agora-python-server-sdk 2.1.4__tar.gz → 2.1.5__tar.gz

agora_python_server_sdk-2.1.5/PKG-INFO

@@ -0,0 +1,140 @@
+Metadata-Version: 2.1
+Name: agora_python_server_sdk
+Version: 2.1.5
+Summary: A Python SDK for Agora Server
+Home-page: https://github.com/AgoraIO-Extensions/Agora-Python-Server-SDK
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Topic :: Multimedia :: Sound/Audio
+Classifier: Topic :: Multimedia :: Video
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3 :: Only
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+
+# Note
+- This is a Python SDK wrapper for the Agora RTC SDK.
+- It supports Linux and Mac platforms.
+- The examples are provided as very simple demonstrations and are not recommended for use in production environments.
+
+# Very Important Notice !!!
+- A process can only have one instance.
+- An instance can have multiple connections.
+- In all observers or callbacks, you must not call the SDK's own APIs, nor perform CPU-intensive tasks in the callbacks; data copying is allowed.
+
+# Required Operating Systems and Python Versions
+- Supported Linux versions:
+  - Ubuntu 18.04 LTS and above
+  - CentOS 7.0 and above
+  
+- Supported Mac versions:
+  - MacOS 13 and above
+
+- Python version:
+  - Python 3.10 and above
+
+# Using Agora-Python-Server-SDK
+```
+pip install agora_python_server_sdk
+```
+
+# Running Examples
+
+## Preparing Test Data
+- Download and unzip [test_data.zip](https://download.agora.io/demo/test/test_data_202408221437.zip) to the Agora-Python-Server-SDK directory.
+
+## Executing Test Script
+```
+python agora_rtc/examples/example_audio_pcm_send.py --appId=xxx --channelId=xxx --userId=xxx --audioFile=./test_data/demo.pcm --sampleRate=16000 --numOfChannels=1
+```
+
+# Change log
+
+## 2024.12.03 release Version 2.1.5
+- Modifications:
+  - LocalUser/audioTrack:
+    -- When the scenario is chorus, developers don't need to call setSendDelayInMs.
+    -- When the scenario is chorus, developers don't need to set the audio scenario of the track to chorus.
+    -- NOTE: This can reduce the difficulty for developers. In AI scenarios, developers only need to set the service to chorus.
+- Additions:
+  -- Added the VadDump class, which can assist in troubleshooting vad issues in the testing environment. However, it should not be enabled in the online env  ironment.
+  -- Added the on_volume_indication callback.
+  -- Added the on_remote_video_track_state_changed callback.
+- Removals:
+  -- Removed Vad V1 version, only retaining the V2 version. Refer to voice_detection.py and sample_audio_vad.py.
+- Updates:
+  -- Updated relevant samples: audioconsume, vad sample.
+
+## 2024.11.12 release 2.1.4
+- Modify the type of metadata in videoFrame from str to bytes type to be consistent with C++; thus, it can support byte streams.
+- The internal encapsulation of ExteranlVideoFrame has been modified to support byte streams. Regarding the support for alpha encoding, a logical judgment has been made. If fill_alpha_buffer is 0, it will not be processed.
+## 2024.11.11 release 2.1.3
+- Added a new sample: example_jpeg_send.py which can push JPEG files or JPEG streams to a channel.
+-
+- Performance overhead, as noted in the example comments, can be summarized as follows:
+- For a 1920x1080 JPEG file, the process from reading the file to converting it to an RGBA bytearray - takes approximately 11 milliseconds.
+
+
+## 2024.11.07 release 2.1.2
+- Updates `user_id` in the `AudioVolumeInfoInner and AudioVolumeInfo` structure to `str` type.
+- Fixes the bug in `_on_audio_volume_indication` callback, where it could only handle one callback to speaker_number 
+- Corrects the parameter type in `IRTCLocalUserObserver::on_audio_volume_indication` callback to `list` type.
+
+## 2024.10.29 release 2.1.1
+
+Add audio VAD interface of version 2 and corresponding example.
+
+## 2024.10.24 release 2.1.0
+
+Fixed some bug.
+
+
+### Common Usage Q&A
+## The relationship between service and process?
+- A process can only have one service, and the service can only be initialized once.
+- A service can only have one media_node_factory.
+- A service can have multiple connections.
+- Release media_node_factory.release() and service.release() when the process exits.
+## If using Docker with one user per Docker, when the user starts Docker and logs out, how should Docker be released?
+- In this case, create service/media_node_factory and connection when the process starts.
+- Release service/media_node_factory and connection when the process exits, ensuring that...
+## If Docker is used to support multiple users and Docker runs for a long time, what should be done?
+- In this case, we recommend using the concept of a connection pool.
+- Create service/media_node_factory and a connection pool (only new connections, without initialization) when the process starts.
+- When a user logs in, get a connection from the connection pool, initialize it, execute con.connect() and set up callbacks, and then join the channel.
+- Handle business operations.
+- When a user logs out, execute con.disconnect() and release the audio/video tracks and observers associated with the connection, but do not call con.release(); then put the connection back into the connection pool.
+- When the process exits, release the connection pool (release each con.release()), service/media_node_factory, and the connection pool (release each con.release()) to ensure resource release and optimal performance.
+
+## Use of VAD
+# Source code: voice_detection.py
+# Sample code: example_audio_vad.py
+# It is recommended to use VAD V2 version, and the class is: AudioVadV2; Reference: voice_detection.py.
+# Use of VAD:
+  1. Call _vad_instance.init(AudioVadConfigV2) to initialize the vad instance. Reference: voice_detection.py. Assume the instance is: _vad_instance
+  2. In audio_frame_observer::on_playback_audio_frame_before_mixing(audio_frame):
+
+  3. Call the process of the vad module: state, bytes = _vad_instance.process(audio_frame)
+Judge the value of state according to the returned state, and do corresponding processing.
+
+    A. If state is _vad_instance._vad_state_startspeaking, it indicates that the user is "starting to speak", and speech recognition (STT/ASR) operations can be started. Remember: be sure to pass the returned bytes to the recognition module instead of the original audio_frame, otherwise the recognition result will be incorrect.
+    B. If state is _vad_instance._vad_state_stopspeaking, it indicates that the user is "stopping speaking", and speech recognition (STT/ASR) operations can be stopped. Remember: be sure to pass the returned bytes to the recognition module instead of the original audio_frame, otherwise the recognition result will be incorrect.
+    C. If state is _vad_instance._vad_state_speaking, it indicates that the user is "speaking", and speech recognition (STT/ASR) operations can be continued. Remember: be sure to pass the returned bytes to the recognition module instead of the original audio_frame, otherwise the recognition result will be incorrect.
+# Note: 
+  If the vad module is used and it is expected to use the vad module for speech recognition (STT/ASR) and other operations, then be sure to pass the returned bytes to the recognition module instead of the original audio_frame, otherwise the recognition result will be incorrect.
+# How to better troubleshoot VAD issues: It includes two aspects, configuration and debugging.
+  1. Ensure that the initialization parameters of the vad module are correct. Reference: voice_detection.py.
+  2. In state, bytes = on_playback_audio_frame_before_mixing(audio_frame):
+
+    - A . Save the data of audio_frame to a local file, reference: example_audio_pcm_send.py. This is to record the original audio data. For example, it can be named: source_{time.time()*1000}.pcm
+    - B.Save the result of each vad processing:
+
+      - a When state == start_speaking: create a new binary file, for example, named: vad_{time.time()*1000}.pcm, and write bytes to the file.
+      - b When state == speaking: write bytes to the file.
+      - c When state == stop_speaking: write bytes to the file and close the file.
+    Note: In this way, problems can be troubleshot based on the original audio file and the audio file processed by vad. This function can be disabled in the production environment.
+### How to push the audio generated by TTS into the channel?
+  # Source code: audio_consumer.py
+  # Sample code: example_audio_consumer.py
+### How to release resources?

