Spaces:
Paused
Paused
rick
commited on
align to PEP8
Browse files
app.py
CHANGED
|
@@ -29,15 +29,13 @@ def load_ui_language(file_path: Optional[str] = "ui_lang_support.json") -> dict:
|
|
| 29 |
with open(file_path, 'r', encoding='utf-8') as file:
|
| 30 |
return json.load(file)
|
| 31 |
except FileNotFoundError:
|
| 32 |
-
print(get_translation(
|
| 33 |
return {}
|
| 34 |
except json.JSONDecodeError:
|
| 35 |
-
print(get_translation(
|
| 36 |
-
"JSON decoding error"
|
| 37 |
-
))
|
| 38 |
return {}
|
| 39 |
-
except
|
| 40 |
-
print(get_translation(
|
| 41 |
return {}
|
| 42 |
|
| 43 |
# Dictionary to store translations
|
|
@@ -73,9 +71,9 @@ def read_file(file_name):
|
|
| 73 |
content = file.read()
|
| 74 |
return content
|
| 75 |
except FileNotFoundError:
|
| 76 |
-
return get_translation(
|
| 77 |
-
except
|
| 78 |
-
return get_translation(
|
| 79 |
|
| 80 |
def split_audio(audio_file, max_size_mb=25):
|
| 81 |
"""
|
|
@@ -88,22 +86,29 @@ def split_audio(audio_file, max_size_mb=25):
|
|
| 88 |
Returns:
|
| 89 |
list: List of paths to the split audio segments.
|
| 90 |
"""
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
segments = []
|
| 99 |
-
for start in range(0, duration_ms, segment_duration_ms):
|
| 100 |
-
end = min(start + segment_duration_ms, duration_ms)
|
| 101 |
-
segment = audio[start:end]
|
| 102 |
-
with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_segment:
|
| 103 |
-
segment.export(temp_segment.name, format="wav")
|
| 104 |
-
segments.append(temp_segment.name)
|
| 105 |
|
| 106 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 107 |
|
| 108 |
# Fonction modifiée pour transcrire l'audio en texte
|
| 109 |
def transcribe_audio(audio_file, language=None):
|
|
@@ -120,63 +125,82 @@ def transcribe_audio(audio_file, language=None):
|
|
| 120 |
max_size_mb = 25
|
| 121 |
file_size_mb = os.path.getsize(audio_file.name) / (1024 * 1024)
|
| 122 |
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 128 |
transcript = client.audio.transcriptions.create(
|
| 129 |
model="whisper-1",
|
| 130 |
-
file=
|
| 131 |
language=language
|
| 132 |
)
|
| 133 |
-
|
| 134 |
-
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
file=audio_file,
|
| 141 |
-
language=language
|
| 142 |
-
)
|
| 143 |
-
return transcript.text
|
| 144 |
|
| 145 |
# Fonction pour détecter la langue d'un texte donné
|
| 146 |
def detect_language(input_text, temperature=0.01):
|
| 147 |
"""
|
| 148 |
-
|
| 149 |
|
| 150 |
Args:
|
| 151 |
-
input_text (str):
|
| 152 |
-
temperature (float, optional):
|
| 153 |
|
| 154 |
Returns:
|
| 155 |
-
str:
|
|
|
|
|
|
|
|
|
|
|
|
|
| 156 |
"""
|
| 157 |
-
system_prompt =
|
| 158 |
-
"
|
| 159 |
-
"
|
| 160 |
-
"
|
| 161 |
-
"
|
| 162 |
-
"
|
| 163 |
-
"
|
| 164 |
-
])
|
| 165 |
-
response = client.chat.completions.create(
|
| 166 |
-
model="gpt-4o-mini",
|
| 167 |
-
temperature=temperature,
|
| 168 |
-
messages=[
|
| 169 |
-
{
|
| 170 |
-
"role": "system",
|
| 171 |
-
"content": system_prompt
|
| 172 |
-
},
|
| 173 |
-
{
|
| 174 |
-
"role": "user",
|
| 175 |
-
"content": f"{input_text}"
|
| 176 |
-
}
|
| 177 |
-
]
|
| 178 |
)
|
| 179 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 180 |
|
| 181 |
def get_duration_pydub(audio_file):
|
| 182 |
"""
|
|
@@ -188,8 +212,15 @@ def get_duration_pydub(audio_file):
|
|
| 188 |
Returns:
|
| 189 |
float: Duration of the audio file in seconds.
|
| 190 |
"""
|
| 191 |
-
|
| 192 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 193 |
|
| 194 |
def text_to_speech(text):
|
| 195 |
"""
|
|
@@ -201,60 +232,71 @@ def text_to_speech(text):
|
|
| 201 |
Returns:
|
| 202 |
tuple: A tuple containing the audio bytes and the duration of the audio in seconds.
