livekit-plugins-volcenginee 1.3.4__tar.gz → 1.3.6__tar.gz

{livekit_plugins_volcenginee-1.3.4 → livekit_plugins_volcenginee-1.3.6}/.gitignore

@@ -4,3 +4,5 @@
 __pycache__/
 .env
 .env.*
+dist/
+latency_agent.py

{livekit_plugins_volcenginee-1.3.4 → livekit_plugins_volcenginee-1.3.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: livekit-plugins-volcenginee
-Version: 1.3.4
+Version: 1.3.6
 Summary: LiveKit Agent Plugins for Volcengine
 Author-email: linsanzhu <2287225730@qq.com>
 Keywords: audio,livekit,realtime,video,webrtc

{livekit_plugins_volcenginee-1.3.4 → livekit_plugins_volcenginee-1.3.6}/livekit/plugins/volcengine/llm.py

@@ -21,7 +21,7 @@ from livekit.agents.types import (
 from livekit.agents.utils import is_given
 
 from .log import logger
-from .utils import to_chat_ctx, to_fnc_ctx
+from .utils import to_fnc_ctx
 
 
 @dataclass
@@ -206,7 +206,7 @@ class LLMStream(llm.LLMStream):
             stream: openai.AsyncStream[
                 ChatCompletionChunk
             ] = await self._client.chat.completions.create(
-                messages=to_chat_ctx(self._chat_ctx, id(self._llm)),
+                messages=self._chat_ctx.to_provider_format(format="openai")[0],
                 tools=to_fnc_ctx(self._tools) if self._tools else openai.NOT_GIVEN,
                 model=self._model,
                 stream_options={"include_usage": True},

{livekit_plugins_volcenginee-1.3.4 → livekit_plugins_volcenginee-1.3.6}/livekit/plugins/volcengine/realtime.py

@@ -259,7 +259,7 @@ class RealtimeModel(llm.RealtimeModel):
         speaking_style: str = "你的说话风格简洁明了，语速适中，语调自然。",
         speaker: str = "zh_female_vv_jupiter_bigtts",
         opening: str | None = None,
-        api_key: str | None = None,
+        api_key: NotGivenOr[str] = NOT_GIVEN,
         system_role: str | None = None,
         character_manifest: str = None,
         model: Literal["O", "SC"] = "O",
@@ -295,9 +295,12 @@ class RealtimeModel(llm.RealtimeModel):
         logger.info(
             f"Volc Websearch No Result Message: {volc_websearch_no_result_message}"
         )
-        api_key = api_key or os.environ.get("VOLCENGINE_REALTIME_API_KEY")
+        api_key = api_key if utils.is_given(api_key) else os.environ.get("VOLCENGINE_REALTIME_API_KEY")
         if api_key is None:
-            raise ValueError("VOLCENGINE_REALTIME_API_KEY is required")
+            raise ValueError(
+                "api_key is required (pass it explicitly or set the "
+                "VOLCENGINE_REALTIME_API_KEY environment variable)"
+            )
         self._opts = _RealtimeOptions(
             api_key=api_key,
             bot_name=bot_name,

{livekit_plugins_volcenginee-1.3.4 → livekit_plugins_volcenginee-1.3.6}/livekit/plugins/volcengine/stt.py

@@ -22,7 +22,7 @@ from livekit.agents.types import (
     NOT_GIVEN,
     NotGivenOr,
 )
-from livekit.agents.utils import AudioBuffer
+from livekit.agents.utils import AudioBuffer, is_given
 
 from .log import logger
 
@@ -180,7 +180,7 @@ class STT(stt.STT):
     def __init__(
         self,
         *,
-        api_key: str | None = None,
+        api_key: NotGivenOr[str] = NOT_GIVEN,
         base_url: str = "wss://openspeech.bytedance.com/api/v3/sauc/bigmodel",
         resource_id: str | None = None,
         model_name: str = "bigmodel",
@@ -199,6 +199,13 @@ class STT(stt.STT):
             )
         )
 