agora_python_server_sdk-2.1.5/README.md

@@ -0,0 +1,125 @@
+# Note
+- This is a Python SDK wrapper for the Agora RTC SDK.
+- It supports Linux and Mac platforms.
+- The examples are provided as very simple demonstrations and are not recommended for use in production environments.
+
+# Very Important Notice !!!
+- A process can only have one instance.
+- An instance can have multiple connections.
+- In all observers or callbacks, you must not call the SDK's own APIs, nor perform CPU-intensive tasks in the callbacks; data copying is allowed.
+
+# Required Operating Systems and Python Versions
+- Supported Linux versions:
+  - Ubuntu 18.04 LTS and above
+  - CentOS 7.0 and above
+  
+- Supported Mac versions:
+  - MacOS 13 and above
+
+- Python version:
+  - Python 3.10 and above
+
+# Using Agora-Python-Server-SDK
+```
+pip install agora_python_server_sdk
+```
+
+# Running Examples
+
+## Preparing Test Data
+- Download and unzip [test_data.zip](https://download.agora.io/demo/test/test_data_202408221437.zip) to the Agora-Python-Server-SDK directory.
+
+## Executing Test Script
+```
+python agora_rtc/examples/example_audio_pcm_send.py --appId=xxx --channelId=xxx --userId=xxx --audioFile=./test_data/demo.pcm --sampleRate=16000 --numOfChannels=1
+```
+
+# Change log
+
+## 2024.12.03 release Version 2.1.5
+- Modifications:
+  - LocalUser/audioTrack:
+    -- When the scenario is chorus, developers don't need to call setSendDelayInMs.
+    -- When the scenario is chorus, developers don't need to set the audio scenario of the track to chorus.
+    -- NOTE: This can reduce the difficulty for developers. In AI scenarios, developers only need to set the service to chorus.
+- Additions:
+  -- Added the VadDump class, which can assist in troubleshooting vad issues in the testing environment. However, it should not be enabled in the online env  ironment.
+  -- Added the on_volume_indication callback.
+  -- Added the on_remote_video_track_state_changed callback.
+- Removals:
+  -- Removed Vad V1 version, only retaining the V2 version. Refer to voice_detection.py and sample_audio_vad.py.
+- Updates:
+  -- Updated relevant samples: audioconsume, vad sample.
+
+## 2024.11.12 release 2.1.4
+- Modify the type of metadata in videoFrame from str to bytes type to be consistent with C++; thus, it can support byte streams.
+- The internal encapsulation of ExteranlVideoFrame has been modified to support byte streams. Regarding the support for alpha encoding, a logical judgment has been made. If fill_alpha_buffer is 0, it will not be processed.
+## 2024.11.11 release 2.1.3
+- Added a new sample: example_jpeg_send.py which can push JPEG files or JPEG streams to a channel.
+-
+- Performance overhead, as noted in the example comments, can be summarized as follows:
+- For a 1920x1080 JPEG file, the process from reading the file to converting it to an RGBA bytearray - takes approximately 11 milliseconds.
+
+
+## 2024.11.07 release 2.1.2
+- Updates `user_id` in the `AudioVolumeInfoInner and AudioVolumeInfo` structure to `str` type.
+- Fixes the bug in `_on_audio_volume_indication` callback, where it could only handle one callback to speaker_number 
+- Corrects the parameter type in `IRTCLocalUserObserver::on_audio_volume_indication` callback to `list` type.
+
+## 2024.10.29 release 2.1.1
+
+Add audio VAD interface of version 2 and corresponding example.
+
+## 2024.10.24 release 2.1.0
+
+Fixed some bug.
+
+
+### Common Usage Q&A
+## The relationship between service and process?
+- A process can only have one service, and the service can only be initialized once.
+- A service can only have one media_node_factory.
+- A service can have multiple connections.
+- Release media_node_factory.release() and service.release() when the process exits.
+## If using Docker with one user per Docker, when the user starts Docker and logs out, how should Docker be released?
+- In this case, create service/media_node_factory and connection when the process starts.
+- Release service/media_node_factory and connection when the process exits, ensuring that...
+## If Docker is used to support multiple users and Docker runs for a long time, what should be done?
+- In this case, we recommend using the concept of a connection pool.
+- Create service/media_node_factory and a connection pool (only new connections, without initialization) when the process starts.
+- When a user logs in, get a connection from the connection pool, initialize it, execute con.connect() and set up callbacks, and then join the channel.
+- Handle business operations.
+- When a user logs out, execute con.disconnect() and release the audio/video tracks and observers associated with the connection, but do not call con.release(); then put the connection back into the connection pool.
+- When the process exits, release the connection pool (release each con.release()), service/media_node_factory, and the connection pool (release each con.release()) to ensure resource release and optimal performance.
+
+## Use of VAD
+# Source code: voice_detection.py
+# Sample code: example_audio_vad.py
+# It is recommended to use VAD V2 version, and the class is: AudioVadV2; Reference: voice_detection.py.
+# Use of VAD:
+  1. Call _vad_instance.init(AudioVadConfigV2) to initialize the vad instance. Reference: voice_detection.py. Assume the instance is: _vad_instance
+  2. In audio_frame_observer::on_playback_audio_frame_before_mixing(audio_frame):
+
+  3. Call the process of the vad module: state, bytes = _vad_instance.process(audio_frame)
+Judge the value of state according to the returned state, and do corresponding processing.
+
+    A. If state is _vad_instance._vad_state_startspeaking, it indicates that the user is "starting to speak", and speech recognition (STT/ASR) operations can be started. Remember: be sure to pass the returned bytes to the recognition module instead of the original audio_frame, otherwise the recognition result will be incorrect.
+    B. If state is _vad_instance._vad_state_stopspeaking, it indicates that the user is "stopping speaking", and speech recognition (STT/ASR) operations can be stopped. Remember: be sure to pass the returned bytes to the recognition module instead of the original audio_frame, otherwise the recognition result will be incorrect.
+    C. If state is _vad_instance._vad_state_speaking, it indicates that the user is "speaking", and speech recognition (STT/ASR) operations can be continued. Remember: be sure to pass the returned bytes to the recognition module instead of the original audio_frame, otherwise the recognition result will be incorrect.
+# Note: 
+  If the vad module is used and it is expected to use the vad module for speech recognition (STT/ASR) and other operations, then be sure to pass the returned bytes to the recognition module instead of the original audio_frame, otherwise the recognition result will be incorrect.
+# How to better troubleshoot VAD issues: It includes two aspects, configuration and debugging.
+  1. Ensure that the initialization parameters of the vad module are correct. Reference: voice_detection.py.
+  2. In state, bytes = on_playback_audio_frame_before_mixing(audio_frame):
+
+    - A . Save the data of audio_frame to a local file, reference: example_audio_pcm_send.py. This is to record the original audio data. For example, it can be named: source_{time.time()*1000}.pcm
+    - B.Save the result of each vad processing:
+
+      - a When state == start_speaking: create a new binary file, for example, named: vad_{time.time()*1000}.pcm, and write bytes to the file.
+      - b When state == speaking: write bytes to the file.
+      - c When state == stop_speaking: write bytes to the file and close the file.
+    Note: In this way, problems can be troubleshot based on the original audio file and the audio file processed by vad. This function can be disabled in the production environment.
+### How to push the audio generated by TTS into the channel?
+  # Source code: audio_consumer.py
+  # Sample code: example_audio_consumer.py
+### How to release resources?

