Issue with plugin code?

[AIWintermuteAI]

I wrote a following plugin for rapiro voice control. Unforunately it doesn't seem to work.

#!/usr/bin/env python
# -*- coding: utf-8 -*-
import os

# Plugin for Rapiro speech recognition
def run(readable_results, data, rawbuf):
    if ('rapiro_activate'  in readable_results):
		os.environ['ISLISTENING'] = '1'
                execute('echo \"#PR000G255B000T010\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')
    if ('rapiro_sleep'  in readable_results):
		os.environ['ISLISTENING'] = '0'
                execute('echo \"#PR000G000B255T010\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')
    if ('rapiro_shutdown'  in readable_results):
		os.environ['ISLISTENING'] = '0'
		execute('echo \"#Z\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')
                execute('echo \"#PR000G000B000T010000\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')
		execute('sudo shutdown now')

		
    if (os.environ['ISLISTENING'] == '1'):
        if ('stop_action' in readable_results):
                execute('echo \"#M0\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')

        if ('go_forward' in readable_results):
                execute('echo \"#M1\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')

        if ('go_back' in readable_results):
                execute('echo \"#M2\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')
				
        if ('go_left' in readable_results):
                execute('echo \"#M4\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')

        if ('go_right' in readable_results):
                execute('echo \"#M3\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')	

        if ('wave_hand' in readable_results):
                execute('echo \"#M5\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')	
       
        if ('action_six' in readable_results):
                execute('echo \"#M6\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')

        if ('action_seven' in readable_results):
                execute('echo \"#M7\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')
				
        if ('wave_goodbye' in readable_results):
                execute('echo \"#M8\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')

        if ('action_nine' in readable_results):
                execute('echo \"#M9\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')	
		
				
def execute(command):
    command_line = command
    os.system(command_line)

I launch it with the following script

import os
import subprocess

os.environ['ISLISTENING'] = '0'
os.system('echo \"#PR000G000B255T010\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0')
subprocess.Popen('python sopare.py -l')

Sopare freezes right after start. If I use ./sopare.py -v -t test for training very often the following error starts appearing "[Errno -9988] Stream closed"

Does somebody else having similiar issues?
Before when i was using fewer commands(3-4) and no environmental variables everything was working fine.

[bishoph]

What does freezes mean? Do you get any error messages? Why do you launch SOPARE via a subprocess and not directly as you can do all the startup stuff within the plugin?

Beside all the questions: executing a process in a plugin should be avoided. Instead, start a thread and launch the process decoupled from the plugin.

[AIWintermuteAI]

No error messages, just doesn't produce any output after launching.
I probably don't understand well enough how the sopare plugins work. My understanding was that the content of the plugin executes after something was recognized in the sound stream.
This is why I created separate python script to be executed by cron on boot - it changes the color of the RGB LED in robot's eyes to Blue(means finished boot procedure), creates environmental variable and spawns sopare process(which inherits that environmental variable).
If I launch sopare with cron, will the plugin script get executed immediately after sopare was launched?

By process in the plugin you mean executing a shell command? It used to work quite good before, when there were less if-conditions. What is a better way?

[bishoph]

I recommend to start and use SOPARE directly to see output like warnings and error messages until you got a working system.

A SOPARE plugin is part of the analysis thread. Each plugin gets initialized shortly after SOPARE starts. And it's called when something gets recognized. As it's part of SOPARE each and every slowdown/hickup/wait-time pulls down the SOPARE main thread. And invoking an external thread is time consuming and can pile up and therefore should be handled by a separate thread or invoked via a queue or whatever you can imagine to decouple it.

Some interesting background information is served here: https://www.bishoph.org/sopare-architecture-and-plugins/

[AIWintermuteAI]

It worked great with a simple example

def run(readable_results, data, rawbuf):
    if ('rapiro_activate'  in readable_results):
            command_line = 'echo \"#M0\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0'
            os.system(command_line)
    if ('rapiro_sleep'  in readable_results):
            command_line = 'echo \"#M1\" | sudo minicom -b 57600 -o -D /dev/ttyAMA0'
            os.system(command_line)

Then I moved the os.system to a separate function for convenience and added the environmental variable ISLISTENING. The code is in the first post.

I had a look at all sopare documentation before starting to work :)
Actually that article was the reason I wrote on_boot.py script.

It says here that plugin gets activated only when sound level is above threshold. Which is not suitable for me for two reasons:

1. I need a script that is executed when SOPARE starts listening(meaning Raspberry Pi boot is finished and robot is ready to take commands)
2. I cannot declare environmental variable in the plugin, since plugin gets called every time when sound is above threshold. I need that environmental variable declared once and then can be changed in the plugin code.
   Admittedly there are other ways to preserve a state, apart from environmental variable. For example writing data to a text file. Or adding that variable to bash shell. I'll see what I can come up with.

What I am trying to achieve is following:
The robot upon recognizing "Rapiro, activate!" will switch to a "listening state", in which it will listen for different commands(go forward, turn left etc). It will switch to sleep mode when "Rapiro, sleep!" is recognized.
Do you have any suggestions how to achieve that in a thread-safely manner?

[bishoph]

The plugin gets initialized and referenced when SOPARE starts once. On top of this the "run" function in the referenced plugin is called whenever there is a match.

Maybe you find this plugin example helpful:
https://github.com/bishoph/Misc/blob/master/robotic_arm_control.py

This means you can declare what you want once and check for matches/follow up stuff in the "run" function.

And I see that there is room for optimization and a more detailed explanation...thx for pointing out.