+        api_key = api_key if is_given(api_key) else os.getenv("VOLCENGINE_STT_API_KEY")
+        if api_key is None:
+            raise ValueError(
+                "api_key is required (pass it explicitly or set the "
+                "VOLCENGINE_STT_API_KEY environment variable)"
+            )
+
         self._opts = STTOptions(
             base_url=base_url,
             api_key=api_key,

{livekit_plugins_volcenginee-1.3.4 → livekit_plugins_volcenginee-1.3.6}/livekit/plugins/volcengine/tts.py

@@ -19,7 +19,12 @@ from livekit.agents import (
     tts,
     utils,
 )
-from livekit.agents.types import DEFAULT_API_CONNECT_OPTIONS
+from livekit.agents.types import (
+    DEFAULT_API_CONNECT_OPTIONS,
+    NOT_GIVEN,
+    NotGivenOr,
+)
+from livekit.agents.utils import is_given
 
 from .log import logger
 
@@ -104,28 +109,44 @@ class _TTSOptions(BaseModel):
 class TTS(tts.TTS):
     def __init__(
         self,
-        api_key: str,
+        *,
+        api_key: NotGivenOr[str] = NOT_GIVEN,
         resource_id: str | None = None,
         voice: str = "zh_female_xiaohe_uranus_bigtts",
         speed: float = 1.0,
         volume: float = 1.0,
         pitch: float = 1.0,
+        min_sentence_length: int = 10,
         sample_rate: Literal[24000, 16000, 8000] = 16000,
         http_session: aiohttp.ClientSession | None = None,
     ):
         """VolcEngine TTS
 
         Args:
-            api_key (str): the API key of the tts, you can get it from the new console.
+            api_key (NotGivenOr[str], optional): the API key of the tts, you can get it from the new console.
+                Falls back to the ``VOLCENGINE_TTS_API_KEY`` environment variable when not provided.
             resource_id (str | None, optional): VolcEngine TTS resource id. Use `seed-tts-2.0` for Doubao TTS 2.0 voices, `seed-tts-1.0` / `seed-tts-1.0-concurr` for Doubao TTS 1.0 voices, `seed-icl-2.0` for voice cloning 2.0, and `seed-icl-1.0` / `seed-icl-1.0-concurr` for voice cloning 1.0. Defaults to None.
             voice (str, optional): the voice id used by the tts request. Defaults to `zh_female_xiaohe_uranus_bigtts`.
             speed (float, optional): speech rate multiplier (OpenAI-style). Range [0.2, 3.0]; 1.0 = normal, 2.0 = 2x, 0.5 = 0.5x. Internally converted to the Volcengine ``speech_rate`` int. Defaults to 1.0.
             volume (float, optional): volume multiplier (OpenAI-style). Range (0.1, 3.0]; 1.0 = normal, 2.0 = 2x, 0.5 = 0.5x. Internally converted to the Volcengine ``loudness_rate`` int. Not supported for mix voices. Defaults to 1.0.
             pitch (float, optional): kept for backwards compatibility, not currently sent to the API. Defaults to 1.0.
+            min_sentence_length (int, optional): minimum number of characters
+                (Chinese chars / tokens) the internal ``TextStreamSentencizer``
+                must accumulate before yielding a sentence to TTS. Lower
+                values trade off TTS efficiency (more / shorter requests)
+                against latency — useful for short-reply agents where the
+                model often returns <10 chars. Defaults to 10 (the upstream
+                default; matches the legacy behavior).
             sample_rate (Literal[24000, 16000, 8000], optional): the sample rate of the tts. Defaults to 16000.
             streaming (bool, optional): whether to use the streaming api. Defaults to True.
             http_session (aiohttp.ClientSession | None, optional): the http session to use. Defaults to None.
         """
+        api_key = api_key if is_given(api_key) else os.getenv("VOLCENGINE_TTS_API_KEY")
+        if api_key is None:
+            raise ValueError(
+                "api_key is required (pass it explicitly or set the "
+                "VOLCENGINE_TTS_API_KEY environment variable)"
+            )
         super().__init__(
             capabilities=tts.TTSCapabilities(streaming=True),
             sample_rate=sample_rate,
@@ -141,6 +162,7 @@ class TTS(tts.TTS):
             pitch=pitch,
         )
         self._session = http_session
+        self._min_sentence_length = min_sentence_length
         self._streams = weakref.WeakSet[SynthesizeStream]()
 