|
| 203 |
"""
|
| 204 |
-
|
| 205 |
-
|
| 206 |
-
|
| 207 |
-
|
| 208 |
-
|
| 209 |
-
|
| 210 |
-
# Save the audio to a temporary file
|
| 211 |
-
with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as temp_audio:
|
| 212 |
-
response.stream_to_file(temp_audio.name)
|
| 213 |
|
| 214 |
-
#
|
| 215 |
-
with
|
| 216 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 217 |
|
| 218 |
-
|
| 219 |
-
|
| 220 |
|
| 221 |
-
|
|
|
|
|
|
|
|
|
|
| 222 |
|
| 223 |
def concatenate_audio_files(audio_list):
|
| 224 |
"""
|
| 225 |
-
|
| 226 |
|
| 227 |
Args:
|
| 228 |
-
audio_list (list):
|
| 229 |
|
| 230 |
Returns:
|
| 231 |
-
bytes:
|
| 232 |
"""
|
| 233 |
-
#
|
| 234 |
final_audio = AudioSegment.empty()
|
| 235 |
|
| 236 |
-
|
| 237 |
-
|
| 238 |
-
|
| 239 |
-
|
| 240 |
-
|
| 241 |
-
|
| 242 |
-
|
| 243 |
-
|
| 244 |
-
# 5 seconds of silence
|
| 245 |
-
silence = AudioSegment.silent(duration=5000) # 5000 ms = 5 seconds
|
| 246 |
-
|
| 247 |
-
for audio_bytes, _ in audio_list:
|
| 248 |
-
# Convert bytes to an audio segment
|
| 249 |
-
segment = AudioSegment.from_mp3(io.BytesIO(audio_bytes))
|
| 250 |
|
| 251 |
-
#
|
| 252 |
-
|
| 253 |
-
|
| 254 |
-
|
| 255 |
-
|
| 256 |
-
|
| 257 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 258 |
|
| 259 |
def process_message(message, operation_prompt="", tts_enabled=False):
|
| 260 |
"""
|
|
@@ -268,7 +310,7 @@ def process_message(message, operation_prompt="", tts_enabled=False):
|
|
| 268 |
Returns:
|
| 269 |
tuple: A tuple containing the TTS audio and its duration, or (None, None) if TTS is disabled.
|
| 270 |
"""
|
| 271 |
-
payload_content =
|
| 272 |
|
| 273 |
st.session_state.messages.append({"role": "user", "content": payload_content})
|
| 274 |
with st.chat_message("user"):
|
|
@@ -277,24 +319,32 @@ def process_message(message, operation_prompt="", tts_enabled=False):
|
|
| 277 |
with st.chat_message("assistant"):
|
| 278 |
message_placeholder = st.empty()
|
| 279 |
full_response = ""
|
| 280 |
-
|
| 281 |
-
|
| 282 |
-
|
| 283 |
-
|
| 284 |
-
|
| 285 |
-
|
| 286 |
-
|
| 287 |
-
|
| 288 |
-
|
| 289 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 290 |
|
| 291 |
st.session_state.messages.append(
|
| 292 |
{"role": "assistant", "content": full_response}
|
| 293 |
)
|
| 294 |
|
| 295 |
if tts_enabled:
|
| 296 |
-
|
| 297 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 298 |
return None, None
|
| 299 |
|
| 300 |
class GlobalSystemPrompts:
|
|
@@ -308,8 +358,15 @@ class GlobalSystemPrompts:
|
|
| 308 |
Returns:
|
| 309 |
str: The system prompt for Linguascribe.
|
| 310 |
"""
|
| 311 |
-
|
| 312 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 313 |
|
| 314 |
# Function to configure the translation mode
|
| 315 |
def set_translation_mode(from_lang, dest_lang):
|
|
@@ -321,7 +378,7 @@ def set_translation_mode(from_lang, dest_lang):
|
|
| 321 |
dest_lang (str): The destination language.
|
| 322 |
"""
|
| 323 |
system_prompt = GlobalSystemPrompts.linguascribe()
|
| 324 |
-
operation_prompt =
|
| 325 |
return system_prompt, operation_prompt
|
| 326 |
|
| 327 |
# List of languages supported by the application
|
|
@@ -337,17 +394,17 @@ SUPPORTED_LANGUAGES = [
|
|
| 337 |
"Turkish", "Ukrainian", "Urdu", "Vietnamese", "Welsh"
|
| 338 |
]
|
| 339 |
|
| 340 |
-
def convert_language_name_to_iso6391(language_data):
|
| 341 |
"""
|
| 342 |
-
|
| 343 |
|
| 344 |
Args:
|
| 345 |
-
language_data (Union[str, dict]):
|
| 346 |
|
| 347 |
Returns:
|
| 348 |
-
str:
|
| 349 |
"""
|
| 350 |
-
#
|
| 351 |
language_to_iso = {
|
| 352 |
"Afrikaans": "af", "Arabic": "ar", "Armenian": "hy", "Azerbaijani": "az",
|
| 353 |
"Belarusian": "be", "Bosnian": "bs", "Bulgarian": "bg", "Catalan": "ca",
|
|
@@ -366,21 +423,26 @@ def convert_language_name_to_iso6391(language_data):
|
|
| 366 |
"Welsh": "cy"
|
| 367 |
}
|
| 368 |
|
| 369 |
-
#
|
| 370 |
if isinstance(language_data, dict):
|
| 371 |
language_name = language_data.get('language')
|
| 372 |
else:
|
| 373 |
language_name = language_data
|
| 374 |
|
| 375 |
-
|
| 376 |