{agora_python_server_sdk-2.1.4 → agora_python_server_sdk-2.1.5}/agora/rtc/agora_service.py

@@ -75,6 +75,8 @@ class AgoraService:
         if result == 0:
             self.inited = True
         logger.debug(f'Initialization result: {result}')
+        self._is_low_delay = True if  config.audio_scenario == AudioScenarioType.AUDIO_SCENARIO_CHORUS  else False
+        
 
         # to enable plugin
         provider = "agora.builtin"
@@ -87,6 +89,9 @@ class AgoraService:
         agora_parameter = self.get_agora_parameter()
         agora_parameter.set_int("rtc.set_app_type", 18)
 
+        # force audio vad v2 to be enabled
+        agora_parameter.set_parameters("{\"che.audio.label.enable\": true}")
+
         if config.log_path:
             log_size = 512 * 1024
             if config.log_size > 0:
@@ -128,7 +133,7 @@ class AgoraService:
         rtc_conn_handle = agora_rtc_conn_create(self.service_handle, ctypes.byref(RTCConnConfigInner.create(con_config)))
         if rtc_conn_handle is None:
             return None
-        return RTCConnection(rtc_conn_handle)
+        return RTCConnection(rtc_conn_handle, self._is_low_delay)
 
     # createCustomAudioTrackPcm: creatae a custom audio track from pcm data sender
     def create_custom_audio_track_pcm(self, audio_pcm_data_sender: AudioPcmDataSender) -> LocalAudioTrack:
@@ -138,7 +143,11 @@ class AgoraService:
         custom_audio_track = agora_service_create_custom_audio_track_pcm(self.service_handle, audio_pcm_data_sender.sender_handle)
         if custom_audio_track is None:
             return None
-        return LocalAudioTrack(custom_audio_track)
+        local_track =  LocalAudioTrack(custom_audio_track)
+        #default for ai senario to set min delay to 10ms
+        if local_track is not None:
+            local_track.set_send_delay_ms(10)
+        return local_track
     # mix_mode: MIX_ENABLED = 0, MIX_DISABLED = 1
 
     def create_custom_audio_track_encoded(self, audio_encoded_frame_sender: AudioEncodedFrameSender, mix_mode: int):

{agora_python_server_sdk-2.1.4 → agora_python_server_sdk-2.1.5}/agora/rtc/local_user.py

@@ -506,7 +506,7 @@ class LocalUser:
         if ret < 0:
             logger.error("Failed to unsubscribe all video")
         else:
-            self.del_remote_video_map_all(None)
+            self.del_remote_video_map(None)
         return ret
 
     def set_audio_volume_indication_parameters(self, interval_in_ms, smooth, report_vad):

{agora_python_server_sdk-2.1.4 → agora_python_server_sdk-2.1.5}/agora/rtc/rtc_connection.py

@@ -53,16 +53,19 @@ agora_rtc_conn_renew_token.argtypes = [AGORA_HANDLE, ctypes.c_char_p]
 
 
 class RTCConnection:
-    def __init__(self, conn_handle) -> None:
+    def __init__(self, conn_handle, is_low_delay: bool = False) -> None:
         self.conn_handle = conn_handle
         self.con_observer = None
         self.local_user = None
         self.local_user_handle = agora_rtc_conn_get_local_user(conn_handle)
         if self.local_user_handle:
             self.local_user = LocalUser(self.local_user_handle, self)
+        #added to set low delay mode
+        if is_low_delay == True: 
+            self.local_user.set_audio_scenario(AudioScenarioType.AUDIO_SCENARIO_CHORUS)
         # add to map
-        AgoraHandleInstanceMap().set_local_user_map(self.conn_handle, self)
-        AgoraHandleInstanceMap().set_con_map(self.conn_handle, self)
+        #AgoraHandleInstanceMap().set_local_user_map(self.conn_handle, self)
+        #AgoraHandleInstanceMap().set_con_map(self.conn_handle, self)
 
