional): Audio encoding of response.
                 Defaults to "pcm_f32le".
             sample_rate (int, optional): Sampling rate to use for the output audio.
-                Defaults to 44100.
+                Defaults to None. If not provided, the default sampling rate for the model will be used.
             stream (bool, optional): If true, output is streamed for several characters at a time.
                 Defaults to False.
 
@@ -57,6 +57,12 @@ class Speech:
             Union[bytes, Iterator[AudioSpeechStreamChunk]]: The generated audio as bytes or an iterator over audio stream chunks.
         """
 
+        if sample_rate is None:
+            if "cartesia" in model:
+                sample_rate = 44100
+            else:
+                sample_rate = 24000
+
         requestor = api_requestor.APIRequestor(
             client=self._client,
         )

{together-1.5.28 → together-1.5.30}/src/together/resources/audio/transcriptions.py

@@ -30,6 +30,7 @@ class Transcriptions:
         timestamp_granularities: Optional[
             Union[str, AudioTimestampGranularities]
         ] = None,
+        diarize: bool = False,
         **kwargs: Any,
     ) -> Union[AudioTranscriptionResponse, AudioTranscriptionVerboseResponse]:
         """
@@ -52,7 +53,11 @@ class Transcriptions:
             timestamp_granularities: The timestamp granularities to populate for this
                 transcription. response_format must be set verbose_json to use timestamp
                 granularities. Either or both of these options are supported: word, or segment.
-
+            diarize: Whether to enable speaker diarization. When enabled, you will get the speaker id for each word in the transcription.
+                In the response, in the words array, you will get the speaker id for each word.
+                In addition, we also return the speaker_segments array which contains the speaker id for each speaker segment along with the start and end time of the segment along with all the words in the segment.
+                You can use the speaker_id to group the words by speaker.
+                You can use the speaker_segments to get the start and end time of each speaker segment.
         Returns:
             The transcribed text in the requested format.
         """
@@ -103,6 +108,9 @@ class Transcriptions:
                 else timestamp_granularities
             )
 
+        if diarize:
+            params_data["diarize"] = diarize
+
         # Add any additional kwargs
         # Convert boolean values to lowercase strings for proper form encoding
         for key, value in kwargs.items():
@@ -135,6 +143,7 @@ class Transcriptions:
         if (
             response_format == "verbose_json"
             or response_format == AudioTranscriptionResponseFormat.VERBOSE_JSON
+            or diarize
         ):
             # Create response with model validation that preserves extra fields
             return AudioTranscriptionVerboseResponse.model_validate(response.data)
@@ -158,6 +167,7 @@ class AsyncTranscriptions:
         timestamp_granularities: Optional[
             Union[str, AudioTimestampGranularities]
         ] = None,
+        diarize: bool = False,
         **kwargs: Any,
     ) -> Union[AudioTranscriptionResponse, AudioTranscriptionVerboseResponse]:
         """
@@ -180,7 +190,11 @@ class AsyncTranscriptions:
             timestamp_granularities: The timestamp granularities to populate for this
                 transcription. response_format must be set verbose_json to use timestamp
                 granularities. Either or both of these options are supported: word, or segment.
-
+            diarize: Whether to enable speaker diarization. When enabled, you will get the speaker id for each word in the transcription.
+                In the response, in the words array, you will get the speaker id for each word.
+                In addition, we also return the speaker_segments array which contains the speaker id for each speaker segment along with the start and end time of the segment along with all the words in the segment.
+                You can use the speaker_id to group the words by speaker.
+                You can use the speaker_segments to get the start and end time of each speaker segment.
         Returns:
             The transcribed text in the requested format.
         """
@@ -239,6 +253,9 @@ class AsyncTranscriptions:
                 )
             )
 