-
|
| 377 |
-
|
| 378 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 379 |
def main():
|
| 380 |
-
"""
|
| 381 |
st.title("------- DEMORRHA -------")
|
| 382 |
|
| 383 |
-
#
|
| 384 |
if "language_detected" not in st.session_state:
|
| 385 |
st.session_state["language_detected"] = None
|
| 386 |
|
|
@@ -402,10 +464,10 @@ def main():
|
|
| 402 |
st.session_state["enable_tts_for_input_from_audio_record"] = True
|
| 403 |
|
| 404 |
if "interface_language" not in st.session_state:
|
| 405 |
-
st.session_state.interface_language = "French" #
|
| 406 |
|
| 407 |
def init_process_mode():
|
| 408 |
-
"""
|
| 409 |
if "translation" == st.session_state["process_mode"]:
|
| 410 |
system_prompt, operation_prompt = set_translation_mode(
|
| 411 |
from_lang=st.session_state.language_detected,
|
|
@@ -416,17 +478,17 @@ def main():
|
|
| 416 |
|
| 417 |
system_prompt, operation_prompt = init_process_mode()
|
| 418 |
|
| 419 |
-
#
|
| 420 |
if "messages" not in st.session_state:
|
| 421 |
st.session_state.messages = []
|
| 422 |
|
| 423 |
-
#
|
| 424 |
if not any(message["role"] == "system" for message in st.session_state.messages):
|
| 425 |
st.session_state.messages.insert(0, {"role": "system", "content": system_prompt})
|
| 426 |
|
| 427 |
-
#
|
| 428 |
if user_input := st.chat_input(get_translation("entrez_message")):
|
| 429 |
-
#
|
| 430 |
if st.session_state.language_detected is None:
|
| 431 |
st.session_state.language_detected = detect_language(
|
| 432 |
input_text=user_input, temperature=0.01
|
|
@@ -436,24 +498,30 @@ def main():
|
|
| 436 |
for cursor_selected_lang in st.session_state.selected_languages:
|
| 437 |
st.session_state.target_language = cursor_selected_lang["iso-639-1"]
|
| 438 |
|
| 439 |
-
#
|
| 440 |
system_prompt, operation_prompt = init_process_mode()
|
| 441 |
|
| 442 |
-
#
|
| 443 |
-
|
| 444 |
-
|
| 445 |
-
|
| 446 |
-
|
| 447 |
-
|
| 448 |
-
|
| 449 |
-
|
|
|
|
|
|
|
|
|
|
| 450 |
|
| 451 |
if audio_list:
|
| 452 |
-
|
| 453 |
-
|
|
|
|
|
|
|
|
|
|
| 454 |
|
| 455 |
with st.container(border=True):
|
| 456 |
-
#
|
| 457 |
st.write(get_translation("enregistrez_message"))
|
| 458 |
audio = audiorecorder(
|
| 459 |
start_prompt=get_translation("cliquez_enregistrer"),
|
|
@@ -463,47 +531,56 @@ def main():
|
|
| 463 |
key="vocal_chat_input"
|
| 464 |
)
|
| 465 |
|
| 466 |
-
#
|
| 467 |
if len(audio) > 0:
|
| 468 |
-
|
| 469 |
-
|
| 470 |
-
|
| 471 |
-
|
| 472 |
-
|
| 473 |
-
st.session_state.language_detected
|
| 474 |
-
|
| 475 |
-
|
| 476 |
-
|
| 477 |
-
|
| 478 |
-
|
| 479 |
-
|
| 480 |
-
|
| 481 |
-
|
| 482 |
-
st.session_state.
|
| 483 |
-
|
| 484 |
-
|
| 485 |
-
|
| 486 |
-
|
| 487 |
-
|
| 488 |
-
|
| 489 |
-
|
| 490 |
-
|
| 491 |
-
|
| 492 |
-
|
| 493 |
-
|
| 494 |
-
|
| 495 |
-
|
| 496 |
-
|
| 497 |
-
|
| 498 |
-
|
| 499 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 500 |
with st.sidebar:
|
| 501 |
st.header(get_translation("sidebar_titre"))
|
| 502 |
-
st.markdown("##
|
| 503 |
st.info(get_translation("info_app"))
|
| 504 |
|
| 505 |
def on_languages_change():
|
| 506 |
-
"""
|
| 507 |
selected_language_names = st.session_state.language_selector
|
| 508 |
st.session_state.selected_languages = [
|
| 509 |
{"language": lang, "iso-639-1": convert_language_name_to_iso6391(lang)}
|
|
@@ -512,7 +589,7 @@ def main():
|
|
| 512 |
|
| 513 |
with st.container(border=True):
|
| 514 |
st.subheader(get_translation("langue_interface"))
|
| 515 |
-
#
|
| 516 |
st.selectbox(
|
| 517 |
label=get_translation("choix_langue_interface"),
|
| 518 |
options=list(translations.keys()),
|
|
@@ -525,10 +602,10 @@ def main():
|
|
| 525 |
)
|
| 526 |
|
| 527 |
with st.container(border=True):
|
| 528 |
-
#
|
| 529 |
st.subheader(get_translation("selection_langue"))
|
| 530 |
|
| 531 |
-
#
|
| 532 |
st.multiselect(
|
| 533 |
label=get_translation("langues_destination"),
|
| 534 |
placeholder=get_translation("placeholder_langues"),
|
|
@@ -544,7 +621,7 @@ def main():
|
|
| 544 |
st.selectbox(
|
| 545 |
get_translation("choix_voix_tts"),
|
| 546 |
