nicer web interface

bits-and-pieces/benchmark-whisper.txt

@@ -69,3 +69,16 @@ whisper_print_timings:   encode time = 195740.19 ms / 32623.37 ms per layer
 whisper_print_timings:   decode time = 57331.04 ms / 9555.17 ms per layer
 whisper_print_timings:    total time = 259338.50 ms
 pi@mema3:~/whisper.cpp $ ./main -h
+
+Thinkpad
+
+whisper_print_timings:     load time =   422.48 ms
+whisper_print_timings:      mel time =  2901.66 ms
+whisper_print_timings:   sample time =   155.64 ms
+whisper_print_timings:   encode time = 26528.44 ms / 4421.41 ms per layer
+whisper_print_timings:   decode time = 11414.45 ms / 1902.41 ms per layer
+whisper_print_timings:    total time = 41441.81 ms
+
+
+
+

etc/mema.ini

@@ -66,8 +66,7 @@ rhasspy_main         = http://localhost:12101
 mimic3               = http://localhost:59125
 node_red             = http://localhost:1880
 
-# this only affects the template directory at the moment
-# Fixme 
+#FIXME: this only affects the template directory at the moment, start of internationalisation 
 language             = en
 
 #FIXME: some small prompts within the intent server, should be [prompts][en] to give full multilingual addressability 
@@ -88,6 +87,7 @@ nope              = I'm_sorry_Dave_I'm_afraid_I_can't_do_that
 ok_then           = OK_then
 done              = Done_and_dusted
 what_kind         = What_kind_of_pie_would_you_like?
+ok_going          = OK_here_we_go
 
 
 #FIXME: should be [literals][en] as above

install.sh

@@ -14,50 +14,29 @@ echo  '*------------------------------------------------------------------------
 echo 'making media directories'
 echo  '*-------------------------------------------------------------------------------*'
 
-mkdir -p static/media/mos
-mkdir -p static/media/tmp
-mkdir -p static/media/rec
-mkdir -p static/media/vid
-mkdir -p static/media/pic
+mkdir static/media/mos
+mkdir static/media/tmp
+mkdir static/media/rec
+mkdir static/media/vid
+mkdir static/media/pic
 
 echo  '*-------------------------------------------------------------------------------*'
 echo 'installing packages'
 echo  '*-------------------------------------------------------------------------------*'
-# may not need some of these
 apt install python3-pip
 apt install python3-numpy
 apt install python3-pycurl
 apt install python3-matplotlib python3-tk
 apt install mosquitto mosquitto-dev
 apt install sqlite3
-apt install cmake
-# makes life simpler for debugging, not in core
-apt install ack
-# this is to enable the xdg-open replacement jaro
-apt-get install guile-2.2
-# audiovisual section
-apt install ffmpeg
-apt install libcamera-apps
-apt install pulseaudio
-apt install vlc
-apt install ntp
-# docker
-# no get_docker instead sudo apt install docker.io
+# may or may not need ffmpeg for sample conversion
+apt install ffmpeg portaudio portaudio19-dev fswebcam curl python3-pyaudio guile2.0
+sudo apt install docker.io
 echo  '*-------------------------------------------------------------------------------*'
 echo 'installing mema3'
 echo 'install python3 packages, make take a while'
 echo  '*-------------------------------------------------------------------------------*'
 sudo -H pip install -r  requirements.txt
-#FIXME: voice bonnet stuff: not happy about this, does tons of 'other stuff' as well!
-# read the docs
-#wget https://raw.githubusercontent.com/adafruit/Raspberry-Pi-Installer-Scripts/master/raspi-blinka.py
-#sudo python3 raspi-blinka.py
-# cant find board
-#sudo pip3 install rpi_ws281x adafruit-circuitpython-neopixel
-#sudo python3 -m pip install --force-reinstall adafruit-blinka
-
-
-
 echo  '*-------------------------------------------------------------------------------*'
 echo 'installing intent server service'
 cp etc/systemd/intent_server.service /etc/systemd/system/
@@ -67,7 +46,7 @@ echo  '*------------------------------------------------------------------------
 echo 'trying to start containers'
 usermod -aG docker pi
 docker run -d --network host --name mema_rhasspy --restart unless-stopped -v "$HOME/.config/rhasspy/profiles:/profiles" -v "/etc/localtime:/etc/localtime:ro" --device /dev/snd:/dev/snd rhasspy/rhasspy --user-profiles /profiles --profile en
-docker run --network host -d --restart unless-stopped --name mema_mimic3 -v "${HOME}/.local/share/mycroft/mimic3:/home/mimic3/.local/share/mycroft/mimic3" 'mycroftai/mimic3'
+sudo docker run --network host -d --restart unless-stopped --name mema_mimic3 -v "${HOME}/.local/share/mycroft/mimic3:/home/mimic3/.local/share/mycroft/mimic3" 'mycroftai/mimic3'
 docker run -d --network host -v node_red_data:/data --restart unless-stopped --name mema_nodered nodered/node-red
 echo  '*-------------------------------------------------------------------------------*'
 echo 'starting intent server'