     def _ensure_session(self) -> aiohttp.ClientSession:
@@ -159,6 +181,7 @@ class TTS(tts.TTS):
             conn_options=conn_options,
             opts=self._opts,
             session=self._ensure_session(),
+            min_sentence_length=self._min_sentence_length,
         )
         self._streams.add(stream)
         return stream
@@ -178,15 +201,19 @@ class SynthesizeStream(tts.SynthesizeStream):
         session: aiohttp.ClientSession,
         tts: TTS,
         conn_options=None,
+        min_sentence_length: int = 10,
     ):
         super().__init__(tts=tts, conn_options=conn_options)
         self._opts: _TTSOptions = opts
         self._session = session
+        self._min_sentence_length = min_sentence_length
 
     async def _run(self, emitter: tts.AudioEmitter):
         request_id = utils.shortuuid()
 
-        sentence_splitter = TextStreamSentencizer()
+        sentence_splitter = TextStreamSentencizer(
+            min_sentence_length=self._min_sentence_length
+        )
         emitter.initialize(
             request_id=request_id,
             sample_rate=self._opts.sample_rate,

livekit_plugins_volcenginee-1.3.6/livekit/plugins/volcengine/utils.py

@@ -0,0 +1,25 @@
+from __future__ import annotations
+
+import os
+from collections.abc import Sequence
+from typing import Awaitable, Callable, Union, cast
+
+from livekit.agents import llm
+from openai.types.chat import ChatCompletionToolParam
+
+AsyncAzureADTokenProvider = Callable[[], Union[str, Awaitable[str]]]
+
+
+def get_base_url(base_url: str | None) -> str:
+    if not base_url:
+        base_url = os.getenv(
+            "VOLCENGINE_LLM_BASE_URL", "https://ark.cn-beijing.volces.com/api/v3"
+        )
+    return base_url
+
+
+def to_fnc_ctx(fnc_ctx: Sequence[llm.Tool]) -> list[ChatCompletionToolParam]:
+    return cast(
+        list[ChatCompletionToolParam],
+        llm.ToolContext(fnc_ctx).parse_function_tools("openai"),
+    )

livekit_plugins_volcenginee-1.3.6/livekit/plugins/volcengine/version.py

@@ -0,0 +1 @@
+__version__ = "1.3.6"