     #
     def connect(self, token: str, chan_id: str, user_id: str) -> int:
@@ -130,7 +133,7 @@ class RTCConnection:
     def release(self):
         # release local user map
         if self.conn_handle:
-            AgoraHandleInstanceMap().del_local_user_map(self.conn_handle)
+            #AgoraHandleInstanceMap().del_local_user_map(self.conn_handle)
             agora_rtc_conn_release(self.conn_handle)
         self.conn_handle = None
         self.local_user = None

agora_python_server_sdk-2.1.5/agora/rtc/utils/audio_consumer.py

@@ -0,0 +1,133 @@
+#!env python
+import threading
+import time
+from agora.rtc.audio_pcm_data_sender import PcmAudioFrame
+from agora.rtc.audio_pcm_data_sender import AudioPcmDataSender
+import logging
+import asyncio
+logger = logging.getLogger(__name__)
+
+
+
+"""
+# AudioConsumer
+# 基础类，用于消费PCM数据，并将PCM数据推送到RTC频道中
+# 在AI场景中：
+#   当TTS有数据返回的时候：调用AudioConsumer::push_pcm_data方法，将返回的TTS数据直接push到AudioConsumer
+#   在另外的一个“timer”的触发函数中，调用 AudioConsumer::consume()方法，将数据推送到rtc
+    # 推荐：
+    # “Timer”可以是asycio的模式；也可以是threading.Timer的模式；也可以和业务已有的timer结合在一起使用，都可以。只需要在timer 触发的函数中，调用 AudioConsumer::consume()即可
+    # “Timer”的触发间隔，可以和业务已有的timer间隔一致，也可以根据业务需求调整，推荐在40～80ms之间  
+ 
+AudioConsumer调用方式：
+1. 使用该类的前提：
+    - 需要客户在应用层自己实现一个timer，该timer间隔需要在[40ms,80ms]之间。这个timer的触发方法下面用app::TimeFunc表示。
+    - 一个用户只能对应一个AudioConsumer对象，也就是保障一个生产者产生的内容对应一个消费者。
+2. 使用方式：
+    A 对每一个“生产pcm数据“的 userid 创建一个AudioConsumer对象，也就是保障一个生产者产生的内容对应一个消费者。
+    B 当有pcm数据生成的时候，比如TTS的返回，调用 AudioConsumer::push_pcm_data(data)
+    C 当需要消费的时候（通常用app::TimerFunc)，调用 AudioConsumer::consume()方法，会自动完成对数据的消费，也就是推送到rtc 频道中
+    D 如果需要打断：也就是AI场景中，要停止播放当前AI的对话：调用 AudioConsumer::clear()方法,会自动清空当前buffer中的数据
+    E 退出的时候，调用release()方法，释放资源
+"""
+class AudioConsumer:
+    def __init__(self, pcm_sender: AudioPcmDataSender, sample_rate: int, channels: int) -> None:
+        self._lock = threading.Lock()
+        self._start_time = 0
+        self._data = bytearray()
+        self._consumed_packages = 0
+        self._pcm_sender = pcm_sender
+        self._frame = PcmAudioFrame()
+        #init sample rate and channels
+        self._frame.sample_rate = sample_rate
+        self._frame.number_of_channels = channels
+
+        #audio parame
+        self._frame.bytes_per_sample = 2
+        self._bytes_per_frame = sample_rate // 100 * channels * 2
+        self._samples_per_channel = sample_rate // 100* channels
+        #init pcmaudioframe
+        self._frame.timestamp = 0
+
+        self._init = True
+        
+        pass
+    def push_pcm_data(self, data) ->None:
+        if self._init == False:
+            return
+        # add to buffer, lock
+        with self._lock:
+            self._data += data
+        pass
+    def _reset(self):
+        if self._init == False:
+            return
+        self._start_time = time.time()*1000
+        self._consumed_packages = 0
+        pass
+ 
+    def consume(self):
+        print("consume begin")
+        if self._init == False:
+            return
+        now = time.time()*1000
+        elapsed_time = int(now - self._start_time)
+        expected_total_packages = int(elapsed_time//10)
+        besent_packages = expected_total_packages - self._consumed_packages
+        data_len = len(self._data)
+
+        if besent_packages > 18 and data_len //self._bytes_per_frame < 18: #for fist time, if data_len is not enough, just return and wait for next time
+            #print("-----underflow data")
+            return
+        if besent_packages > 18: #rest to start state, push 18 packs in Start_STATE
+            self._reset()
+            besent_packages = min(18, data_len//self._bytes_per_frame)
+            self._consumed_packages = -besent_packages
+            
+
+        #get min packages
+        act_besent_packages = (int)(min(besent_packages, data_len//self._bytes_per_frame))
+        #print("consume 1:", act_besent_packages, data_len)
+        if act_besent_packages < 1:
+            return
+       
+        #construct an audio frame to push
+        #frame = PcmAudioFrame()
+        with self._lock:
+            #frame = PcmAudioFrame()
+            self._frame.data = self._data[:self._bytes_per_frame*act_besent_packages]
+            self._frame.timestamp = 0
+            self._frame.samples_per_channel = self._samples_per_channel*act_besent_packages
+           
+            #reset data
+            self._data = self._data[self._bytes_per_frame*act_besent_packages:]
+            self._consumed_packages += act_besent_packages
+
+        self._pcm_sender.send_audio_pcm_data(self._frame)
+        #print(f"act_besent_packages: {now},{now - self._start_time}, {besent_packages}, {act_besent_packages},{self._consumed_packages},{data_len}")
+        pass
+
+    def len(self) -> int:
+        if self._init == False:
+            return 0
+        with self._lock:
+            return len(self._data)
+    def clear(self):
+        if self._init == False:
+            return
+        with self._lock:
+            self._data = bytearray()
+        pass
+    def release(self):
+        if self._init == False:
+            return
+        
+        self._init = False
+        with self._lock:
+            self._data = None
+            self._frame = None
+            self._pcm_sender = None
+        self._lock = None
+        pass
+
+