options=["alloy", "echo", "fable", "onyx", "nova", "shimmer"],
|
| 547 |
-
index=3, # "onyx"
|
| 548 |
key="tts_voice"
|
| 549 |
)
|
| 550 |
st.checkbox(
|
|
@@ -556,6 +633,6 @@ def main():
|
|
| 556 |
key="enable_tts_for_input_from_audio_record"
|
| 557 |
)
|
| 558 |
|
| 559 |
-
#
|
| 560 |
if __name__ == "__main__":
|
| 561 |
main()
|
|
|
|
| 29 |
with open(file_path, 'r', encoding='utf-8') as file:
|
| 30 |
return json.load(file)
|
| 31 |
except FileNotFoundError:
|
| 32 |
+
print(f"{get_translation('erreur_fichier_non_trouve')} {file_path}")
|
| 33 |
return {}
|
| 34 |
except json.JSONDecodeError:
|
| 35 |
+
print(f"{get_translation('erreur_lecture_fichier')} JSON decoding error")
|
|
|
|
|
|
|
| 36 |
return {}
|
| 37 |
+
except IOError as e:
|
| 38 |
+
print(f"{get_translation('erreur_lecture_fichier')} {e}")
|
| 39 |
return {}
|
| 40 |
|
| 41 |
# Dictionary to store translations
|
|
|
|
| 71 |
content = file.read()
|
| 72 |
return content
|
| 73 |
except FileNotFoundError:
|
| 74 |
+
return f"{get_translation('erreur_fichier_non_trouve')} {file_name}"
|
| 75 |
+
except IOError as e:
|
| 76 |
+
return f"{get_translation('erreur_lecture_fichier')} {str(e)}"
|
| 77 |
|
| 78 |
def split_audio(audio_file, max_size_mb=25):
|
| 79 |
"""
|
|
|
|
| 86 |
Returns:
|
| 87 |
list: List of paths to the split audio segments.
|
| 88 |
"""
|
| 89 |
+
try:
|
| 90 |
+
audio = AudioSegment.from_wav(audio_file)
|
| 91 |
+
duration_ms = len(audio)
|
| 92 |
+
segment_duration_ms = int(
|
| 93 |
+
(max_size_mb * 1024 * 1024 * 8) /
|
| 94 |
+
(audio.frame_rate * audio.sample_width * audio.channels)
|
| 95 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 96 |
|
| 97 |
+
segments = []
|
| 98 |
+
for start in range(0, duration_ms, segment_duration_ms):
|
| 99 |
+
end = min(start + segment_duration_ms, duration_ms)
|
| 100 |
+
segment = audio[start:end]
|
| 101 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_segment:
|
| 102 |
+
segment.export(temp_segment.name, format="wav")
|
| 103 |
+
segments.append(temp_segment.name)
|
| 104 |
+
|
| 105 |
+
return segments
|
| 106 |
+
except IOError as e:
|
| 107 |
+
print(f"Erreur lors de la lecture ou de l'écriture du fichier audio : {e}")
|
| 108 |
+
return []
|
| 109 |
+
except ValueError as e:
|
| 110 |
+
print(f"Erreur de valeur lors du traitement de l'audio : {e}")
|
| 111 |
+
return []
|
| 112 |
|
| 113 |
# Fonction modifiée pour transcrire l'audio en texte
|
| 114 |
def transcribe_audio(audio_file, language=None):
|
|
|
|
| 125 |
max_size_mb = 25
|
| 126 |
file_size_mb = os.path.getsize(audio_file.name) / (1024 * 1024)
|
| 127 |
|
| 128 |
+
try:
|
| 129 |
+
if file_size_mb > max_size_mb:
|
| 130 |
+
segments = split_audio(audio_file.name, max_size_mb)
|
| 131 |
+
full_transcript = ""
|
| 132 |
+
for segment in segments:
|
| 133 |
+
with open(segment, "rb") as audio_segment:
|
| 134 |
+
transcript = client.audio.transcriptions.create(
|
| 135 |
+
model="whisper-1",
|
| 136 |
+
file=audio_segment,
|
| 137 |
+
language=language
|
| 138 |
+
)
|
| 139 |
+
full_transcript += f"{transcript.text} "
|
| 140 |
+
os.unlink(segment) # Delete temporary file
|
| 141 |
+
return full_transcript.strip()
|
| 142 |
+
else:
|
| 143 |
+
with open(audio_file.name, "rb") as audio_file:
|
| 144 |
transcript = client.audio.transcriptions.create(
|
| 145 |
model="whisper-1",
|
| 146 |
+
file=audio_file,
|
| 147 |
language=language
|
| 148 |
)
|
| 149 |
+
return transcript.text
|
| 150 |
+
except IOError as e:
|
| 151 |
+
print(f"Erreur d'entrée/sortie lors de la transcription : {e}")
|
| 152 |
+
return ""
|
| 153 |
+
except client.APIError as e:
|
| 154 |
+
print(f"Erreur API lors de la transcription : {e}")
|
| 155 |
+
return ""
|
|
|
|
|
|
|
|
|
|
|
|
|
| 156 |
|
| 157 |
# Fonction pour détecter la langue d'un texte donné
|
| 158 |
def detect_language(input_text, temperature=0.01):
|
| 159 |
"""
|
| 160 |
+
Détecte la langue d'un texte donné.
|
| 161 |
|
| 162 |
Args:
|
| 163 |
+
input_text (str): Le texte dont il faut détecter la langue.