intent_server.py

@@ -2,6 +2,8 @@
 import sys
 import os, pwd
 
+import webbrowser as wb
+
 # for system health, but pycurl should be used elsewhere
 # get rid of one of these, pycurl_request is probably problematic
 
@@ -52,6 +54,7 @@
 
 #FIXME: Problems with xdg-open
 os.environ["DISPLAY"] = ":0.0"
+#xset -display :0 -dpms
 
 pi  = False 
 # no test on system name in os now, unreliable, changed from arm to aaarch    
@@ -262,12 +265,58 @@ def run_pie(number,please):
 #FIXME: Oh boy, what a problem for something v. simple!
 # make general!
 
+'''[
 def run_front_page(number,please):
 
     Call_URL = "http://localhost:8000/memories.html"
     mycmd = r'handlr open  {}'.format(Call_URL)
     subprocess.Popen(mycmd,shell = True) 
-
+'''
+
+def run_front_page(number,please):
+
+
+    mu.curl_speak(config['en_prompts']['ok_going'])    
+    #wb.open('http://localhost:8000/memories.html', new=2)
+    wb.get('/usr/bin/chromium').open('http://localhost:8000/memories.html', new=0)
+'''     
+    try:
+        Call_URL = "http://localhost:8000/memories.html"
+        mycmd = r'jaro  {}'.format(Call_URL)
+        logging.debug('in jaro command ' + mycmd)
+        mycmd = 'chromium-browser --display=:0 --kiosk --incognito --window-position=0,0 https://reelyactive.github.io/diy/pi-kiosk/'
+        #subprocess.run([mycmd, number[0] ],  stdout=subprocess.PIPE, stderr=subprocess.PIPE).stdout
+        subprocess.run(mycmd, shell=True, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
+        #subprocess.Popen(mycmd,shell = True)   
+    except subprocess.CalledProcessError as e:
+        logging.debug('in jaro command error' + mycmd)
+        raise RuntimeError("command '{}' return here with error (code {}): {}".format(e.cmd, e.returncode, e.output))
+'''        
+
+#FIXME: Probably merge this with memories fetch? Don't let these one page calls multiply?
+@app.get("/privacy.html")
+def fetch_privacy(request: Request):
+    mema_health = system_health()
+    if mema_health['wifi'] == 'dotgreen':
+        mu.curl_speak('the_wifi_network_is_connected')
+    else:
+       mu.curl_speak('the_wifi_network_is_off')
+
+    if config['main']['use_external_ai'] == 'yes':
+        mu.curl_speak('external_services_for_labelling_and_transcription_are_on')
+    else:
+        mu.curl_speak('external_services_are_off')
+
+    #FIXME: wordcloud() not here, in a cron
+    return TEMPLATES.TemplateResponse(
+        "privacy.html",
+        {"request": {}, "results" :{}, "mema_health": system_health()}
+    )
+
+
+
+
+
 
 def run_search_memories(number,please):
     print('in search memories')

