chatloop.py

import queue
import sounddevice as sd
from vosk import Model, KaldiRecognizer
import json
from eliza.eliza import Eliza
import re

# --- ASR Setup ---
model = Model("./models/vosk-model-small-de-0.15")
rec = KaldiRecognizer(model, 32000)

# --- Audio Queue ---
q = queue.Queue()
audio_active = True  # Flag to control audio input

def callback(indata, frames, time, status):
    """This function is called for each audio block."""
    if status:
        print(status, flush=True)
    if audio_active:  # Only queue audio if active
        q.put(bytes(indata))

# --- Eliza Chatbot Setup ---
bot = Eliza()
bot.load("buddy.txt")

# --- TTS ---
def speak(text):
    import subprocess
    global audio_active
    
    # Pause audio input before speaking
    audio_active = False
    
    # Run TTS
    subprocess.run(["espeak-ng", "-v", "de", text])
    
    # Resume audio input after speaking
    audio_active = True

def extract_tags_and_text(text):
    # Alle Tags extrahieren
    tags = re.findall(r'<[^>]+>', text)

    # Tags sortieren (lexikographisch)
    tags_sorted = sorted(tags)

    # Zu einem String verketten (mit Leerzeichen getrennt)
    tag_string = " ".join(tags_sorted) or ""

    # Tags aus dem Text entfernen
    clean_text = re.sub(r'\s*<[^>]+>\s*', ' ', text).strip()
    clean_text = re.sub(r'\s+', ' ', clean_text)

    return tag_string, clean_text

# --- Loop ---

def output(response):
    print("Eliza: " + response)
    speak(response)

with sd.RawInputStream(samplerate=32000, blocksize=8000, dtype='int16', channels=1, callback=callback):
    response = bot.initial()
    output(response)

    context = ""
    while True:
        data = q.get()
        if rec.AcceptWaveform(data):
            result = rec.Result()
            user_input = json.loads(result).get("text", "")
            if user_input:
                print("You: " + user_input)
                input_combined = " ".join([context, user_input]).strip()
                print("[Input combined: " + input_combined + "]")
                response = bot.respond(input_combined)
                if response is None:
                    response = bot.final()
                    output(response)
                    break

                print("[Raw response: " + response + "]")
                context, response = extract_tags_and_text(response)

                output(response)