|
| 164 |
+
temperature (float, optional): La température pour le modèle de langage. Par défaut à 0.01.
|
| 165 |
|
| 166 |
Returns:
|
| 167 |
+
str: La langue détectée au format ISO-639-1.
|
| 168 |
+
|
| 169 |
+
Raises:
|
| 170 |
+
ValueError: Si la réponse de l'API est invalide.
|
| 171 |
+
requests.RequestException: En cas d'erreur de communication avec l'API.
|
| 172 |
"""
|
| 173 |
+
system_prompt = (
|
| 174 |
+
"Agissez comme une fonction de détection de langue. "
|
| 175 |
+
"Je fournirai du texte dans n'importe quelle langue, et vous détecterez sa langue. "
|
| 176 |
+
"Fournissez le résultat de votre détection au format ISO-639-1. "
|
| 177 |
+
"Votre réponse doit représenter l'argument `language` et ne contenir "
|
| 178 |
+
"que sa valeur sous forme de chaîne. "
|
| 179 |
+
"Fournir la langue d'entrée au format ISO-639-1 améliorera la précision et la latence."
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 180 |
)
|
| 181 |
+
try:
|
| 182 |
+
response = client.chat.completions.create(
|
| 183 |
+
model="gpt-4o-mini",
|
| 184 |
+
temperature=temperature,
|
| 185 |
+
messages=[
|
| 186 |
+
{
|
| 187 |
+
"role": "system",
|
| 188 |
+
"content": system_prompt
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"role": "user",
|
| 192 |
+
"content": input_text
|
| 193 |
+
}
|
| 194 |
+
]
|
| 195 |
+
)
|
| 196 |
+
detected_language = response.choices[0].message.content
|
| 197 |
+
if not detected_language:
|
| 198 |
+
raise ValueError("La réponse de l'API est vide")
|
| 199 |
+
return detected_language
|
| 200 |
+
except requests.RequestException as e:
|
| 201 |
+
raise requests.RequestException(f"Erreur de communication avec l'API : {str(e)}")
|
| 202 |
+
except Exception as e:
|
| 203 |
+
raise ValueError(f"Erreur inattendue lors de la détection de la langue : {str(e)}")
|
| 204 |
|
| 205 |
def get_duration_pydub(audio_file):
|
| 206 |
"""
|
|
|
|
| 212 |
Returns:
|
| 213 |
float: Duration of the audio file in seconds.
|
| 214 |
"""
|
| 215 |
+
try:
|
| 216 |
+
audio = AudioSegment.from_file(audio_file)
|
| 217 |
+
return audio.duration_seconds
|
| 218 |
+
except FileNotFoundError:
|
| 219 |
+
print(f"Erreur : Le fichier audio '{audio_file}' n'a pas été trouvé.")
|
| 220 |
+
return 0.0
|
| 221 |
+
except Exception as e:
|
| 222 |
+
print(f"Erreur lors de la lecture du fichier audio : {str(e)}")
|
| 223 |
+
return 0.0
|
| 224 |
|
| 225 |
def text_to_speech(text):
|
| 226 |
"""
|
|
|
|
| 232 |
Returns:
|
| 233 |
tuple: A tuple containing the audio bytes and the duration of the audio in seconds.
|
| 234 |
"""
|
| 235 |
+
try:
|
| 236 |
+
response = client.audio.speech.create(
|
| 237 |
+
model="tts-1",
|
| 238 |
+
voice=st.session_state.tts_voice,
|
| 239 |
+
input=text
|
| 240 |
+
)
|
|
|
|
|
|
|
|
|
|
| 241 |
|
| 242 |
+
# Save the audio to a temporary file
|
| 243 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as temp_audio:
|
| 244 |
+
response.stream_to_file(temp_audio.name)
|
| 245 |
+
|
| 246 |
+
# Read the content of the audio file
|
| 247 |
+
with open(temp_audio.name, "rb") as audio_file:
|
| 248 |
+
audio_bytes = audio_file.read()
|
| 249 |
|
| 250 |
+
# Get the duration of the audio in seconds
|
| 251 |
+
audio_duration = get_duration_pydub(temp_audio.name)
|
| 252 |
|
| 253 |
+
return audio_bytes, audio_duration
|
| 254 |
+
except Exception as e:
|
| 255 |
+
print(f"Erreur lors de la conversion texte-parole : {str(e)}")
|
| 256 |
+
return None, 0.0
|
| 257 |
|
| 258 |
def concatenate_audio_files(audio_list):
|
| 259 |
"""
|
| 260 |
+
Concatène plusieurs fichiers audio avec des effets sonores.
|
| 261 |
|
| 262 |
Args:
|
| 263 |
+
audio_list (list): Une liste de tuples, chacun contenant des octets audio et la durée.
|
| 264 |
|
| 265 |
Returns:
|
| 266 |
+
bytes: L'audio concaténé sous forme d'octets.
