File size: 10,126 Bytes
ca8512a 36436fd ca8512a 0f3e652 ca8512a 0f3e652 ca8512a 0f3e652 ca8512a 0f3e652 ca8512a | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 | import os
import time
import yaml
import json
from decouple import config as decouple_config # if still using .env for other values
from groq import Groq
# import sounddevice as sd
import soundfile as sf
import numpy as np
import noisereduce as nr
# import keyboard
import queue
import sys
import datetime
# from pydub import AudioSegment
import os
import tempfile
class AudioProcessor:
def __init__(self):
"""
Initialize the AudioProcessor by loading configuration and setting up the Groq client.
"""
# Compute the absolute path to the config file located at <project_root>/config/config.yml.
base_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
config_path = os.path.join(base_dir, "config", "config.yml")
# Load the configuration from the YAML file
with open(config_path, "r") as file:
self.config_data = yaml.safe_load(file)
# Access Groq settings - Get API key from Hugging Face secrets
self.groq_api_key = os.getenv("GROQ_API_KEY") # From environment variables
self.groq_model = self.config_data["groq"]["model"] # Keep model name in config
# Validate configuration
if not self.groq_api_key:
raise ValueError("Groq API key not found in environment variables. "
"Set GROQ_API_KEY in Hugging Face secrets.")
if not self.groq_model:
raise ValueError("Groq model not configured in config.yml")
# Initialize the Groq client
self.groq_client = Groq(api_key=self.groq_api_key)
# Store default audio settings from the config.
self.sample_rate = self.config_data.get("audio", {}).get("sample_rate", 44100)
self.channels = self.config_data.get("audio", {}).get("channels", 1)
self.default_duration = self.config_data.get("audio", {}).get("duration", 5)
self.default_output_file = self.config_data.get("audio", {}).get("output_file", "denoised_output.wav")
# Setup logging folder (assuming logs folder is at project_root/logs)
base_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
self.logs_folder = os.path.join(base_dir, "logs")
os.makedirs(self.logs_folder, exist_ok=True)
self.log_file = os.path.join(self.logs_folder, "conversation_logs.json")
def wait_for_key(self, target_key):
"""
Wait for the target key to be pressed. If 's' is pressed, exit the program.
"""
while True:
if keyboard.is_pressed('s'):
print("Stop key 's' pressed. Exiting program.")
sys.exit(0)
if keyboard.is_pressed(target_key):
# Debounce: wait until key is released.
while keyboard.is_pressed(target_key):
time.sleep(0.1)
break
time.sleep(0.1)
def record_and_denoise(self, fs=None, channels=None, filename=None):
"""
Waits for the user to press 'q' to start recording and again 'q' to stop.
Also checks if 's' is pressed to exit immediately.
Then it performs noise reduction and saves the cleaned audio.
Parameters:
fs (int): Sampling rate (e.g., 44100 Hz). Defaults to configuration value.
channels (int): Number of audio channels (1 for mono, 2 for stereo). Defaults to configuration value.
filename (str): Output filename for the denoised audio. Defaults to configuration value.
Returns:
np.ndarray: The denoised audio signal.
int: The sampling rate.
"""
fs = fs if fs is not None else self.sample_rate
channels = channels if channels is not None else self.channels
filename = filename if filename is not None else self.default_output_file
print("Press 'q' to start recording (or 's' to exit).")
self.wait_for_key('q')
print("Recording started. Press 'q' again to stop (or 's' to exit).")