agora_python_server_sdk-2.1.5/agora/rtc/utils/vad_dump.py

@@ -0,0 +1,104 @@
+#!env python
+import time
+from datetime import datetime
+import logging
+import os
+from agora.rtc.agora_base import AudioFrame
+logger = logging.getLogger(__name__)
+
+
+
+"""
+## VadDump helper class
+"""
+class VadDump():
+    def __init__(self, path: str) -> None:
+        self._file_path = path
+        self._count = 0
+        self._frame_count = 0
+        self._is_open = False
+        self._source_file = None
+        self._label_file = None
+        self._vad_file = None
+        #check path is existed or not? if not, create new dir
+        if self._check_directory_exists(path) is False:
+            os.makedirs(path)
+        # make suddirectory : ("%s/%04d%02d%02d%02d%02d%02d
+        now = datetime.now()
+        #format to YYYYMMDDHHMMSS
+        self._file_path = "%s/%04d%02d%02d%02d%02d%02d" % (path, now.year, now.month, now.day, now.hour, now.minute, now.second)
+        os.makedirs(self._file_path)
+
+
+        pass
+    def _check_directory_exists(self, path: str) -> bool:
+        return os.path.exists(path) and os.path.isdir(path)
+    def _create_vad_file(self) -> None:
+        self._close_vad_file()
+        #create a new one
+        vad_file_path = "%s/vad_%d.pcm" % (self._file_path, self._count)
+        self._vad_file = open(vad_file_path, "wb")
+        #increment the count
+        self._count += 1
+        pass
+    def _close_vad_file(self) -> None:
+        if self._vad_file:
+            self._vad_file.close()
+            self._vad_file = None
+        pass
+    def open(self) -> int:
+        if self._is_open is True:
+            return 1
+        self._is_open = True
+        #open source file
+        source_file_path = self._file_path + "/source.pcm"
+        self._source_file = open(source_file_path, "wb") 
+        #open label file
+        label_file_path = self._file_path + "/label.txt"
+        self._label_file = open(label_file_path, "w")
+        #open vad file
+        pass
+    def write(self, frame:AudioFrame, vad_result_bytes: bytearray, vad_result_state : int) -> None:
+        #write pcm to source
+        if self._is_open is False:
+            return
+        if self._source_file:
+            self._source_file.write(frame.buffer)
+        # fomat frame 's label informaiton and write to label file
+        if self._label_file:
+            label_str = "ct:%d fct:%d state:%d far:%d vop:%d rms:%d pitch:%d mup:%d\n" % (self._count, self._frame_count,vad_result_state, frame.far_field_flag, frame.voice_prob, frame.rms, frame.pitch, frame.music_prob)
+            self._label_file.write(label_str)
+        #write to vad result
+        if vad_result_state == 1: # start speaking
+            #open new vad file and write header
+            self._create_vad_file()
+            if self._vad_file:
+                self._vad_file.write(vad_result_bytes)
+        if vad_result_state == 2:
+            if self._vad_file:
+                self._vad_file.write(vad_result_bytes)
+        if vad_result_state == 3:
+            if self._vad_file:
+                self._vad_file.write(vad_result_bytes)
+            self._close_vad_file()
+        #increment frame counter
+        self._frame_count += 1
+        pass
+    def close(self) -> None:
+        if self._is_open == False:
+            return 
+        self._is_open = False
+        if self._vad_file:
+            self._close_vad_file()
+            self._vad_file = None
+        if self._label_file:
+            self._label_file.close()
+            self._label_file = None
+        self._close_vad_file()
+
+        # assign to None
+        self._count = 0
+        self._frame_count = 0
+        self._file_path = None
+
+        pass