|
| 267 |
"""
|
| 268 |
+
# Créer un segment audio vide
|
| 269 |
final_audio = AudioSegment.empty()
|
| 270 |
|
| 271 |
+
try:
|
| 272 |
+
# Charger les effets sonores
|
| 273 |
+
begin_sound = AudioSegment.from_mp3(
|
| 274 |
+
"sound-effects/voice-message-play-begin/voice-message-play-begin-1.mp3"
|
| 275 |
+
)
|
| 276 |
+
end_sound = AudioSegment.from_mp3(
|
| 277 |
+
"sound-effects/voice-message-play-ending/voice-message-play-ending-1.mp3"
|
| 278 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 279 |
|
| 280 |
+
# 5 secondes de silence
|
| 281 |
+
silence = AudioSegment.silent(duration=5000) # 5000 ms = 5 secondes
|
| 282 |
+
|
| 283 |
+
for audio_bytes, _ in audio_list:
|
| 284 |
+
# Convertir les octets en un segment audio
|
| 285 |
+
segment = AudioSegment.from_mp3(io.BytesIO(audio_bytes))
|
| 286 |
+
|
| 287 |
+
# Ajouter le son de début, le segment TTS, le son de fin et le silence
|
| 288 |
+
final_audio += begin_sound + segment + end_sound + silence
|
| 289 |
+
|
| 290 |
+
# Convertir le segment audio final en octets
|
| 291 |
+
buffer = io.BytesIO()
|
| 292 |
+
final_audio.export(buffer, format="mp3")
|
| 293 |
+
return buffer.getvalue()
|
| 294 |
+
except IOError as e:
|
| 295 |
+
print(f"Erreur lors de la lecture ou de l'écriture des fichiers audio : {e}")
|
| 296 |
+
return None
|
| 297 |
+
except Exception as e:
|
| 298 |
+
print(f"Une erreur inattendue s'est produite : {e}")
|
| 299 |
+
return None
|
| 300 |
|
| 301 |
def process_message(message, operation_prompt="", tts_enabled=False):
|
| 302 |
"""
|
|
|
|
| 310 |
Returns:
|
| 311 |
tuple: A tuple containing the TTS audio and its duration, or (None, None) if TTS is disabled.
|
| 312 |
"""
|
| 313 |
+
payload_content = '{} :\n"""\n{}\n"""'.format(operation_prompt, message)
|
| 314 |
|
| 315 |
st.session_state.messages.append({"role": "user", "content": payload_content})
|
| 316 |
with st.chat_message("user"):
|
|
|
|
| 319 |
with st.chat_message("assistant"):
|
| 320 |
message_placeholder = st.empty()
|
| 321 |
full_response = ""
|
| 322 |
+
try:
|
| 323 |
+
for response in client.chat.completions.create(
|
| 324 |
+
model="gpt-4o-mini",
|
| 325 |
+
messages=st.session_state.messages,
|
| 326 |
+
stream=True,
|
| 327 |
+
temperature=0.1):
|
| 328 |
+
full_response += (response.choices[0].delta.content or "")
|
| 329 |
+
message_placeholder.markdown(full_response + "▌")
|
| 330 |
+
# Use regex to remove the first and last three double quotes
|
| 331 |
+
full_response = re.sub(r'^"{3}|"{3}$', '', full_response.strip())
|
| 332 |
+
message_placeholder.markdown(full_response)
|
| 333 |
+
except Exception as e:
|
| 334 |
+
st.error("Une erreur s'est produite lors de la génération de la réponse : {}".format(str(e)))
|
| 335 |
+
return None, None
|
| 336 |
|
| 337 |
st.session_state.messages.append(
|
| 338 |
{"role": "assistant", "content": full_response}
|
| 339 |
)
|
| 340 |
|
| 341 |
if tts_enabled:
|
| 342 |
+
try:
|
| 343 |
+
tts_audio, tts_duration = text_to_speech(full_response)
|
| 344 |
+
return tts_audio, tts_duration
|
| 345 |
+
except Exception as e:
|
| 346 |
+
st.error("Une erreur s'est produite lors de la conversion texte-parole : {}".format(str(e)))
|
| 347 |
+
return None, None
|
| 348 |
return None, None
|
| 349 |
|
| 350 |
class GlobalSystemPrompts:
|
|
|
|
| 358 |
Returns:
|
| 359 |
str: The system prompt for Linguascribe.
|
| 360 |
"""
|
| 361 |
+
try:
|
| 362 |
+
system_prompt = read_file('linguascribe.prompt')
|
| 363 |
+
return system_prompt
|
| 364 |
+
except FileNotFoundError:
|
| 365 |
+
print("Le fichier 'linguascribe.prompt' n'a pas été trouvé.")
|
| 366 |
+
return ""
|
| 367 |
+
except IOError as e:
|
| 368 |
+
print(f"Erreur lors de la lecture du fichier 'linguascribe.prompt': {e}")
|
| 369 |
+
return ""
|
| 370 |
|
| 371 |
# Function to configure the translation mode
|
| 372 |
def set_translation_mode(from_lang, dest_lang):
|
|
|
|
| 378 |
dest_lang (str): The destination language.
|
| 379 |
"""
|
| 380 |
system_prompt = GlobalSystemPrompts.linguascribe()
|
| 381 |
+
operation_prompt = "Translate({} to {})".format(from_lang, dest_lang)
|
| 382 |
return system_prompt, operation_prompt
|
| 383 |
|
| 384 |
# List of languages supported by the application
|
|
|
|
| 394 |
"Turkish", "Ukrainian", "Urdu", "Vietnamese", "Welsh"
|
| 395 |
]
|
| 396 |
|
| 397 |
+
def convert_language_name_to_iso6391(language_data: Union[str, dict]) -> str:
|
| 398 |
"""
|
| 399 |
+
Convertit un nom de langue en son code ISO 639-1.
