livekit-plugins-volcengine 1.2.8__tar.gz → 1.2.9__tar.gz

{livekit_plugins_volcengine-1.2.8 → livekit_plugins_volcengine-1.2.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: livekit-plugins-volcengine
-Version: 1.2.8
+Version: 1.2.9
 Summary: LiveKit Agent Plugins for Volcengine
 Author-email: wangmengdi <790990241@qq.com>
 Keywords: audio,livekit,realtime,video,webrtc
@@ -14,10 +14,10 @@ Classifier: Topic :: Multimedia :: Sound/Audio
 Classifier: Topic :: Multimedia :: Video
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Requires-Python: >=3.9
-Requires-Dist: livekit-agents>=1.2.8
+Requires-Dist: livekit-agents>=1.2.9
 Requires-Dist: numpy
 Requires-Dist: openai>=1.75.0
-Requires-Dist: osc-data>=0.2.2
+Requires-Dist: osc-data==0.2.2
 Requires-Dist: pydantic
 Description-Content-Type: text/markdown
 

{livekit_plugins_volcengine-1.2.8 → livekit_plugins_volcengine-1.2.9}/livekit/plugins/volcengine/realtime.py

@@ -11,7 +11,7 @@ import gzip
 import uuid
 from collections.abc import Iterator
 from dataclasses import dataclass
-from typing import Literal
+from typing import Literal, Callable
 
 import aiohttp
 import numpy as np
@@ -176,6 +176,13 @@ class _RealtimeOptions:
     sample_rate: int = 24000
     num_channels: int = 1
     format: str = "pcm"
+    model: Literal["O", "SC"] = "O"
+    character_manifest: str = None
+    end_smooth_window_ms: int = 500
+    enable_volc_websearch: bool = False
+    volc_websearch_type: Literal["web_summary", "web"] = "web_summary"
+    volc_websearch_api_key: str = None
+    volc_websearch_no_result_message: str = "抱歉，我找不到相关信息。"
 
     @property
     def ws_url(self) -> str:
@@ -193,6 +200,11 @@ class _RealtimeOptions:
 
     def get_start_session_reqs(self, dialog_id: str | None) -> dict:
         start_session_req = {
+            "asr": {
+                "extra": {
+                    "end_smooth_window_ms": self.end_smooth_window_ms,
+                }
+            },
             "tts": {
                 "audio_config": {
                     "channel": self.num_channels,
@@ -206,7 +218,15 @@ class _RealtimeOptions:
                 "system_role": self.system_role,
                 "dialog_id": dialog_id or str(utils.shortuuid()),
                 "speaking_style": self.speaking_style,
-                "extra": {"strict_audit": False},
+                "character_manifest": self.character_manifest,
+                "extra": {
+                    "strict_audit": False,
+                    "enable_volc_websearch": self.enable_volc_websearch,
+                    "volc_websearch_type": self.volc_websearch_type,
+                    "volc_websearch_api_key": self.volc_websearch_api_key,
+                    "volc_websearch_no_result_message": self.volc_websearch_no_result_message,
+                    "model": self.model,
+                },
             },
         }
         return start_session_req
@@ -245,6 +265,14 @@ class RealtimeModel(llm.RealtimeModel):
         app_id: str | None = None,
         access_token: str | None = None,
         system_role: str | None = None,
+        character_manifest: str = None,
+        model: Literal["O", "SC"] = "O",
+        end_smooth_window_ms: int = 500,
+        enable_volc_websearch: bool = False,
+        volc_websearch_type: Literal["web_summary", "web"] = "web_summary",
+        volc_websearch_api_key: str = None,
+        volc_websearch_no_result_message: str = "抱歉，我找不到相关信息。",
+        rag_fn: Callable[[str], str] = None,
         audio_output: bool = True,
         modalities: NotGivenOr[list[Literal["text", "audio"]]] = NOT_GIVEN,
         http_session: aiohttp.ClientSession | None = None,
@@ -259,9 +287,18 @@ class RealtimeModel(llm.RealtimeModel):
                 user_transcription=True,
                 auto_tool_reply_generation=False,
                 audio_output=("audio" in modalities),
+                manual_function_calls=True
             )
         )
-
+        logger.info(f"Model: {model}")
+        logger.info(f"Character Manifest: {character_manifest}")
+        logger.info(f"End Smooth Window MS: {end_smooth_window_ms}")
+        logger.info(f"Enable Volc Websearch: {enable_volc_websearch}")
+        logger.info(f"Volc Websearch Type: {volc_websearch_type}")
+        logger.info(f"Volc Websearch API Key: {volc_websearch_api_key}")
+        logger.info(
+            f"Volc Websearch No Result Message: {volc_websearch_no_result_message}"
+        )
         app_id = app_id or os.environ.get("VOLCENGINE_REALTIME_APP_ID")
         if app_id is None:
             raise ValueError("VOLCENGINE_REALTIME_APP_ID is required")
@@ -273,15 +310,23 @@ class RealtimeModel(llm.RealtimeModel):
         self._opts = _RealtimeOptions(
             app_id=app_id,
             access_token=access_token,
-            max_session_duration=max_session_duration,
-            conn_options=conn_options,
             bot_name=bot_name,
             system_role=system_role,
             speaker=speaker,
             opening=opening,
             speaking_style=speaking_style,
+            character_manifest=character_manifest,
+            model=model,
+            end_smooth_window_ms=end_smooth_window_ms,
+            enable_volc_websearch=enable_volc_websearch,
+            volc_websearch_type=volc_websearch_type,
+            volc_websearch_api_key=volc_websearch_api_key,
+            volc_websearch_no_result_message=volc_websearch_no_result_message,
             modalities=modalities,
+            max_session_duration=max_session_duration,
+            conn_options=conn_options,
         )
+        self._rag_fn = rag_fn
         self._http_session = http_session
         self._sessions = weakref.WeakSet[RealtimeSession]()
 
