tst

bot.py

@@ -122,7 +122,6 @@ class MusicFileList(object):
 
 
 # *=========================================== Predefines
-formatter = logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")
 MASTER_TIMEOUT = datetime.now()
 INITIAL_TIME_WAIT = 500
 MUZYKA: Music_Config = {"ctx": None, "queue": [], "requester": []}
@@ -243,6 +242,7 @@ handler = handlers.RotatingFileHandler(
     maxBytes=6 * 1024 * 1024,
     backupCount=6,
 )
+formatter = logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")
 handler.setFormatter(formatter)
 logger.addHandler(handler)
 random.seed()

requirements_bot.txt

@@ -15,4 +15,5 @@ flask
 waitress
 clickupython
 assemblyai[extras]
+SpeechRecognition
 git+https://github.com/imayhaveborkedit/discord-ext-voice-recv

voice_recognition.py

@@ -1,13 +1,10 @@
-# Start by making sure the `assemblyai` package is installed.
-# If not, you can install it by running the following command:
-# pip install -U assemblyai
-#
-# Note: Some macOS users may need to use `pip3` instead of `pip`.
-
 import assemblyai as aai
 import discord
 from discord.ext import commands, voice_recv
 import logging
+from discord.opus import Decoder as OpusDecoder
+import wave
+
 # Replace with your API key
 aai.settings.api_key = "aa9962f0088a449a9c4ab2361e96cc08"
 discord.opus._load_default()
@@ -20,28 +17,55 @@ FILE_URL = "https://github.com/AssemblyAI-Community/audio-examples/raw/main/2023
 
 logger = logging.getLogger("discord")
 
-'''
-class SRAudioSinkHammerVersion(AudioSink):
-    def __init__(
-    self,
-    *,
-    process_cb: Optional[SRProcessDataCB] = None,
-    text_cb: Optional[SRTextCB] = None,
-    default_recognizer: SRRecognizerMethod = 'google',
-    phrase_time_limit: int = 10,
-    ignore_silence_packets: bool = True,
-    ):
-    super().__init__(None)
-    self.process_cb: Optional[SRProcessDataCB] = process_cb
-    self.text_cb: Optional[SRTextCB] = text_cb
-    self.phrase_time_limmit: int = phrase_time_limit
-    self.ignore_silence_packets: bool = ignore_silence_packets
 
-    self.default_recognizer: SRRecognizerMethod = default_recognizer
-    self._stream_data: defaultdict[int, _StreamData] = defaultdict(
-        lambda: _StreamData(stopper=None, recognizer=sr.Recognizer(), buffer=array.array('B'))
-    )
-'''
+class SRBuffer(voice_recv.AudioSink):
+    """Endpoint AudioSink that generates a wav file.
+    Best used in conjunction with a silence generating sink. (TBD)
+    """
+
+    CHANNELS = OpusDecoder.CHANNELS
+    SAMPLE_WIDTH = OpusDecoder.SAMPLE_SIZE // OpusDecoder.CHANNELS
+    SAMPLING_RATE = OpusDecoder.SAMPLING_RATE
+
+    ##on member join - 5 plików buforowych na człowieka otwarty jest tylko aktualny i następny
+    #on member disconnect - dropujemy go
+
+
+    #        event: BasicSinkWriteCB,
+    #        rtcp_event: Optional[BasicSinkWriteRTCPCB] = None,
+    #        self.cb = event
+    #        self.cb_rtcp = rtcp_event
+    #   def write(self, user: Optional[User], data: VoiceData) -> None:
+    #       self.cb(user, data)
+    #   @AudioSink.listener()
+    #   def on_rtcp_packet(self, packet: RTCPPacket, guild: discord.Guild) -> None:
+    #       self.cb_rtcp(packet) if self.cb_rtcp else None
+
+
+
+    def __init__(self, destination: wave._File):
+        super().__init__()
+
+        self._file: wave.Wave_write = wave.open(destination, 'wb')
+        self._file.setnchannels(self.CHANNELS)
+        self._file.setsampwidth(self.SAMPLE_WIDTH)
+        self._file.setframerate(self.SAMPLING_RATE)
+        self.user_list = []
+
+    def wants_opus(self) -> bool:
+        return False
+
+    def write(self, user, data) -> None:
+        logger.info("writing to file %s", self._file)
+        logger.info("user: %s", user)
+        self._file.writeframes(data.pcm)
+
+    def cleanup(self) -> None:
+        try:
+            self._file.close()
+        except Exception:
+            logger.warning("WaveSink got error closing file on cleanup", exc_info=True)
+
 
 
 class Transcriber(commands.Cog):
@@ -51,10 +75,7 @@ class Transcriber(commands.Cog):
 
         self.bot = bot
         self._last_member = None
-        bsink = voice_recv.BasicSink(callback)
-        wsink = voice_recv.WaveSink(destination="/home/pi/Conjurer/wav.wav")
-        fsink = voice_recv.FFmpegSink(filename = ".home/pi/Conjurer/mp3.mp3")
-        self.sink_list = [bsink, wsink, fsink]
+        self.wsink = voice_recv.SRBuffer(destination="/home/pi/Conjurer/wav.wav")
 
     async def transcribe():
         config = aai.TranscriptionConfig(speaker_labels=True)
@@ -71,8 +92,7 @@ class Transcriber(commands.Cog):
         logger.info("Attempt transcribe")
         vc = await ctx.author.voice.channel.connect(cls=voice_recv.VoiceRecvClient)
         logger.info("Connected")
-        msink = voice_recv.MultiAudioSink(destinations= self.sink_list)
-        vc.listen(msink)
+        vc.listen(self.wsink)
 
 
     @commands.command(name="stop_transcribe")
@@ -80,4 +100,8 @@ class Transcriber(commands.Cog):
         await ctx.voice_client.disconnect()
 
 async def setup(bot):
-    await bot.add_cog(Transcriber(bot))
+    await bot.add_cog(Transcriber(bot))
+
+
+#1. zapisuj kwestie człowieka do pliku w którym będzie można stwierdzić kto co powiedział (callback z basicaudio + write z wavesinka). Zamykaj plik i wysyłaj do transkrypcji w momencie ciszy dłuższej niż 0.5s. Jeśli człowiek nadaje cały czas dawaj sygnał że nie można go transkrybować - i wyłaczaj zapis. 
+#2. Transkrypcja to abstrakt - w zależności od tego która metoda jest włączona wysyła do odpowiedniego silnika, silniki offline odpalam na activcomie. Zaczynamy od assemblyai bo najlepiej wspiera polski. Potem zobaczymy co dalej.