agora_python_server_sdk-2.1.5/agora_python_server_sdk.egg-info/PKG-INFO

@@ -0,0 +1,140 @@
+Metadata-Version: 2.1
+Name: agora_python_server_sdk
+Version: 2.1.5
+Summary: A Python SDK for Agora Server
+Home-page: https://github.com/AgoraIO-Extensions/Agora-Python-Server-SDK
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Topic :: Multimedia :: Sound/Audio
+Classifier: Topic :: Multimedia :: Video
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3 :: Only
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+
+# Note
+- This is a Python SDK wrapper for the Agora RTC SDK.
+- It supports Linux and Mac platforms.
+- The examples are provided as very simple demonstrations and are not recommended for use in production environments.
+
+# Very Important Notice !!!
+- A process can only have one instance.
+- An instance can have multiple connections.
+- In all observers or callbacks, you must not call the SDK's own APIs, nor perform CPU-intensive tasks in the callbacks; data copying is allowed.
+
+# Required Operating Systems and Python Versions
+- Supported Linux versions:
+  - Ubuntu 18.04 LTS and above
+  - CentOS 7.0 and above
+  
+- Supported Mac versions:
+  - MacOS 13 and above
+
+- Python version:
+  - Python 3.10 and above
+
+# Using Agora-Python-Server-SDK
+```
+pip install agora_python_server_sdk
+```
+
+# Running Examples
+
+## Preparing Test Data
+- Download and unzip [test_data.zip](https://download.agora.io/demo/test/test_data_202408221437.zip) to the Agora-Python-Server-SDK directory.
+
+## Executing Test Script
+```
+python agora_rtc/examples/example_audio_pcm_send.py --appId=xxx --channelId=xxx --userId=xxx --audioFile=./test_data/demo.pcm --sampleRate=16000 --numOfChannels=1
+```
+
+# Change log
+
+## 2024.12.03 release Version 2.1.5
+- Modifications:
+  - LocalUser/audioTrack:
+    -- When the scenario is chorus, developers don't need to call setSendDelayInMs.
+    -- When the scenario is chorus, developers don't need to set the audio scenario of the track to chorus.
+    -- NOTE: This can reduce the difficulty for developers. In AI scenarios, developers only need to set the service to chorus.
+- Additions:
+  -- Added the VadDump class, which can assist in troubleshooting vad issues in the testing environment. However, it should not be enabled in the online env  ironment.
+  -- Added the on_volume_indication callback.
+  -- Added the on_remote_video_track_state_changed callback.
+- Removals:
+  -- Removed Vad V1 version, only retaining the V2 version. Refer to voice_detection.py and sample_audio_vad.py.
+- Updates:
+  -- Updated relevant samples: audioconsume, vad sample.
+
+## 2024.11.12 release 2.1.4
+- Modify the type of metadata in videoFrame from str to bytes type to be consistent with C++; thus, it can support byte streams.
+- The internal encapsulation of ExteranlVideoFrame has been modified to support byte streams. Regarding the support for alpha encoding, a logical judgment has been made. If fill_alpha_buffer is 0, it will not be processed.
+## 2024.11.11 release 2.1.3
+- Added a new sample: example_jpeg_send.py which can push JPEG files or JPEG streams to a channel.
+-
+- Performance overhead, as noted in the example comments, can be summarized as follows:
+- For a 1920x1080 JPEG file, the process from reading the file to converting it to an RGBA bytearray - takes approximately 11 milliseconds.
+
+
+## 2024.11.07 release 2.1.2
+- Updates `user_id` in the `AudioVolumeInfoInner and AudioVolumeInfo` structure to `str` type.
+- Fixes the bug in `_on_audio_volume_indication` callback, where it could only handle one callback to speaker_number 
+- Corrects the parameter type in `IRTCLocalUserObserver::on_audio_volume_indication` callback to `list` type.
+
+## 2024.10.29 release 2.1.1
+
+Add audio VAD interface of version 2 and corresponding example.
+
+## 2024.10.24 release 2.1.0
+
+Fixed some bug.
+
+
+### Common Usage Q&A
+## The relationship between service and process?
+- A process can only have one service, and the service can only be initialized once.
+- A service can only have one media_node_factory.
+- A service can have multiple connections.
+- Release media_node_factory.release() and service.release() when the process exits.
+## If using Docker with one user per Docker, when the user starts Docker and logs out, how should Docker be released?
+- In this case, create service/media_node_factory and connection when the process starts.
+- Release service/media_node_factory and connection when the process exits, ensuring that...
+## If Docker is used to support multiple users and Docker runs for a long time, what should be done?
+- In this case, we recommend using the concept of a connection pool.
+- Create service/media_node_factory and a connection pool (only new connections, without initialization) when the process starts.
+- When a user logs in, get a connection from the connection pool, initialize it, execute con.connect() and set up callbacks, and then join the channel.
+- Handle business operations.
+- When a user logs out, execute con.disconnect() and release the audio/video tracks and observers associated with the connection, but do not call con.release(); then put the connection back into the connection pool.
+- When the process exits, release the connection pool (release each con.release()), service/media_node_factory, and the connection pool (release each con.release()) to ensure resource release and optimal performance.
+
+## Use of VAD
+# Source code: voice_detection.py
+# Sample code: example_audio_vad.py
+# It is recommended to use VAD V2 version, and the class is: AudioVadV2; Reference: voice_detection.py.
+# Use of VAD:
+  1. Call _vad_instance.init(AudioVadConfigV2) to initialize the vad instance. Reference: voice_detection.py. Assume the instance is: _vad_instance
+  2. In audio_frame_observer::on_playback_audio_frame_before_mixing(audio_frame):
+
+  3. Call the process of the vad module: state, bytes = _vad_instance.process(audio_frame)
+Judge the value of state according to the returned state, and do corresponding processing.
+
+    A. If state is _vad_instance._vad_state_startspeaking, it indicates that the user is "starting to speak", and speech recognition (STT/ASR) operations can be started. Remember: be sure to pass the returned bytes to the recognition module instead of the original audio_frame, otherwise the recognition result will be incorrect.
+    B. If state is _vad_instance._vad_state_stopspeaking, it indicates that the user is "stopping speaking", and speech recognition (STT/ASR) operations can be stopped. Remember: be sure to pass the returned bytes to the recognition module instead of the original audio_frame, otherwise the recognition result will be incorrect.
+    C. If state is _vad_instance._vad_state_speaking, it indicates that the user is "speaking", and speech recognition (STT/ASR) operations can be continued. Remember: be sure to pass the returned bytes to the recognition module instead of the original audio_frame, otherwise the recognition result will be incorrect.
+# Note: 
+  If the vad module is used and it is expected to use the vad module for speech recognition (STT/ASR) and other operations, then be sure to pass the returned bytes to the recognition module instead of the original audio_frame, otherwise the recognition result will be incorrect.
+# How to better troubleshoot VAD issues: It includes two aspects, configuration and debugging.
+  1. Ensure that the initialization parameters of the vad module are correct. Reference: voice_detection.py.
+  2. In state, bytes = on_playback_audio_frame_before_mixing(audio_frame):
+
+    - A . Save the data of audio_frame to a local file, reference: example_audio_pcm_send.py. This is to record the original audio data. For example, it can be named: source_{time.time()*1000}.pcm
+    - B.Save the result of each vad processing:
+
+      - a When state == start_speaking: create a new binary file, for example, named: vad_{time.time()*1000}.pcm, and write bytes to the file.
+      - b When state == speaking: write bytes to the file.
+      - c When state == stop_speaking: write bytes to the file and close the file.
+    Note: In this way, problems can be troubleshot based on the original audio file and the audio file processed by vad. This function can be disabled in the production environment.
+### How to push the audio generated by TTS into the channel?
+  # Source code: audio_consumer.py
+  # Sample code: example_audio_consumer.py
+### How to release resources?

{agora_python_server_sdk-2.1.4 → agora_python_server_sdk-2.1.5}/agora_python_server_sdk.egg-info/SOURCES.txt

@@ -10,7 +10,6 @@ agora/rtc/audio_encoded_frame_sender.py
 agora/rtc/audio_frame_observer.py
 agora/rtc/audio_pcm_data_sender.py
 agora/rtc/audio_sessionctrl.py
-agora/rtc/audio_vad.py
 agora/rtc/local_audio_track.py
 agora/rtc/local_user.py
 agora/rtc/local_user_observer.py
@@ -32,6 +31,8 @@ agora/rtc/_ctypes_handle/_rtc_connection_observer.py
 agora/rtc/_ctypes_handle/_video_encoded_frame_observer.py
 agora/rtc/_ctypes_handle/_video_frame_observer.py
 agora/rtc/_utils/globals.py
+agora/rtc/utils/audio_consumer.py
+agora/rtc/utils/vad_dump.py
 agora_python_server_sdk.egg-info/PKG-INFO
 agora_python_server_sdk.egg-info/SOURCES.txt
 agora_python_server_sdk.egg-info/dependency_links.txt