kiosk.sh

@@ -0,0 +1,7 @@
+#!/bin/sh
+xset -dpms     # disable DPMS (Energy Star) features.
+xset s off     # disable screen saver
+xset s noblank # don't blank the video device
+#matchbox-window-manager -use_titlebar no &
+#unclutter &    # hide X mouse cursor unless mouse activated
+#/usr/bin/chromium --display=:0 --kiosk --incognito --window-position=0,0 http://localhost:8000/memories.html

memalib/mema_utility.py

@@ -8,6 +8,7 @@
 import re
 import datetime
 from configobj import ConfigObj
+from fastapi import WebSocket
 
 config = ConfigObj('etc/mema.ini')
 logging.basicConfig(filename=config['main']['logfile_name'], format='%(asctime)s %(message)s', encoding='utf-8', level=logging.DEBUG)
@@ -73,7 +74,7 @@ def docker_control(command,instance_name):
     return int(datetime.datetime.now().timestamp())
 
 
-#FIXME: not used current, but useful
+#FIXME: not used currently, but useful
 
 def system_health():
 
@@ -102,14 +103,33 @@ def system_health():
             mema_health[name] = 'dotred'
     c.close()
 
-    # test whether wifi is up
-    text_string = Path('/proc/net/wireless').read_text()
-    m = re.match(r"wlp2s0", text_string)
+    # test whether wifi is up, make a long line rather than use complex regex!
+    t = Path('/proc/net/wireless').read_text()
+    t = ''.join(t.splitlines())
+    m = re.search(r'wlp', t)
+    print('m is ', m, t)  
     if m:
         mema_health['wifi'] = 'dotgreen'
     else:
         mema_health['wifi'] = 'dotred'
     return mema_health
 
+
 def timer_function():  
    print("timed out \n")  
+
+
+# websocket manager, inspiration is https://gealber.com/simple-chat-app-websockets-fastapi   
+
+class ConnectionManager:
+    def __init__(self):
+        self.connections: List[WebSocket] = []
+
+    async def connect(self, websocket: WebSocket):
+        await websocket.accept()
+        self.connections.append(websocket)
+
+    async def broadcast(self, data: str):
+        for connection in self.connections:
+            await connection.send_text(data)
+

record_story.py

@@ -7,73 +7,104 @@
 import datetime
 from pathlib import Path
 import os
-import re
-import memalib.mema_utility as mu
+#from plumbum import local, FG, BG
+
 import replicate
-from configobj import ConfigObj
-import logging
+
+import configparser
+
+# convenience for separating Pi and a random laptop
+pi = False
+if os.uname()[4].startswith("arm"): 
+    pi = True 
+
+if pi:
+    import board
+    from picamera import PiCamera
+    # coloured LEDS on front of voice bonnet, for primitive feedback
+    from digitalio import DigitalInOut, Direction, Pull
+    import adafruit_dotstar
+    DOTSTAR_DATA = board.D5
+    DOTSTAR_CLOCK = board.D6
+    dots = adafruit_dotstar.DotStar(DOTSTAR_CLOCK, DOTSTAR_DATA, 3, brightness=0.2)
+
+
+# spoken prompts without going back into node red
+
+def curl_speak(phrase):
+    cl = '''curl -s --header "Content-Type: text/utf-8"   --request POST  --data '{speech}'   http://localhost:12101/api/text-to-speech > /dev/null'''.format(speech = phrase)
+    cl_array = cl.split()
+    subprocess.call(cl_array,stdout=subprocess.DEVNULL,stderr=subprocess.STDOUT)
+    return
+
 
 # main script
 def main():
+    dots = {}
+    config = configparser.ConfigParser()
+    config.read('etc/mema.ini')
 
-    config = ConfigObj('etc/mema.ini')
-    logging.basicConfig(filename=config['main']['logfile_name'], format='%(asctime)s %(message)s', encoding='utf-8', level=logging.DEBUG)
-
-    pi  = False 
-    #FIXME: mema.ini produces strings! also no test on system name now, unreliable    
-    if config['main']['pi'] == 'yes' : 
-        pi = True 
-        import board
-        # coloured LEDS on front of voice bonnet, for primitive feedback
-        from digitalio import DigitalInOut, Direction, Pull
-        import adafruit_dotstar
-        DOTSTAR_DATA = board.D5
-        DOTSTAR_CLOCK = board.D6
-        dots = adafruit_dotstar.DotStar(DOTSTAR_CLOCK, DOTSTAR_DATA, 3, brightness=0.2)
-        dots.deinit()
+    phrase = config['en_prompts']['start_record'].replace(' ','_')
+    curl_speak(phrase)
+    sleep(1)
 
