modeltest.py

#EPI-Talking
#Copyright (C) 2024  Maria Tavemark
#
#This program is free software: you can redistribute it and/or modify
#it under the terms of the GNU Affero General Public License as published by
#the Free Software Foundation, either version 3 of the License, or
#(at your option) any later version.
#
#This program is distributed in the hope that it will be useful,
#but WITHOUT ANY WARRANTY; without even the implied warranty of
#MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
#GNU Affero General Public License for more details.
#
#You should have received a copy of the GNU Affero General Public License
#along with this program.  If not, see <http://www.gnu.org/licenses/>.

#/Users/epi/miniforge3/bin/python3

#Global
from random import random
import sys
import select
from configobj import ConfigObj
import time
import traceback

from tts import TTS
from llm import LLM
from stt import STT
from epi import EPI

# Read our config from config.ini
config = ConfigObj("config.ini")

tts = TTS(config)
llm = LLM(config)
stt = STT(config)
epi = EPI(config)

def checkKeypress():
    while sys.stdin in select.select([sys.stdin], [], [], 0)[0]:
        ch = sys.stdin.read(1)
        if ch == 'p':
            if stt.paused:
                stt.resume()
            else:
                stt.pause()
            print("EPI is now " + ("" if not stt.paused else "not ") + "listening. Press 'p' to make EPI " + ("" if stt.paused else "not ") + "listen")
        elif ch == 'r':
            llm.clearHistory()
            print("Chat history cleared, the conversation with EPI has now started over.")
        elif ch == 'i':
            epi.restartIkaros()
        elif ch == 'q':
            raise KeyboardInterrupt


def run_stt_to_llm():
    print("System ready")

    for mood in epi.moods:
        print("EPI is " + mood)
        epi.setMood(mood)
        time.sleep(1)

    epi._nod_()
    epi._shakeHead_()
    epi.setMood("neutral")
    if config["Global"]["language"] == "sv":
        print("Saying hej!")
        tts.say("Hej!")
    else:
        print("Saying Hello!")
        tts.say("Hello!")

    while tts.isTalking():
        time.sleep(0.1)
    
    stt.resume()

    print("EPI is listening")
    print("To make EPI pause (and not listen), press 'p'")
    print("Press 'r' to reset the conversation with EPI")
    print("Press 'q' to quit")
    print("If EPI is not moving/changing colors, press 'i' to restart Ikaros")
    print("Please note that you need to press 'Enter' after each keypress for me to understand it")

    while True:
        checkKeypress()

        if not epi.ikarosRunning():
            print("Ikaros server has crashed... Trying to restart!")
            stt.pause()
            epi.startIkaros()
            stt.resume()

        if stt.paused:
            time.sleep(0.2)
            continue

        result = stt.recognize()
        #Test if we recognized any speech
        if result:
            print("You said:", result)

            #Pass to LLM
            epi.setMood("thinking")
            print("EPI is thinking....")

            answer = llm.generateAnswer(result).split()
            epi.setMood("done_thinking")

            #Pass to TTS
            stt.pause()

            print("EPI is talking")
            if llm.llm_type == "local":
                print("EPI said: ", " ".join(answer))
                mood = "neutral"
            else:
                print("EPI said: ", " ".join(answer[:-1]))
                mood = answer[-1:][0]
                answer = answer[:-1]

            happy_moods = ["glad", "happy"]
            angry_moods = ["arg", "angry"]
            sad_moods = ["ledsen", "sad"]
            neutral_moods = ["neutral", "neutral"]

            if any([True for x in happy_moods if x in mood.lower()]):
                print("EPI is happy")
                epi.setMood("happy")
            elif any([True for x in angry_moods if x in mood.lower()]):
                print("EPI is angry")
                epi.setMood("angry")
            elif any([True for x in sad_moods if x in mood.lower()]):
                print("EPI is sad")
                epi.setMood("sad")
            else:
                print("Epi had the mood", mood.lower())
                epi.setMood("neutral")
            

            no_codes = ["nej,", "nej", "nej.", "nej!", "nej?", "no,", "no", "no.", "no!", "no?"]
            yes_codes = ["ja,", "ja", "ja.", "ja!", "ja?", "yes,", "yes", "yes.", "yes!", "yes?"]
            if any([True for x in answer if x.lower() in no_codes]):
                epi.shakeHead()
            elif any([True for x in answer if x.lower() in yes_codes]):
                epi.nod()

            tts.say(" ".join(answer))

            #print("EPI is talking and flashing")
            while tts.isTalking():
                intensity = round(random()* 0.7, 1) + 0.1
                #print("Random mouth intensity: ", intensity)
                epi.controlEpi("mouth_intensity", intensity)
                time.sleep(0.1)

            epi.setMood("neutral")
            stt.resume()
            print("EPI is listening")


if __name__ == "__main__":
    try:
        for _ in range(100):
            print()
        run_stt_to_llm()
    except KeyboardInterrupt as k:
        print("Goodbye!")
    except Exception as err:
        print("An error occured during the conversation with EPI:", err)
        print(traceback.format_exc())
    finally:
        llm.shutdown()
        epi.shutdown()
        print("Goodbye!")
        exit(0)


#Problem: 

# Dålig STT på svenska, kan inte många ord..
# Lösning: Byt till openAIs modell whisper som kan alla ord på alla språk.. (kostar $0.006 /minut )

# Dålig TTS - kanske fixat?
# Lösning: Byt till Whisper (kostar 15 USD/1M tecken)

#TODO:
# Nada

#Done:
#Testa barn engelska - dålig röst just nu
#Testa vuxen engelska - dålig just nu
#     Kommentar - EPI på engelska lokalt klarar ibland av "mood"

#Testa vuxen svenska - fungerar bra.
#Test barn svenska - liiiiite creepy röst, ändra kanske röst+pitch

#Testa pyttsx3 svenska - fungerar bra oftast.
#Testa pyttsx3 engelska - fungerar bra

#Testa online LLM engelska - perfekt utöver röst.
#Testa online LLM svenska - perfekt utöver STT