+        if diarize:
+            params_data["diarize"] = diarize
+
         # Add any additional kwargs
         # Convert boolean values to lowercase strings for proper form encoding
         for key, value in kwargs.items():
@@ -271,6 +288,7 @@ class AsyncTranscriptions:
         if (
             response_format == "verbose_json"
             or response_format == AudioTranscriptionResponseFormat.VERBOSE_JSON
+            or diarize
         ):
             # Create response with model validation that preserves extra fields
             return AudioTranscriptionVerboseResponse.model_validate(response.data)

together-1.5.30/src/together/resources/audio/voices.py

@@ -0,0 +1,65 @@
+from __future__ import annotations
+
+from together.abstract import api_requestor
+from together.together_response import TogetherResponse
+from together.types import (
+    TogetherClient,
+    TogetherRequest,
+    VoiceListResponse,
+)
+
+
+class Voices:
+    def __init__(self, client: TogetherClient) -> None:
+        self._client = client
+
+    def list(self) -> VoiceListResponse:
+        """
+        Method to return list of available voices on the API
+
+        Returns:
+            VoiceListResponse: Response containing models and their available voices
+        """
+        requestor = api_requestor.APIRequestor(
+            client=self._client,
+        )
+
+        response, _, _ = requestor.request(
+            options=TogetherRequest(
+                method="GET",
+                url="voices",
+            ),
+            stream=False,
+        )
+
+        assert isinstance(response, TogetherResponse)
+
+        return VoiceListResponse(**response.data)
+
+
+class AsyncVoices:
+    def __init__(self, client: TogetherClient) -> None:
+        self._client = client
+
+    async def list(self) -> VoiceListResponse:
+        """
+        Async method to return list of available voices on the API
+
+        Returns:
+            VoiceListResponse: Response containing models and their available voices
+        """
+        requestor = api_requestor.APIRequestor(
+            client=self._client,
+        )
+
+        response, _, _ = await requestor.arequest(
+            options=TogetherRequest(
+                method="GET",
+                url="voices",
+            ),
+            stream=False,
+        )
+
+        assert isinstance(response, TogetherResponse)
+
+        return VoiceListResponse(**response.data)

{together-1.5.28 → together-1.5.30}/src/together/resources/endpoints.py

@@ -13,13 +13,18 @@ class Endpoints:
         self._client = client
 
     def list(
-        self, type: Optional[Literal["dedicated", "serverless"]] = None
+        self,
+        type: Optional[Literal["dedicated", "serverless"]] = None,
+        usage_type: Optional[Literal["on-demand", "reserved"]] = None,
+        mine: Optional[bool] = None,
     ) -> List[ListEndpoint]:
         """
-        List all endpoints, can be filtered by type.
+        List all endpoints, can be filtered by endpoint type and ownership.
 
         Args:
-            type (str, optional): Filter endpoints by type ("dedicated" or "serverless"). Defaults to None.
+            type (str, optional): Filter endpoints by endpoint type ("dedicated" or "serverless"). Defaults to None.
+            usage_type (str, optional): Filter endpoints by usage type ("on-demand" or "reserved"). Defaults to None.
+            mine (bool, optional): If True, return only endpoints owned by the caller. Defaults to None.
 
         Returns:
             List[ListEndpoint]: List of endpoint objects
@@ -28,9 +33,20 @@ class Endpoints:
             client=self._client,
         )
 
-        params = {}
+        params: Dict[
+            str,
+            Union[
+                Literal["dedicated", "serverless"],
+                Literal["on-demand", "reserved"],
+                bool,
+            ],
+        ] = {}
         if type is not None:
             params["type"] = type
+        if usage_type is not None:
+            params["usage_type"] = usage_type
+        if mine is not None:
+            params["mine"] = mine
 
         response, _, _ = requestor.request(
             options=TogetherRequest(
@@ -60,6 +76,7 @@ class Endpoints:
         disable_speculative_decoding: bool = True,
         state: Literal["STARTED", "STOPPED"] = "STARTED",
         inactive_timeout: Optional[int] = None,
+        availability_zone: Optional[str] = None,
     ) -> DedicatedEndpoint:
         """
         Create a new dedicated endpoint.
@@ -74,6 +91,7 @@ class Endpoints:
             disable_speculative_decoding (bool, optional): Whether to disable speculative decoding. Defaults to False.
             state (str, optional): The desired state of the endpoint. Defaults to "STARTED".
             inactive_timeout (int, optional): The number of minutes of inactivity after which the endpoint will be automatically stopped. Set to 0 to disable automatic timeout.
+            availability_zone (str, optional): Start endpoint in specified availability zone (e.g., us-central-4b).
 