livekit_plugins_volcenginee-1.3.4/livekit/plugins/volcengine/utils.py

@@ -1,154 +0,0 @@
-from __future__ import annotations
-
-import base64
-import os
-from collections import OrderedDict
-from collections.abc import Awaitable, Sequence
-from typing import Any, Callable, Union, cast
-
-from openai.types.chat import (
-    ChatCompletionContentPartParam,
-    ChatCompletionMessageParam,
-    ChatCompletionToolParam,
-)
-
-from livekit.agents import llm
-
-AsyncAzureADTokenProvider = Callable[[], Union[str, Awaitable[str]]]
-
-
-def get_base_url(base_url: str | None) -> str:
-    if not base_url:
-        base_url = os.getenv(
-            "VOLCENGINE_LLM_BASE_URL", "https://ark.cn-beijing.volces.com/api/v3"
-        )
-    return base_url
-
-
-def to_fnc_ctx(fnc_ctx: Sequence[llm.Tool]) -> list[ChatCompletionToolParam]:
-    return cast(
-        list[ChatCompletionToolParam],
-        llm.ToolContext(fnc_ctx).parse_function_tools("openai"),
-    )
-
-
-def to_chat_ctx(
-    chat_ctx: llm.ChatContext, cache_key: Any
-) -> list[ChatCompletionMessageParam]:
-    # group the message and function_calls
-    item_groups: dict[str, list[llm.ChatItem]] = OrderedDict()
-    for item in chat_ctx.items:
-        if (
-            item.type == "message" and item.role == "assistant"
-        ) or item.type == "function_call":
-            group_id = item.id.split("/")[0]
-            if group_id not in item_groups:
-                item_groups[group_id] = []
-            item_groups[group_id].append(item)
-        else:
-            item_groups[item.id] = [item]
-
-    return [_group_to_chat_item(items, cache_key) for items in item_groups.values()]
-
-
-def _group_to_chat_item(
-    items: list[llm.ChatItem], cache_key: Any
-) -> ChatCompletionMessageParam:
-    if len(items) == 1:
-        return _to_chat_item(items[0], cache_key)
-    else:
-        msg = {"role": "assistant", "tool_calls": []}
-        for item in items:
-            if item.type == "message":
-                assert item.role == "assistant", (
-                    "only assistant messages can be grouped"
-                )
-                assert "content" not in msg, (
-                    "only one assistant message is allowed in a group"
-                )
-
-                msg.update(_to_chat_item(item, cache_key))
-            elif item.type == "function_call":
-                msg["tool_calls"].append(
-                    {
-                        "id": item.call_id,
-                        "type": "function",
-                        "function": {"name": item.name, "arguments": item.arguments},
-                    }
-                )
-        return msg
-
-
-def _to_chat_item(msg: llm.ChatItem, cache_key: Any) -> ChatCompletionMessageParam:
-    if msg.type == "message":
-        list_content: list[ChatCompletionContentPartParam] = []
-        text_content = ""
-        for content in msg.content:
-            if isinstance(content, str):
-                if text_content:
-                    text_content += "\n"
-                text_content += content
-            elif isinstance(content, llm.ImageContent):
-                list_content.append(_to_image_content(content, cache_key))
-
-        if not list_content:
-            # certain providers require text-only content in a string vs a list.
-            # for max-compatibility, we will combine all text content into a single string.
-            return {
-                "role": msg.role,  # type: ignore
-                "content": text_content,
-            }
-
-        if text_content:
-            list_content.append({"type": "text", "text": text_content})
-
-        return {
-            "role": msg.role,  # type: ignore
-            "content": list_content,
-        }
-
-    elif msg.type == "function_call":
-        return {
-            "role": "assistant",
-            "tool_calls": [
-                {
-                    "id": msg.call_id,
-                    "type": "function",
-                    "function": {
-                        "name": msg.name,
-                        "arguments": msg.arguments,
-                    },
-                }
-            ],
-        }
-
-    elif msg.type == "function_call_output":
-        return {
-            "role": "tool",
-            "tool_call_id": msg.call_id,
-            "content": msg.output,
-        }
-
-
-def _to_image_content(
-    image: llm.ImageContent, cache_key: Any
-) -> ChatCompletionContentPartParam:
-    img = llm.utils.serialize_image(image)
-    if img.external_url:
-        return {
-            "type": "image_url",
-            "image_url": {
-                "url": img.external_url,
-                "detail": img.inference_detail,
-            },
-        }
-    if cache_key not in image._cache:
-        image._cache[cache_key] = img.data_bytes
-    b64_data = base64.b64encode(image._cache[cache_key]).decode("utf-8")
-    return {
-        "type": "image_url",
-        "image_url": {
-            "url": f"data:{img.mime_type};base64,{b64_data}",
-            "detail": img.inference_detail,
-        },
-    }

livekit_plugins_volcenginee-1.3.4/livekit/plugins/volcengine/version.py

@@ -1 +0,0 @@
-__version__ = "1.3.4"