|
| 400 |
|
| 401 |
Args:
|
| 402 |
+
language_data (Union[str, dict]): Le nom de la langue ou un dictionnaire contenant le nom de la langue.
|
| 403 |
|
| 404 |
Returns:
|
| 405 |
+
str: Le code ISO 639-1 pour la langue donnée, ou 'en' si non trouvé.
|
| 406 |
"""
|
| 407 |
+
# Dictionnaire associant les noms de langues aux codes ISO 639-1
|
| 408 |
language_to_iso = {
|
| 409 |
"Afrikaans": "af", "Arabic": "ar", "Armenian": "hy", "Azerbaijani": "az",
|
| 410 |
"Belarusian": "be", "Bosnian": "bs", "Bulgarian": "bg", "Catalan": "ca",
|
|
|
|
| 423 |
"Welsh": "cy"
|
| 424 |
}
|
| 425 |
|
| 426 |
+
# Vérifier si language_data est un dictionnaire
|
| 427 |
if isinstance(language_data, dict):
|
| 428 |
language_name = language_data.get('language')
|
| 429 |
else:
|
| 430 |
language_name = language_data
|
| 431 |
|
| 432 |
+
try:
|
| 433 |
+
# Retourner le code ISO 639-1 correspondant au nom de la langue
|
| 434 |
+
return language_to_iso[language_name]
|
| 435 |
+
except KeyError:
|
| 436 |
+
# Gérer spécifiquement l'exception KeyError
|
| 437 |
+
print(f"Langue non trouvée : {language_name}")
|
| 438 |
+
return "en" # Par défaut 'en' si la langue n'est pas trouvée
|
| 439 |
+
|
| 440 |
+
# Fonction principale de l'application
|
| 441 |
def main():
|
| 442 |
+
"""Fonction principale qui configure et exécute l'application Streamlit."""
|
| 443 |
st.title("------- DEMORRHA -------")
|
| 444 |
|
| 445 |
+
# Initialisation des variables d'état de session
|
| 446 |
if "language_detected" not in st.session_state:
|
| 447 |
st.session_state["language_detected"] = None
|
| 448 |
|
|
|
|
| 464 |
st.session_state["enable_tts_for_input_from_audio_record"] = True
|
| 465 |
|
| 466 |
if "interface_language" not in st.session_state:
|
| 467 |
+
st.session_state.interface_language = "French" # Langue par défaut
|
| 468 |
|
| 469 |
def init_process_mode():
|
| 470 |
+
"""Initialise le mode de traitement pour la traduction si nécessaire."""
|
| 471 |
if "translation" == st.session_state["process_mode"]:
|
| 472 |
system_prompt, operation_prompt = set_translation_mode(
|
| 473 |
from_lang=st.session_state.language_detected,
|
|
|
|
| 478 |
|
| 479 |
system_prompt, operation_prompt = init_process_mode()
|
| 480 |
|
| 481 |
+
# Initialisation de l'historique des messages avec le prompt système
|
| 482 |
if "messages" not in st.session_state:
|
| 483 |
st.session_state.messages = []
|
| 484 |
|
| 485 |
+
# Vérification de l'existence d'un message système dans st.session_state.messages
|
| 486 |
if not any(message["role"] == "system" for message in st.session_state.messages):
|
| 487 |
st.session_state.messages.insert(0, {"role": "system", "content": system_prompt})
|
| 488 |
|
| 489 |
+
# Interface utilisateur pour le chat textuel
|
| 490 |
if user_input := st.chat_input(get_translation("entrez_message")):
|
| 491 |
+
# Traitement du message texte de l'utilisateur
|
| 492 |
if st.session_state.language_detected is None:
|
| 493 |
st.session_state.language_detected = detect_language(
|
| 494 |
input_text=user_input, temperature=0.01
|
|
|
|
| 498 |
for cursor_selected_lang in st.session_state.selected_languages:
|
| 499 |
st.session_state.target_language = cursor_selected_lang["iso-639-1"]
|
| 500 |
|
| 501 |
+
# Initialisation du mode de traitement pour la langue cible actuelle
|
| 502 |
system_prompt, operation_prompt = init_process_mode()
|
| 503 |
|
| 504 |
+
# Traitement du message utilisateur pour la langue cible actuelle
|
| 505 |
+
try:
|
| 506 |
+
tts_audio, tts_duration = process_message(
|
| 507 |
+
user_input,
|
| 508 |
+
operation_prompt=f"{operation_prompt}",
|
| 509 |
+
tts_enabled=st.session_state.enable_tts_for_input_from_text_field
|
| 510 |
+
)
|
| 511 |
+
if tts_audio is not None:
|
| 512 |
+
audio_list.append((tts_audio, tts_duration))
|
| 513 |
+
except Exception as e:
|
| 514 |
+
st.error(f"Erreur lors du traitement du message : {str(e)}")
|
| 515 |
|
| 516 |
if audio_list:
|
| 517 |
+
try:
|
| 518 |
+
final_audio = concatenate_audio_files(audio_list)
|
| 519 |
+
st.audio(final_audio, format="audio/mp3", autoplay=True)
|
| 520 |
+
except Exception as e:
|
| 521 |
+
st.error(f"Erreur lors de la concaténation des fichiers audio : {str(e)}")