         Returns:
             DedicatedEndpoint: Object containing endpoint information
@@ -100,6 +118,9 @@ class Endpoints:
         if inactive_timeout is not None:
             data["inactive_timeout"] = inactive_timeout
 
+        if availability_zone is not None:
+            data["availability_zone"] = availability_zone
+
         response, _, _ = requestor.request(
             options=TogetherRequest(
                 method="POST",
@@ -257,19 +278,49 @@ class Endpoints:
 
         return [HardwareWithStatus(**item) for item in response.data["data"]]
 
+    def list_avzones(self) -> List[str]:
+        """
+        List all available availability zones.
+
+        Returns:
+            List[str]: List of unique availability zones
+        """
+        requestor = api_requestor.APIRequestor(
+            client=self._client,
+        )
+
+        response, _, _ = requestor.request(
+            options=TogetherRequest(
+                method="GET",
+                url="clusters/availability-zones",
+            ),
+            stream=False,
+        )
+
+        assert isinstance(response, TogetherResponse)
+        assert isinstance(response.data, dict)
+        assert isinstance(response.data["avzones"], list)
+
+        return response.data["avzones"]
+
 
 class AsyncEndpoints:
     def __init__(self, client: TogetherClient) -> None:
         self._client = client
 
     async def list(
-        self, type: Optional[Literal["dedicated", "serverless"]] = None
+        self,
+        type: Optional[Literal["dedicated", "serverless"]] = None,
+        usage_type: Optional[Literal["on-demand", "reserved"]] = None,
+        mine: Optional[bool] = None,
     ) -> List[ListEndpoint]:
         """
-        List all endpoints, can be filtered by type.
+        List all endpoints, can be filtered by type and ownership.
 
         Args:
             type (str, optional): Filter endpoints by type ("dedicated" or "serverless"). Defaults to None.
+            usage_type (str, optional): Filter endpoints by usage type ("on-demand" or "reserved"). Defaults to None.
+            mine (bool, optional): If True, return only endpoints owned by the caller. Defaults to None.
 
         Returns:
             List[ListEndpoint]: List of endpoint objects
@@ -278,9 +329,20 @@ class AsyncEndpoints:
             client=self._client,
         )
 
-        params = {}
+        params: Dict[
+            str,
+            Union[
+                Literal["dedicated", "serverless"],
+                Literal["on-demand", "reserved"],
+                bool,
+            ],
+        ] = {}
         if type is not None:
             params["type"] = type
+        if usage_type is not None:
+            params["usage_type"] = usage_type
+        if mine is not None:
+            params["mine"] = mine
 
         response, _, _ = await requestor.arequest(
             options=TogetherRequest(
@@ -308,6 +370,7 @@ class AsyncEndpoints:
         disable_speculative_decoding: bool = True,
         state: Literal["STARTED", "STOPPED"] = "STARTED",
         inactive_timeout: Optional[int] = None,
+        availability_zone: Optional[str] = None,
     ) -> DedicatedEndpoint:
         """
         Create a new dedicated endpoint.
@@ -348,6 +411,9 @@ class AsyncEndpoints:
         if inactive_timeout is not None:
             data["inactive_timeout"] = inactive_timeout
 
