Fix #27 and fix #17 - Transcriptions and recordings - two birds, one stone

hugvey/central_command.py

@@ -22,6 +22,7 @@ import queue
 import threading
 from hugvey.voice import VoiceStorage
 import multiprocessing
+from hugvey.speech.recorder import Recorder
 
 mainLogger = logging.getLogger("hugvey")
 
@@ -314,6 +315,7 @@ class HugveyState(object):
         self.streamer = None
         self.status = self.STATE_PAUSE
         self.google = None
+        self.recorder = None
         self.notShuttingDown = True # TODO: allow shutdown of object
         self.startMsgId = None
         self.eventLogger = eventLogger.getChild(f"{self.id}")
@@ -494,6 +496,10 @@ class HugveyState(object):
             startMsgId = self.startMsgId
             self.startMsgId = None # use only once, reset before 'run'
             self.logger.warn(f"Starting from {startMsgId}")
+            if not self.streamer:
+                await asyncio.sleep(1)
+                
+            self.streamer.triggerStart()
             self.story.setStoryData(self.command.languages[self.language_code])
             await self.story.run(startMsgId)
 #             self.story = None
@@ -511,6 +517,12 @@ class HugveyState(object):
                     self.command.config['voice']['src_rate'], self.command.config['voice']['out_rate'])
                 self.streamer.addConsumer(self.player)
             
+            if self.command.config['voice']['record_dir']:
+                self.logger.warn("Record Audio of conversation")
+                self.recorder = Recorder( self.id,
+                    self.command.config['voice']['src_rate'], self.command.config['voice']['record_dir'])
+                self.streamer.addConsumer(self.recorder)
+    
             self.logger.debug("Start Speech")
             self.google = GoogleVoiceClient(
                 hugvey=self,

hugvey/speech/google.py

@@ -210,5 +210,8 @@ class GoogleVoiceClient(object):
     def shutdown(self):
         self.toBeShutdown = True
     
+    def triggerStart(self):
+        pass
+        
     
            

hugvey/speech/player.py

@@ -61,3 +61,6 @@ class Player:
         
         self.stream.close()
         self.p.terminate()
+    
+    def triggerStart(self):
+        pass

hugvey/speech/recorder.py

@@ -0,0 +1,105 @@
+import pyaudio
+import logging
+import audioop
+import wave
+import time
+import os
+import chunk
+from _struct import pack
+from array import array
+
+mainLogger = logging.getLogger("hugvey")
+logger = mainLogger.getChild("recorder")
+
+class Recorder:
+    """
+    Record the streamed audio
+    """
+
+    def __init__(self, hv_id, src_rate, out_folder):
+        if not os.path.exists(out_folder):
+            raise Exception(f"Invalid output folder for recordings: {out_folder}")
+
+        self.hv_id = hv_id
+        self.src_rate = src_rate
+        self.main_folder = out_folder # unfortunately not every device plays 16kHz audio streams
+        self.running = False
+        
+    def start(self):
+        self.subsequentMutedFrames = 0
+        self.fragmentNr = 0
+        
+        self.data = array('h')
+        self.currentTranscription = ""
+        
+        t = time.strftime("%Y%m%d-%H%M")
+        
+        self.out_folder = os.path.join(self.main_folder, f"{self.hv_id}", t)
+        if not os.path.exists(self.out_folder):
+            logger.debug(f"Create directory {self.out_folder}")
+            self.target_folder = os.makedirs(self.out_folder, exist_ok=True)
+        
+        self.running = True
+    
+    
+    def writeData(self):
+        if len(self.data) < 1:
+            return
+            
+        self.fragmentNr += 1
+        
+        self.wf = wave.open(os.path.join(self.out_folder, f"{self.fragmentNr}.wav"), 'wb')
+        self.wf.setnchannels(1)
+        self.wf.setsampwidth(2)
+        self.wf.setframerate(self.src_rate)
+        # adapted from https://stackoverflow.com/questions/892199/detect-record-audio-in-python#6743593
+        self.wf.writeframes(pack('<' + ('h'*len(self.data)), *self.data))
+        self.wf.close()
+        
+        with open(os.path.join(self.out_folder, "transcriptions.txt"), "a") as fp:
+            fp.write(f"{self.fragmentNr}\t{self.currentTranscription}\n")
+        self.log('-',self.currentTranscription)
+        
+        self.data = array('h')
+        self.currentTranscription = ""
+    
+    def receive(self, chunk):
+        if not self.running:
+            return 
+        
+#           logger.debug('receive {}'.format(len(chunk)))
+        if audioop.max(chunk, 2) == 0:
+            # mic is muted on client side.
+            
+            self.subsequentMutedFrames += 1
+            logger.warn(f"Empty frame {self.subsequentMutedFrames}")
+#             self.logger.debug("Muted")
+            if self.subsequentMutedFrames == 4:
+                logger.warn("Create new wave!")
+#                 self.createWave()
+                self.writeData()
+            if self.subsequentMutedFrames > 4:
+#                 Don't write more muted frames to the audio file.
+                return
+        else:
+            self.subsequentMutedFrames = 0
+        
+        d = array('h', chunk)
+        self.data.extend(d)
+#         self.wf.writeframes(chunk)
+    
+    def shutdown(self):
+        self.writeData()
+        self.running = False
+#         self.wf.close()
+    
+    def triggerStart(self):
+        self.start()
+
+    def updateTranscription(self, text):
+        self.currentTranscription = text
+        
+    def log(self, origin, msg):
+        with open(os.path.join(self.out_folder, "log.txt"), "a") as fp:
+            fp.write(f"{origin}: {msg}\n")
+            

hugvey/speech/streamer.py

@@ -60,4 +60,7 @@ class AudioStreamer(object):
         for consumer in self.consumers:
             consumer.receive(chunk)
         
-    
+    def triggerStart(self):
+        # start a (new) run on the hugvey. Send it to the consumers that need it
+        for consumer in self.consumers:
+            consumer.triggerStart()

hugvey/story.py

@@ -657,7 +657,7 @@ class Story(object):
         
         # Register variables
         for msg in self.getMessages():
-            print(msg.id, msg.hasVariables())
+#             print(msg.id, msg.hasVariables())
             if not msg.hasVariables():
                 continue
             
@@ -795,6 +795,9 @@ class Story(object):
                     utterance.setFinished(self.timer.getElapsed())
                     self.hugvey.eventLogger.info("speaking: stop {}".format(id(utterance)))
                 
+                if self.hugvey.recorder:
+                    self.hugvey.recorder.updateTranscription(self.currentReply.getText())
+                    
         
     async def _processDirections(self, directions):
         ':type directions: list(Direction)'
@@ -874,6 +877,7 @@ class Story(object):
                 'id': self.currentMessage.id,
             })
             
+        
         message.uuid = shortuuid.uuid()
         self.currentMessage = message
         self.lastMsgTime = time.time()
@@ -891,6 +895,8 @@ class Story(object):
             message.id, message.text))
         self.addToLog(message)
         self.hugvey.eventLogger.info(f"message: {message.id} {message.uuid} start \"{message.text}\"")
+        if self.hugvey.recorder:
+            self.hugvey.recorder.log('h',message.text)
         
         # TODO: prep events & timer etc.
         fn = await message.getAudioFilePath()

recordings/.gitignore

@@ -0,0 +1 @@
+*