|
| 522 |
|
| 523 |
with st.container(border=True):
|
| 524 |
+
# Interface utilisateur pour l'enregistrement audio
|
| 525 |
st.write(get_translation("enregistrez_message"))
|
| 526 |
audio = audiorecorder(
|
| 527 |
start_prompt=get_translation("cliquez_enregistrer"),
|
|
|
|
| 531 |
key="vocal_chat_input"
|
| 532 |
)
|
| 533 |
|
| 534 |
+
# Traitement de l'entrée audio de l'utilisateur
|
| 535 |
if len(audio) > 0:
|
| 536 |
+
try:
|
| 537 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio:
|
| 538 |
+
audio.export(temp_audio.name, format="wav")
|
| 539 |
+
transcription = transcribe_audio(temp_audio, language=st.session_state.language_detected)
|
| 540 |
+
os.unlink(temp_audio.name) # Suppression du fichier temporaire
|
| 541 |
+
if st.session_state.language_detected is None:
|
| 542 |
+
st.session_state.language_detected = detect_language(
|
| 543 |
+
input_text=transcription, temperature=0.01
|
| 544 |
+
)
|
| 545 |
+
st.write(get_translation("langue_detectee").format(st.session_state.language_detected))
|
| 546 |
+
|
| 547 |
+
st.write(get_translation("transcription").format(transcription))
|
| 548 |
+
|
| 549 |
+
audio_list = []
|
| 550 |
+
for cursor_selected_lang in st.session_state.selected_languages:
|
| 551 |
+
st.session_state.target_language = cursor_selected_lang["iso-639-1"]
|
| 552 |
+
|
| 553 |
+
# Initialisation du mode de traitement pour la langue cible actuelle
|
| 554 |
+
system_prompt, operation_prompt = init_process_mode()
|
| 555 |
+
|
| 556 |
+
# Traitement du message utilisateur pour la langue cible actuelle
|
| 557 |
+
try:
|
| 558 |
+
tts_audio, tts_duration = process_message(
|
| 559 |
+
transcription,
|
| 560 |
+
operation_prompt=f"{operation_prompt}",
|
| 561 |
+
tts_enabled=st.session_state.enable_tts_for_input_from_audio_record
|
| 562 |
+
)
|
| 563 |
+
if tts_audio is not None:
|
| 564 |
+
audio_list.append((tts_audio, tts_duration))
|
| 565 |
+
except Exception as e:
|
| 566 |
+
st.error(f"Erreur lors du traitement du message audio : {str(e)}")
|
| 567 |
+
if audio_list:
|
| 568 |
+
try:
|
| 569 |
+
final_audio = concatenate_audio_files(audio_list)
|
| 570 |
+
st.audio(final_audio, format="audio/mp3", autoplay=True)
|
| 571 |
+
except Exception as e:
|
| 572 |
+
st.error(f"Erreur lors de la concaténation des fichiers audio : {str(e)}")
|
| 573 |
+
except Exception as e:
|
| 574 |
+
st.error(f"Erreur lors du traitement de l'audio : {str(e)}")
|
| 575 |
+
|
| 576 |
+
# Configuration de la barre latérale
|
| 577 |
with st.sidebar:
|
| 578 |
st.header(get_translation("sidebar_titre"))
|
| 579 |
+
st.markdown(f"## {get_translation('a_propos')}")
|
| 580 |
st.info(get_translation("info_app"))
|
| 581 |
|
| 582 |
def on_languages_change():
|
| 583 |
+
"""Fonction de rappel pour le changement de langue(s) de destination."""
|
| 584 |
selected_language_names = st.session_state.language_selector
|
| 585 |
st.session_state.selected_languages = [
|
| 586 |
{"language": lang, "iso-639-1": convert_language_name_to_iso6391(lang)}
|
|
|
|
| 589 |
|
| 590 |
with st.container(border=True):
|
| 591 |
st.subheader(get_translation("langue_interface"))
|
| 592 |
+
# Sélection de la langue de l'interface
|
| 593 |
st.selectbox(
|
| 594 |
label=get_translation("choix_langue_interface"),
|
| 595 |
options=list(translations.keys()),
|
|
|
|
| 602 |
)
|
| 603 |
|
| 604 |
with st.container(border=True):
|
| 605 |
+
# Conteneur pour la sélection de langue
|
| 606 |
st.subheader(get_translation("selection_langue"))
|
| 607 |
|
| 608 |
+
# Sélection multiple des langues de destination
|
| 609 |
st.multiselect(
|
| 610 |
label=get_translation("langues_destination"),
|
| 611 |
placeholder=get_translation("placeholder_langues"),
|
|
|
|
| 621 |
st.selectbox(
|
| 622 |
get_translation("choix_voix_tts"),
|
| 623 |
options=["alloy", "echo", "fable", "onyx", "nova", "shimmer"],
|
| 624 |
+
index=3, # "onyx" est à l'index 3
|
| 625 |
key="tts_voice"
|
| 626 |
)
|
| 627 |
st.checkbox(
|
|
|
|
| 633 |
key="enable_tts_for_input_from_audio_record"
|
| 634 |
)
|
| 635 |
|
| 636 |
+
# Point d'entrée de l'application
|
| 637 |
if __name__ == "__main__":
|
| 638 |
main()
|