+        if availability_zone is not None:
+            data["availability_zone"] = availability_zone
+
         response, _, _ = await requestor.arequest(
             options=TogetherRequest(
                 method="POST",
@@ -506,3 +572,28 @@ class AsyncEndpoints:
         assert isinstance(response.data["data"], list)
 
         return [HardwareWithStatus(**item) for item in response.data["data"]]
+
+    async def list_avzones(self) -> List[str]:
+        """
+        List all availability zones.
+
+        Returns:
+            List[str]: List of unique availability zones
+        """
+        requestor = api_requestor.APIRequestor(
+            client=self._client,
+        )
+
+        response, _, _ = await requestor.arequest(
+            options=TogetherRequest(
+                method="GET",
+                url="clusters/availability-zones",
+            ),
+            stream=False,
+        )
+
+        assert isinstance(response, TogetherResponse)
+        assert isinstance(response.data, dict)
+        assert isinstance(response.data["avzones"], list)
+
+        return response.data["avzones"]

{together-1.5.28 → together-1.5.30}/src/together/resources/images.py

@@ -21,7 +21,6 @@ class Images:
         *,
         prompt: str,
         model: str,
-        steps: int | None = 20,
         seed: int | None = None,
         n: int | None = 1,
         height: int | None = 1024,
@@ -37,8 +36,6 @@ class Images:
 
             model (str, optional): The model to use for image generation.
 
-            steps (int, optional): Number of generation steps. Defaults to 20
-
             seed (int, optional): Seed used for generation. Can be used to reproduce image generations.
                 Defaults to None.
 
@@ -51,7 +48,8 @@ class Images:
             negative_prompt (str, optional): The prompt or prompts not to guide the image generation.
                 Defaults to None
 
-            image_base64: (str, optional): Reference image used for generation. Defaults to None.
+            **kwargs: Additional parameters like steps (int, optional): Number of generation steps,
+                image_base64 (str, optional): Reference image used for generation, etc.
 
         Returns:
             ImageResponse: Object containing image data
@@ -64,7 +62,6 @@ class Images:
         parameter_payload = ImageRequest(
             prompt=prompt,
             model=model,
-            steps=steps,
             seed=seed,
             n=n,
             height=height,
@@ -96,7 +93,6 @@ class AsyncImages:
         *,
         prompt: str,
         model: str,
-        steps: int | None = 20,
         seed: int | None = None,
         n: int | None = 1,
         height: int | None = 1024,
@@ -112,8 +108,6 @@ class AsyncImages:
 
             model (str, optional): The model to use for image generation.
 
-            steps (int, optional): Number of generation steps. Defaults to 20
-
             seed (int, optional): Seed used for generation. Can be used to reproduce image generations.
                 Defaults to None.
 
@@ -126,7 +120,8 @@ class AsyncImages:
             negative_prompt (str, optional): The prompt or prompts not to guide the image generation.
                 Defaults to None
 
-            image_base64: (str, optional): Reference image used for generation. Defaults to None.
+            **kwargs: Additional parameters like steps (int, optional): Number of generation steps,
+                image_base64 (str, optional): Reference image used for generation, etc.
 
         Returns:
             ImageResponse: Object containing image data
@@ -139,7 +134,6 @@ class AsyncImages:
         parameter_payload = ImageRequest(
             prompt=prompt,
             model=model,
-            steps=steps,
             seed=seed,
             n=n,
             height=height,

{together-1.5.28 → together-1.5.30}/src/together/types/__init__.py

@@ -15,6 +15,8 @@ from together.types.audio_speech import (
     AudioTranslationVerboseResponse,
     AudioTranscriptionResponseFormat,
     AudioTimestampGranularities,
+    ModelVoices,
+    VoiceListResponse,
 )
 from together.types.chat_completions import (
     ChatCompletionChunk,
@@ -140,6 +142,8 @@ __all__ = [
     "AudioTranslationVerboseResponse",
     "AudioTranscriptionResponseFormat",
     "AudioTimestampGranularities",
+    "ModelVoices",
+    "VoiceListResponse",
     "DedicatedEndpoint",
     "ListEndpoint",
     "Autoscaling",