Fixed service. Added some instructions

README.md

@@ -9,7 +9,9 @@ I just made some cosmetical changes to it and swapped the funky python voice lib
 
 This will be a part of the pet project I'm working on - [to build an ultimate entertainment / desktop replacement machine out of cheap ARM SBC](https://orange-pi-4-lts.blogspot.com/p/todo.html). So this "voice assistant" will be running as background daemon with the system startup.
 
-## Installation on Debian Linux
+But it doesn't mean you need to have the same device as me - it could run on any Linux machine. 
+
+## Installation (Debian Bullseye 11)
 
 
 1. Install all dependancies
@@ -35,35 +37,78 @@ ln -s $(pwd)/sayit /usr/local/bin/sayit
 
 4. Make sure you have /usr/local/bin in your $PATH, if not - change the profile
 
-5. Create a service, that will be executed for multi-user.target (look inside .service file, you might need to update your path)
+5. Test the voice assistant by running it stand alone:
+
+```
+python3 main.py
+```
+
+Give it 10 or so seconds to warm up, look after logs it's printing and then say out loud "Orange, who is Donald Trump?"
 
+6. Create a service, that will be executed for multi-user.target.
+Before doing that, look inside voiceassistant.service file. 
+You might need to update the path to main.py script, if you installed it to different than mine location.
+Also make sure that xxx in "User=xxx" is actually the user, who is running PulseAudio process. To get that to know run 
+```
+ps -ef | grep -i pulseaudio
+```
+In my case PulseAudio is running under pi, so the service will be also runing under pi
 
+Install the service:
 ```
-sudo cp $(pwd)/voiceassistant.service /lib/systemd/system/voiceassistant.service 
+sudo cp ./voiceassistant.service /lib/systemd/system/voiceassistant.service 
 sudo systemctl daemon-reload 
 sudo systemctl enable voiceassistant.service 
 sudo systemctl start voiceassistant.service 
 sudo systemctl status voiceassistant.service
 ```
 
-## Uninstall
+## Uninstall instructions
 
 ```
 sudo systemctl stop voiceassistant.service
 sudo systemctl disable voiceassistant.service
 rm /lib/systemd/system/voiceassistant.service 
 systemctl daemon-reload
+rm /usr/local/bin/sayit
 ```
 
-This step doesn't work yet. Run  it manually:
-python3 main.py
+## Usage notes / troubleshooting
+
+### Where to see logs
+
+When the voice assistant is installed as service, logs will be collected by journald, and you can see them like this:
+```
+journalctl -x
+```
+... then press [SHIFT] + [F]
+
+If you executed the assistant from the command line, logs will be printed there.
+
+### Audio issues
+
+I only tested it on my Orange PI 4 LTS which is having a built-in microphone.
+pyAudio / SpeechRecognition puthon packages did a great job detecting it and using it as the default input device. But you might be less lucky than me.
+If you have troubles with the built-in microphone or if your SBC doesn't have it, the usual suggesting would be to buy a decent USB microphone 
+instead and make sure it will be detected by pyAudio / SpeechRecognition, by running it as stand-alone (not as deamon)
+
+First make sure you can run shell **sayit** script under the non-privileged user and you hear the sound:
+```
+/usr/local/bin/sayit "This is the voice of Google. Testing, testing, one, two, three"
+```
+
+Also make sure you can run sayit with using sudo, but under the same user, who's owning PulseAudio (in my case it's **pi**):
+
+```
+# first become a root with clean env
+sudo su -
+# then attempt to run sayit under your original user
+sudo -u pi sayit "Say it again"
+```
+I had to export XDG_RUNTIME_DIR in that script, otherwise all audio playback tools using PulseAudio didn't know which one to use. 
 
-## TODO:
 
-0. Fix bug with service
-  - where the heck is log written to?
-  - hearing doesn't work
-  - playback doesn't work
+## Project TODO
 
 1. Multi-language support
    https://gtts.readthedocs.io/en/latest/module.html#languages-gtts-lang

main.py

@@ -6,22 +6,32 @@
 import pyjokes
 import subprocess
 
-listener = sr.Recognizer()
-engine = pyttsx3.init()
-voices = engine.getProperty('voices')
-engine.setProperty('voice', voices[0].id)
+## configuration ##
+last_text = 'I did not say anything'
+activation_word = 'orange'
+phraseRepeat = ['repeat', 'say it again','pardon me']
+phraseJoke = ['tell me a joke','say something funny','make me smile']
+phraseExit = ['stop listening','terminate yourself','exit']
+###################
 
+listener = sr.Recognizer()
 
-def talk_orig(text):
-    engine.say(text)
-    engine.runAndWait()
+## disabled this as we're using external shell script
+# engine = pyttsx3.init()
+# voices = engine.getProperty('voices')
+# engine.setProperty('voice', voices[0].id)
+# 
+# def talk_orig(text):
+#     engine.say(text)
+#     engine.runAndWait()
 
 
 def talk(text):
     if not text:
         print("Warning: no text was given to talk function")
         return 
     print("Talking this text: "+text)
+    last_text = text;
     process = subprocess.Popen('/usr/local/bin/sayit "%s"' % str(text), shell=True)
     process.wait()
     print("Return code is: " + str(process.returncode))
@@ -62,11 +72,12 @@ def run_alexa():
             info = wikipedia.summary(person, 1)
         except:
             info = "I have no idea"
-        print(info)
         talk(info)
-    elif 'joke' in command:
+    elif any(x in command for x in phraseRepeat):
+        talk(last_text)
+    elif any(x in command for x in phraseJoke):
         talk(pyjokes.get_joke())
-    elif 'stop listening' in command:
+    elif any(x in command for x in phraseExit):
         talk("Bye-bye!")
         quit()
     else:

sayit

@@ -5,11 +5,13 @@ then
   exit 0
 fi
 
-# implementation with using ffplay 
+export XDG_RUNTIME_DIR=/run/user/1000
+
+# implementation with using ffplay
 #tfile="$(mktemp /tmp/tts.XXXXXXXXX.mp3)" || exit 1
 #~pi/.local/bin/gtts-cli "$1" --output $tfile
 #ffplay -nodisp -autoexit -loglevel quiet $tfile
 #rm $tfile
 
-# implementation with using SoX
-~/.local/bin/gtts-cli "$1" | play -t mp3 -
+# implementation with using SoX play
+~pi/.local/bin/gtts-cli "$1" | play -t mp3 -