@@ -441,7 +486,9 @@ class RealtimeSession(
             )
             self.emit("generation_created", generation_ev)
             item_id = utils.shortuuid()
-            modalities_fut: asyncio.Future[list[Literal["text", "audio"]]] = asyncio.Future()
+            modalities_fut: asyncio.Future[list[Literal["text", "audio"]]] = (
+                asyncio.Future()
+            )
             self._current_item = _MessageGeneration(
                 message_id=item_id,
                 text_ch=utils.aio.Chan(),
@@ -536,7 +583,9 @@ class RealtimeSession(
 
                                 self.emit("generation_created", generation_ev)
                                 item_id = utils.shortuuid()
-                                modalities_fut: asyncio.Future[list[Literal["text", "audio"]]] = asyncio.Future()
+                                modalities_fut: asyncio.Future[
+                                    list[Literal["text", "audio"]]
+                                ] = asyncio.Future()
                                 self._current_item = _MessageGeneration(
                                     message_id=item_id,
                                     text_ch=utils.aio.Chan(),
@@ -547,7 +596,9 @@ class RealtimeSession(
                                     self._current_item.audio_ch.close()
                                     self._current_item.modalities.set_result(["text"])  # type: ignore[union-attr]
                                 else:
-                                    self._current_item.modalities.set_result(["audio", "text"])  # type: ignore[union-attr]
+                                    self._current_item.modalities.set_result(
+                                        ["audio", "text"]
+                                    )  # type: ignore[union-attr]
                                 self._current_generation.message_ch.send_nowait(
                                     llm.MessageGeneration(
                                         message_id=item_id,
@@ -565,6 +616,22 @@ class RealtimeSession(
                                 user_transcription_enabled=False
                             ),
                         )
+                        if self._realtime_model._rag_fn is not None:
+                            logger.info("rag start")
+                            rag_result = self._realtime_model._rag_fn(transcription)
+                            payload = {
+                                "external_rag": rag_result,
+                            }
+                            payload_bytes = str.encode(json.dumps(payload))
+                            payload_bytes = gzip.compress(payload_bytes)
+                            chat_rag_text_request = bytearray(generate_header())
+                            chat_rag_text_request.extend(int(502).to_bytes(4, "big"))
+                            chat_rag_text_request.extend((len(self.session_id)).to_bytes(4, "big"))
+                            chat_rag_text_request.extend(str.encode(self.session_id))
+                            chat_rag_text_request.extend((len(payload_bytes)).to_bytes(4, "big"))
+                            chat_rag_text_request.extend(payload_bytes)
+                            await ws_conn.send_bytes(chat_rag_text_request)
+                            logger.info("rag end")
                         logger.info("llm start")
                         logger.info("tts start")
 

livekit_plugins_volcengine-1.2.9/livekit/plugins/volcengine/version.py

@@ -0,0 +1 @@
+__version__ = "1.2.9"

{livekit_plugins_volcengine-1.2.8 → livekit_plugins_volcengine-1.2.9}/pyproject.toml

@@ -10,10 +10,10 @@ keywords = ["webrtc", "realtime", "audio", "video", "livekit"]
 requires-python = ">=3.9"
 dependencies = [
     "openai>=1.75.0",
-    "osc-data>=0.2.2",
+    "osc-data==0.2.2",
     "pydantic",
     "numpy",
-    "livekit-agents>=1.2.8",
+    "livekit-agents>=1.2.9",
 ]
 
 classifiers = [

livekit_plugins_volcengine-1.2.8/livekit/plugins/volcengine/version.py

@@ -1 +0,0 @@
-__version__ = "1.2.8"