{agora_python_server_sdk-2.1.4 → agora_python_server_sdk-2.1.5}/setup.py

@@ -45,12 +45,12 @@ class CustomInstallCommand(install):
 
 setup(
     name='agora_python_server_sdk',
-    version='2.1.4',
+    version='2.1.5',
     description='A Python SDK for Agora Server',
     long_description=open('README.md').read(),
     long_description_content_type='text/markdown',
     url='https://github.com/AgoraIO-Extensions/Agora-Python-Server-SDK',
-    packages=["agora.rtc", "agora.rtc._ctypes_handle", "agora.rtc._utils"],
+    packages=["agora.rtc", "agora.rtc._ctypes_handle", "agora.rtc._utils","agora.rtc.utils"],
     classifiers=[
         "Intended Audience :: Developers",
         'License :: OSI Approved :: MIT License',

agora_python_server_sdk-2.1.4/PKG-INFO

@@ -1,51 +0,0 @@
-Metadata-Version: 2.1
-Name: agora_python_server_sdk
-Version: 2.1.4
-Summary: A Python SDK for Agora Server
-Home-page: https://github.com/AgoraIO-Extensions/Agora-Python-Server-SDK
-Classifier: Intended Audience :: Developers
-Classifier: License :: OSI Approved :: MIT License
-Classifier: Topic :: Multimedia :: Sound/Audio
-Classifier: Topic :: Multimedia :: Video
-Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
-Classifier: Programming Language :: Python :: 3.10
-Classifier: Programming Language :: Python :: 3 :: Only
-Requires-Python: >=3.10
-Description-Content-Type: text/markdown
-
-# Note
-- This is a Python SDK wrapper for the Agora RTC SDK.
-- It supports Linux and Mac platforms.
-- The examples are provided as very simple demonstrations and are not recommended for use in production environments.
-
-# Very Important Notice !!!
-- A process can only have one instance.
-- An instance can have multiple connections.
-- In all observers or callbacks, you must not call the SDK's own APIs, nor perform CPU-intensive tasks in the callbacks; data copying is allowed.
-
-# Required Operating Systems and Python Versions
-- Supported Linux versions:
-  - Ubuntu 18.04 LTS and above
-  - CentOS 7.0 and above
-  
-- Supported Mac versions:
-  - MacOS 13 and above
-
-- Python version:
-  - Python 3.10 and above
-
-# Using Agora-Python-Server-SDK
-```
-pip install agora_python_server_sdk
-```
-
-# Running Examples
-
-## Preparing Test Data
-- Download and unzip [test_data.zip](https://download.agora.io/demo/test/test_data_202408221437.zip) to the Agora-Python-Server-SDK directory.
-
-## Executing Test Script
-```
-python agora_rtc/examples/example_audio_pcm_send.py --appId=xxx --channelId=xxx --audioFile=./test_data/demo.pcm --sampleRate=16000 --numOfChannels=1
-```
-

agora_python_server_sdk-2.1.4/README.md

@@ -1,36 +0,0 @@
-# Note
-- This is a Python SDK wrapper for the Agora RTC SDK.
-- It supports Linux and Mac platforms.
-- The examples are provided as very simple demonstrations and are not recommended for use in production environments.
-
-# Very Important Notice !!!
-- A process can only have one instance.
-- An instance can have multiple connections.
-- In all observers or callbacks, you must not call the SDK's own APIs, nor perform CPU-intensive tasks in the callbacks; data copying is allowed.
-
-# Required Operating Systems and Python Versions
-- Supported Linux versions:
-  - Ubuntu 18.04 LTS and above
-  - CentOS 7.0 and above
-  
-- Supported Mac versions:
-  - MacOS 13 and above
-
-- Python version:
-  - Python 3.10 and above
-
-# Using Agora-Python-Server-SDK
-```
-pip install agora_python_server_sdk
-```
-
-# Running Examples
-
-## Preparing Test Data
-- Download and unzip [test_data.zip](https://download.agora.io/demo/test/test_data_202408221437.zip) to the Agora-Python-Server-SDK directory.
-
-## Executing Test Script
-```
-python agora_rtc/examples/example_audio_pcm_send.py --appId=xxx --channelId=xxx --audioFile=./test_data/demo.pcm --sampleRate=16000 --numOfChannels=1
-```
-

