Speech To Text To Gemini To Speech#

# setup instructions are at the bottom

import os
import random
import subprocess
import threading
import time

import dotenv
import gpiozero  # gpiozero is good people https://gpiozero.readthedocs.io
from google.cloud import speech  # will need to go through the rigmarole of creating
                                 # a project, enabling billing, etc.
import google.generativeai as gemini  # it's actually super easy to tinker on gemini
                                      # now e.g. no google cloud setup stuff required
                                      # https://ai.google.dev/tutorials/python_quickstart

os.environ['GOOGLE_APPLICATION_CREDENTIALS'] = '...'  # Absolute path to service acct JSON file
env = dotenv.dotenv_values('.env')
gemini.configure(api_key=env['GEMINI_API_KEY'])
model = gemini.GenerativeModel('gemini-pro')

# https://projects.raspberrypi.org/en/projects/physical-computing/5
button = gpiozero.Button(2)
audio_file = 'user.flac'
led = gpiozero.RGBLED(16, 20, 21)

# https://codelabs.developers.google.com/codelabs/cloud-speech-text-python3#3
# https://cloud.google.com/speech-to-text/docs/sync-recognize
def speech_to_text():
    client = speech.SpeechClient()
    with open(audio_file, 'rb') as f:
        content = f.read()
    config = speech.RecognitionConfig(
        language_code='en',
        encoding=speech.RecognitionConfig.AudioEncoding.FLAC,
        sample_rate_hertz=48000,
        audio_channel_count=2
    )
    audio = speech.RecognitionAudio(content=content)
    response = client.recognize(config=config, audio=audio)
    transcript = ''
    # full transcript is weirdly fragmented across this iterable thing...
    for result in response.results:
        transcript += result.alternatives[0].transcript
    return transcript

def record():
    led.on()
    led.color = (1, 0, 0)  # red
    # start the audio recording
    process = subprocess.Popen(['rec', audio_file, 'rate', '48k'])
    # https://unix.stackexchange.com/a/57593
    while not stop_recording.is_set():
        time.sleep(1)
    process.terminate()
    led.off()

def thinky_blinky():
    """blink like an octopus dreaming... https://youtu.be/0vKCLJZbytU"""
    colors = [
        (0, 0, 0),
        (1, 0, 0),
        (1, 1, 0),
        (1, 0, 1),
        (0, 1, 0),
        (0, 1, 1),
        (1, 1, 1)
    ]
    led.on()
    while not stop_blinking.is_set():
        led.color = random.choice(colors)
        time.sleep(0.1)
    led.off()

while True:
    button.wait_for_press()
    stop_recording = threading.Event()
    record_thread = threading.Thread(target=record)
    record_thread.start()
    button.wait_for_release()
    stop_recording.set()
    record_thread.join()
    stop_blinking = threading.Event()
    blink_thread = threading.Thread(target=thinky_blinky)
    blink_thread.start()
    # subprocess.run(['play', '-v', '3.0', audio_file])
    text = speech_to_text()
    print(text)
    response = model.generate_content(text)
    print(response.text)
    stop_blinking.set()
    blink_thread.join()
    led.on()
    led.color = (0, 0.2, 1)  # blueish
    # yes, i'm aware that there are now uncanny valley TTS services,
    # i like my old school friendly robot voice thank you very much
    p = subprocess.Popen(['spd-say', '--wait', '--volume', '+100', f'"{response.text}"'])
    while p.poll() is None:
        time.sleep(1)
    led.off()

"""
##### SETUP #####
----------------------------------------------------------
using the rinkydink sparkfun usb microphone dingle dongle and speaker:
https://www.sparkfun.com/products/18488
https://www.sparkfun.com/products/18343
don't think the mic or speaker required any setup...
just plugged them in and Stuff Just Worked...

--------------------------------------------
create .env and put Gemini API key there:
GEMINI_API_KEY=...

----------------------------------------------------
create requirements.txt and put these deps in there:
gpiozero==2.0
RPi.GPIO==0.7.1
google-cloud-speech==2.24.1
google-generativeai==0.3.2
python-dotenv==1.0.1

------------------------------------------
install the above deps into a virtual env:
python3 -m venv venv
source venv/bin/activate  # remember to do this before running the app too
python3 -m pip install -r requirements.txt

-------------------------------
install these libs system-wide:
sudo apt install -y sox  # audio recording command (`rec`)
sudo apt install -y speech-dispatcher  # text-to-speech command (`spd-say`)
wget http://abyz.me.uk/lg/lg.zip   # https://abyz.me.uk/lg/download.html
unzip lg.zip
cd lg  # low-level GPIO lib used by gpiozero (I think)
make
sudo make install
"""