# Set up a queue to collect recorded audio data.
audio_queue = queue.Queue()
def callback(indata, frames, time_info, status):
if status:
print(status)
audio_queue.put(indata.copy())
# Start the input stream.
stream = sd.InputStream(samplerate=fs, channels=channels, callback=callback)
stream.start()
# Instead of a single blocking wait, check in a loop.
while True:
if keyboard.is_pressed('s'):
print("Stop key 's' pressed during recording. Exiting program.")
stream.stop()
sys.exit(0)
if keyboard.is_pressed('q'):
# Debounce the key press.
while keyboard.is_pressed('q'):
time.sleep(0.1)
break
time.sleep(0.1)
stream.stop()
# Combine all audio chunks from the queue.
audio_chunks = []
while not audio_queue.empty():
audio_chunks.append(audio_queue.get())
if audio_chunks:
audio = np.concatenate(audio_chunks, axis=0)
else:
print("No audio was recorded.")
return None, fs
# If mono, remove the extra dimension.
if channels == 1 and audio.ndim > 1:
audio = np.squeeze(audio)
print("Recording stopped. Reducing background noise...")
noise_duration = 0.5
noise_samples = int(noise_duration * fs)
noise_sample = audio if audio.shape[0] < noise_samples else audio[:noise_samples]
# Perform noise reduction.
denoised_audio = nr.reduce_noise(y=audio, sr=fs, y_noise=noise_sample, prop_decrease=1.0)
# Save the denoised audio to a file.
sf.write(filename, denoised_audio, fs)
print(f"Denoised audio saved as '{filename}'")
return denoised_audio, fs
def prepare_audio(self, file_path):
"""
Check if the provided audio file is in WAV format.
If not, convert it to WAV and return the path to the converted file.
"""
ext = os.path.splitext(file_path)[1].lower()
if ext != ".wav":
try:
audio = AudioSegment.from_file(file_path)
# Create a temporary file to hold the WAV data.
temp_wav_fd, temp_wav_path = tempfile.mkstemp(suffix=".wav")
os.close(temp_wav_fd) # Close the file descriptor.
audio.export(temp_wav_path, format="wav")
print(f"Converted {file_path} to WAV format at {temp_wav_path}")
return temp_wav_path
except Exception as e:
print(f"Error converting audio file: {e}")
return file_path
return file_path
def transcribe_audio(self, file_path, language="en", prompt=None, temperature=0.0):
"""
Transcribe an audio file using the Groq API with the specified Whisper model.
This function now accepts any voice input format by converting non-WAV files to WAV.
Parameters:
file_path (str): Path to the audio file.
language (str): Language code (default "en").
prompt (str): Optional transcription prompt.
temperature (float): Temperature parameter for transcription (default 0.0).
Returns:
str: The transcribed text.
"""
# Ensure the audio is in WAV format.
prepared_file = self.prepare_audio(file_path)
with open(prepared_file, "rb") as f:
audio_bytes = f.read()
transcription = self.groq_client.audio.transcriptions.create(
file=(os.path.basename(prepared_file), audio_bytes),
model=self.groq_model,
prompt=prompt,
response_format="json",
language=language,
temperature=temperature
)
# If a temporary file was created, remove it.
if prepared_file != file_path:
try:
os.remove(prepared_file)
except Exception as e:
print(f"Error removing temporary file: {e}")
return transcription.text
def log_conversation(self, user_text, bot_text=""):
"""
Append a conversation entry to the JSON log file.
Each entry has the following format:
{
"timestamp": "YYYY-MM-DD HH:MM:SS",
"User": { "text": "User's transcribed text" },
"bot": { "text": "Bot's reply text" }
}
Parameters:
user_text (str): The text from the user (transcribed).
bot_text (str): The bot's reply text (empty until LLM integration).
"""
entry = {
"timestamp": datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
"User": {"text": user_text},
"bot": {"text": bot_text}
}
# Load existing logs if the file exists, else start with an empty list.
if os.path.exists(self.log_file):
try:
with open(self.log_file, "r", encoding="utf-8") as f:
logs = json.load(f)
except json.JSONDecodeError:
logs = []
else:
logs = []
logs.append(entry)
# Save the updated logs.
with open(self.log_file, "w", encoding="utf-8") as f:
json.dump(logs, f, indent=4)
print("Conversation logged.")
# Example usage:
if __name__ == "__main__":
processor = AudioProcessor()
denoised_audio, fs = processor.record_and_denoise()
if denoised_audio:
transcription = processor.transcribe_audio(processor.default_output_file, language="en")
print("Transcribed Text:\n", transcription)
processor.log_conversation(transcription, bot_text="")
|