agora_python_server_sdk-2.1.4/agora/rtc/audio_vad.py

@@ -1,164 +0,0 @@
-from . import lib_path
-import ctypes
-import os
-import sys
-from enum import Enum, IntEnum
-import logging
-logger = logging.getLogger(__name__)
-
-
-if sys.platform == 'darwin':
-    agora_vad_lib_path = os.path.join(lib_path, 'libuap_aed.dylib')
-elif sys.platform == 'linux':
-    agora_vad_lib_path = os.path.join(lib_path, 'libagora_uap_aed.so')
-try:
-    agora_vad_lib = ctypes.CDLL(agora_vad_lib_path)
-except OSError as e:
-    logger.error(f"Error loading the library: {e}")
-    logger.error(f"Attempted to load from: {agora_vad_lib_path}")
-    sys.exit(1)
-
-
-class VAD_STATE(ctypes.c_int):
-    VAD_STATE_NONE_SPEAKING = 0
-    VAD_STATE_START_SPEAKING = 1
-    VAD_STATE_SPEAKING = 2
-    VAD_STATE_STOP_SPEAKING = 3
-
-
-# struct def
-"""
-typedef struct Vad_Config_ {
-  int fftSz;  // fft-size, only support: 128, 256, 512, 1024, default value is 1024
-  int hopSz;  // fft-Hop Size, will be used to check, default value is 160
-  int anaWindowSz;  // fft-window Size, will be used to calc rms, default value is 768
-  int frqInputAvailableFlag;  // whether Aed_InputData will contain external freq. power-sepctra, default value is 0
-  int useCVersionAIModule; // whether to use the C version of AI submodules, default value is 0
-  float voiceProbThr;  // voice probability threshold 0.0f ~ 1.0f, default value is 0.8
-  float rmsThr; // rms threshold in dB, default value is -40.0
-  float jointThr; // joint threshold in dB, default value is 0.0
-  float aggressive; // aggressive factor, greater value means more aggressive, default value is 5.0
-  int startRecognizeCount; // start recognize count, buffer size for 10ms 16KHz 16bit 1channel PCM, default value is 10
-  int stopRecognizeCount; // max recognize count, buffer size for 10ms 16KHz 16bit 1channel PCM, default value is 6
-  int preStartRecognizeCount; // pre start recognize count, buffer size for 10ms 16KHz 16bit 1channel PCM, default value is 10
-  float activePercent; // active percent, if over this percent, will be recognized as speaking, default value is 0.6
-  float inactivePercent; // inactive percent, if below this percent, will be recognized as non-speaking, default value is 0.2
-} Vad_Config;
-"""
-
-
-class VadConfig(ctypes.Structure):
-    _fields_ = [
-        ("fftSz", ctypes.c_int),
-        ("hopSz", ctypes.c_int),
-        ("anaWindowSz", ctypes.c_int),
-        ("frqInputAvailableFlag", ctypes.c_int),
-        ("useCVersionAIModule", ctypes.c_int),
-        ("voiceProbThr", ctypes.c_float),
-        ("rmsThr", ctypes.c_float),
-        ("jointThr", ctypes.c_float),
-        ("aggressive", ctypes.c_float),
-        ("startRecognizeCount", ctypes.c_int),
-        ("stopRecognizeCount", ctypes.c_int),
-        ("preStartRecognizeCount", ctypes.c_int),
-        ("activePercent", ctypes.c_float),
-        ("inactivePercent", ctypes.c_float)
-    ]
-
-    def __init__(self) -> None:
-        self.fftSz = 1024
-        self.hopSz = 160
-        self.anaWindowSz = 768
-        self.frqInputAvailableFlag = 0
-        self.useCVersionAIModule = 0
-        self.voiceProbThr = 0.8
-        self.rmsThr = -40.0
-        self.jointThr = 0.0
-        self.aggressive = 5.0
-        self.startRecognizeCount = 10
-        self.stopRecognizeCount = 6
-        self.preStartRecognizeCount = 10
-        self.activePercent = 0.6
-        self.inactivePercent = 0.2
-
-
-# struct def
-class VadAudioData(ctypes.Structure):
-    _fields_ = [
-        ("audioData", ctypes.c_void_p),
-        ("size", ctypes.c_int)
-    ]
-    # def __init__(self) -> None:
-    # self.data = None
-
-
-"""
-int Agora_UAP_VAD_Create(void** stPtr, const Vad_Config* config);
-int Agora_UAP_VAD_Destroy(void** stPtr);
-int Agora_UAP_VAD_Proc(void* stPtr, const Vad_AudioData* pIn, Vad_AudioData* pOut, VAD_STATE* state);
-"""
-agora_uap_vad_create = agora_vad_lib.Agora_UAP_VAD_Create
-agora_uap_vad_create.restype = ctypes.c_int
-agora_uap_vad_create.argtypes = [ctypes.POINTER(ctypes.c_void_p), ctypes.POINTER(VadConfig)]
-
-agora_uap_vad_destroy = agora_vad_lib.Agora_UAP_VAD_Destroy
-agora_uap_vad_destroy.restype = ctypes.c_int
-agora_uap_vad_destroy.argtypes = [ctypes.POINTER(ctypes.c_void_p)]
-
-agora_uap_vad_proc = agora_vad_lib.Agora_UAP_VAD_Proc
-agora_uap_vad_proc.restype = ctypes.c_int
-agora_uap_vad_proc.argtypes = [ctypes.c_void_p, ctypes.POINTER(VadAudioData), ctypes.POINTER(VadAudioData), ctypes.POINTER(VAD_STATE)]
-
-
-class AudioVad:
-    def __init__(self) -> None:
-        self.vadCfg = VadConfig()
-
-        self.handler = None
-        self.lastOutTs = 0
-        self.initialized = False
-    # return 0 if success， -1 if failed
-
-    def Create(self, vadCfg):
-        if self.initialized:
-            return 0
-        self.vadCfg = vadCfg
-        self.initialized = True
-        # creat handler
-        self.handler = ctypes.c_void_p()
-        ret = agora_uap_vad_create(ctypes.byref(self.handler), ctypes.byref(self.vadCfg))
-        return ret
-
-    # Destroy
-    # return 0 if success， -1 if failed
-    def Destroy(self):
-        if self.initialized:
-            agora_uap_vad_destroy(ctypes.byref(self.handler))
-        self.initialized = False
-        self.handler = None
-        return 0
-
-    # Proc
-    # framein: bytearray object, include audio data
-    # return ret, frameout, flag, ret: 0 if success， -1 if failed; frameout: bytearray object, include audio data; flag: 0 if non-speaking, 1 if speaking
-    def Proc(self, framein):
-        ret = -1
-        if not self.initialized:
-            return -1
-
-        # supporse vadout is empty,vadin byte array
-        inVadData = VadAudioData()
-        buffer = (ctypes.c_ubyte * len(framein)).from_buffer(framein)  # only a pointer to the buffer is needed, not a copy
-        inVadData.audioData = ctypes.cast(buffer, ctypes.c_void_p)
-        inVadData.size = len(framein)
-
-        outVadData = VadAudioData(None, 0)  # c api will allocate memory
-        vadflag = VAD_STATE(0)
-        ret = agora_uap_vad_proc(self.handler, ctypes.byref(inVadData), ctypes.byref(outVadData), ctypes.byref(vadflag))
-
-        # convert from c_char to bytearray
-        bytes_from_c = ctypes.string_at(outVadData.audioData, outVadData.size)
-        frameout = bytearray(bytes_from_c)
-        flag = vadflag.value
-
-        return ret, frameout, flag

agora_python_server_sdk-2.1.4/agora_python_server_sdk.egg-info/PKG-INFO

@@ -1,51 +0,0 @@
-Metadata-Version: 2.1
-Name: agora_python_server_sdk
-Version: 2.1.4
-Summary: A Python SDK for Agora Server
-Home-page: https://github.com/AgoraIO-Extensions/Agora-Python-Server-SDK
-Classifier: Intended Audience :: Developers
-Classifier: License :: OSI Approved :: MIT License
-Classifier: Topic :: Multimedia :: Sound/Audio
-Classifier: Topic :: Multimedia :: Video
-Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
-Classifier: Programming Language :: Python :: 3.10
-Classifier: Programming Language :: Python :: 3 :: Only
-Requires-Python: >=3.10
-Description-Content-Type: text/markdown
-
-# Note
-- This is a Python SDK wrapper for the Agora RTC SDK.
-- It supports Linux and Mac platforms.
-- The examples are provided as very simple demonstrations and are not recommended for use in production environments.
-
-# Very Important Notice !!!
-- A process can only have one instance.
-- An instance can have multiple connections.
-- In all observers or callbacks, you must not call the SDK's own APIs, nor perform CPU-intensive tasks in the callbacks; data copying is allowed.
-
-# Required Operating Systems and Python Versions
-- Supported Linux versions:
-  - Ubuntu 18.04 LTS and above
-  - CentOS 7.0 and above
-  
-- Supported Mac versions:
-  - MacOS 13 and above
-
-- Python version:
-  - Python 3.10 and above
-
-# Using Agora-Python-Server-SDK
-```
-pip install agora_python_server_sdk
-```
-
-# Running Examples
-
-## Preparing Test Data
-- Download and unzip [test_data.zip](https://download.agora.io/demo/test/test_data_202408221437.zip) to the Agora-Python-Server-SDK directory.
-
-## Executing Test Script
-```
-python agora_rtc/examples/example_audio_pcm_send.py --appId=xxx --channelId=xxx --audioFile=./test_data/demo.pcm --sampleRate=16000 --numOfChannels=1
-```
-