-    mu.curl_speak(config['en_prompts']['start_record'])    
-    unix_time = mu.docker_control('stop', 'mema_rhasspy')
+    try:
+        subprocess.run(["docker", "stop", "mema_rhasspy"], check=True, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
+    except subprocess.CalledProcessError as e:
+        raise RuntimeError("command '{}' return with error (code {}): {}".format(e.cmd, e.returncode, e.output))
+
+    # make a file name from the current unix timestamp
+    unix_time = int(datetime.datetime.now().timestamp())
     file_name = str(unix_time) + ".wav" 
 
     file_path = config['main']['media_directory'] + "rec/" + file_name
     media_path = config['main']['media_directory_url'] + "rec/" + file_name
-
-    dots[0] = (255,0,0) if pi else None
+
+    sleep(1)
+    if pi:
+        dots[0] = (255,0,0)  # green
 
     try:
-        #FIXME: this all goes wrong because of the comma -D Hw3,0 for example: see: https://github.com/shivammathur/setup-php/issues/392
-        record_command = config['main']['record_command'] + ' ' + config['main']['audio_maximum'] + ' ' + file_path
-        #record_array = record_command.split()
-        logging.debug('record command is: ' + record_command)
-        subprocess.call(record_command, shell=True, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
+        record_command = config['main']['record_command'] + ' ' + file_path
+        record_array = record_command.split()
+        subprocess.call(record_array, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
     except subprocess.CalledProcessError as e:
         raise RuntimeError("command '{}' return here with error (code {}): {}".format(e.cmd, e.returncode, e.output))
 
-    dots[0] = (0,0,255)  if pi else None  # red
+    if pi:
+        dots[0] = (0,0,255)  # red
+
+    try:
+        subprocess.run(["docker", "start", "mema_rhasspy"], check=True, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
+    except subprocess.CalledProcessError as e:
+        raise RuntimeError("command '{}' return with error (code {}): {}".format(e.cmd, e.returncode, e.output))
+
+    sleep(5)  # give mema_rhasspy time to reload!
+
+    phrase =  config['en_prompts']['end_record'].replace(' ','_')
+    curl_speak(phrase)
+
 
-    unix_time = mu.docker_control('start', 'mema_rhasspy')
-    sleep(int(config['main']['rhasspy_reload']))  # give mema_rhasspy time to reload!
-
-    mu.curl_speak(config['en_prompts']['end_record'])
 
     # give a little feedback
-    dots[0] = (0,255,0)  if pi else None
-
-    text = config['en_literals']['unlabelled_audio']
+    if pi:
+        dots[0] = (0,255,0)  # blue
+
+    # probably in configuration later
+    text = 'no label'
 
-    #FIXME speech to text on remote server, replace with whisper.cpp cron
-    if config['main']['use_external_ai'] == 'yes' :
-        logging.debug('in external AI transcribe')
+    # speech to text on remote server
+    if config['main']['use_external_ai']:
         # select speech to text model
         model = replicate.models.get("openai/whisper")
         # format file as path object (openai needs this)
         audio_file = Path(file_path)
         result = model.predict(audio=audio_file)
         text = result['transcription'] 
-        mu.curl_speak(config['en_prompts']['end_transcription'])
-
-    mu.curl_speak(config['en_prompts']['done'])
+        phrase = config['en_prompts']['end_transcription'].replace(' ','_')
+    else:        
+        phrase = config['en_prompts']['done']
+        curl_speak(phrase)
 
     # done, feedback, stop blinking lights
     if pi:

record_video.py

@@ -14,7 +14,7 @@
 
 # main script
 def main():
-
+    dots = {}
     config = ConfigObj('etc/mema.ini')
     logging.basicConfig(filename=config['main']['logfile_name'], format='%(asctime)s %(message)s', encoding='utf-8', level=logging.DEBUG)