picovoice configuration

wake-word-detect.py

@@ -0,0 +1,197 @@
+from openai import OpenAI
+from dotenv import load_dotenv
+import sounddevice as sd
+import struct
+import pyaudio
+import numpy as np
+import tempfile
+import pvporcupine
+import wavio
+import wave
+import os
+import time
+from pvrecorder import PvRecorder
+
+
+load_dotenv()
+# Retrieve the OpenAI API key and Porcupine access key from environment variables
+openai_api_key = os.getenv("OPENAI_API_KEY")
+porcupine_access_key = os.getenv("PORCUPINE_ACCESS_KEY")
+
+sd.default.device = 'seeed-2mic-voicecard'
+
+if not openai_api_key:
+    raise ValueError("OpenAI API key is not set in environment variables.")
+if not porcupine_access_key:
+    raise ValueError("Porcupine access key is not set in environment variables.")
+
+# Initialize OpenAI client
+client = OpenAI(api_key=openai_api_key, default_headers={"OpenAI-Beta": "assistants=v2"})
+
+# Initialize Porcupine
+porcupine = pvporcupine.create(
+    access_key=porcupine_access_key,
+    keywords=["picovoice", "bumblebee"]
+)
+
+# paud = pyaudio.PyAudio()
+# audio_frame = paud.open(rate=porcupine.sample_rate, channels=1, format=pyaudio.paInt16, input=True, frames_per_buffer=porcupine.frame_length)
+
+def get_next_audio_frame():
+    """
+    Record a chunk of audio from the microphone.
+    """
+
+
+
+    # return sd.rec(int(porcupine.frame_length), samplerate=porcupine.sample_rate, channels=1, dtype='int16')
+
+def query_and_record(prompt, mp3_filename):
+    """
+    Send a prompt to the OpenAI assistant and record the response as an MP3 file.
+    """
+    # Create an assistant instance
+    assistant = client.beta.assistants.create(
+        name="Senior Tech Help",
+        instructions="You are a helpful tech teacher specifically for seniors. You will help older adults (ages 50+) with quick questions about smartphones, voice assistants, computers, cameras, the internet, digital shopping, or any other technology-related topic. You will always ask for specifics, like what device or phone they are using, and provide them with step-by-step instructions for their response.",
+        model="gpt-4o"
+    )
+
+    # Create a thread for communication
+    thread = client.beta.threads.create()
+
+    # Send user's prompt to the AI
+    message = client.beta.threads.messages.create(
+        thread_id=thread.id,
+        role="user",
+        content=prompt
+    )
+
+    # Start the AI to process the user prompt
+    run = client.beta.threads.runs.create(
+        thread_id=thread.id,
+        assistant_id=assistant.id,
+        instructions="Please address the user as Jane Doe. The user has a premium account."
+    )
+
+    # Wait until AI is complete with processing
+    while run.status in ["in_progress", "queued"]:
+        time.sleep(1)
+        run = client.beta.threads.runs.retrieve(
+            thread_id=thread.id,
+            run_id=run.id
+        )
+
+    if run.status == "completed":
+        message_list = client.beta.threads.messages.list(
+            thread_id=thread.id
+        )
+
+        # Extract the text from the response
+        text_response = message_list.data[-1].content
+
+        # Generate an audio response from the text
+        response = client.audio.speech.create(
+            model="tts-1-hd",
+            voice="echo",
+            input=text_response,
+        )
+
+        response.stream_to_file(mp3_filename)
+
+        print("Response recorded to " + mp3_filename)
+
+def is_silent(file, threshold=500):
+    """
+    Returns 'True' if below the 'silent' threshold.
+    """
+    return np.abs(file).mean() < threshold
+
+def record_audio(samplerate=44100, chunk_duration=1, silence_threshold=500, min_chunks=5):
+    """
+    Record audio from the microphone until silence is detected.
+    """
+    print("Recording... Press Ctrl+C to stop.")
+    audio_file = []
+
+    try:
+        while True:
+            recording = sd.rec(int(chunk_duration * samplerate), samplerate=samplerate, channels=1, dtype='int16')
+            sd.wait()
+            audio_file.append(recording)
+            # Ensure minimum recording length before checking for silence
+            if len(audio_file) >= min_chunks and is_silent(recording, threshold=silence_threshold):
+                print("Silence detected, stopping recording.")
+                break
+    except KeyboardInterrupt:
+        print("Recording stopped manually.")
+
+    # Concatenate all recorded chunks
+    if audio_file:
+        audio_file = np.concatenate(audio_file, axis=0)
+        return audio_file
+    else:
+        raise ValueError("No audio file recorded.")
+
+# Main loop for keyword detection and interaction
+recorder = PvRecorder(
+    frame_length=porcupine.frame_length,
+    device_index=3
+)
+recorder.start()
+wav_file = None
+wav_file = wave.open("./test.wav", "w")
+wav_file.setnchannels(1)
+wav_file.setsampwidth(2)
+wav_file.setframerate(16000)
+print('Listening ... ')
+
+try:
+    while True:
+        # keyword = audio_frame.read(porcupine.frame_length)
+        # keyword = struct.unpack_from ("h" * porcupine.frame_length, keyword)
+        # keyword_index= porcupine.process(keyword)
+
+
+
+
+
+        pcm = recorder.read()
+        keyword_index = porcupine.process(pcm)
+
+        if wav_file is not None:
+            wav_file.writeframes(struct.pack("h" * len(pcm), *pcm))
+
+        print(keyword_index)
+        if keyword_index == 0:
+            print("Detected 'picovoice'")
+        elif keyword_index == 1:
+            recorder.delete()
+            print("Detected 'bumblebee'")
+
+            # Record audio from the microphone
+            audio_file = record_audio()
+
+            # Convert audio to text using OpenAI API
+            with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as tmpfile:
+                tmpfilename = tmpfile.name
+                wavio.write(tmpfilename, audio_file, 44100, sampwidth=2)
+
+            transcription = client.audio.transcriptions.create(
+                model="whisper-1",
+                file=open(tmpfilename, "rb"),
+            )
+
+            print("Transcription:", transcription.text)
+
+            # Example usage
+            prompt = transcription.text
+            mp3_filename = "response.mp3"
+            query_and_record(prompt, mp3_filename)
+
+finally:
+# Ensuring proper release of resources
+    recorder.delete()
+    porcupine.delete()
+    if wav_file is not None:
+        wav_file.close()