Spaces:
Building
Building
Update stt_google.py
Browse files- stt_google.py +145 -129
stt_google.py
CHANGED
@@ -110,116 +110,120 @@ class GoogleCloudSTT(STTInterface):
|
|
110 |
log_error(f"❌ Error queuing result: {e}")
|
111 |
|
112 |
def _run_stream(self):
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
def request_generator():
|
118 |
-
"""Generate streaming requests"""
|
119 |
-
chunk_count = 0
|
120 |
-
start_time = datetime.now()
|
121 |
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
-
|
129 |
-
|
130 |
-
|
131 |
-
|
132 |
-
|
133 |
-
|
134 |
-
break
|
135 |
-
chunk_count += 1
|
136 |
-
|
137 |
-
# Sadece önemli milestone'larda logla
|
138 |
-
if chunk_count == 1:
|
139 |
-
log_info(f"📤 First chunk sent to Google STT, size: {len(chunk)} bytes")
|
140 |
-
elif chunk_count % 100 == 0:
|
141 |
-
log_info(f"📤 Sent {chunk_count} chunks to Google STT (elapsed: {elapsed:.1f}s)")
|
142 |
|
143 |
-
|
144 |
-
|
145 |
-
|
146 |
-
|
147 |
-
|
148 |
-
|
149 |
-
|
150 |
-
|
151 |
-
|
152 |
-
|
153 |
-
|
154 |
-
|
155 |
-
|
156 |
-
|
157 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
158 |
|
159 |
-
|
160 |
-
response_count = 0
|
161 |
-
empty_response_count = 0
|
162 |
|
163 |
-
|
164 |
-
|
|
|
165 |
|
166 |
-
|
167 |
-
|
168 |
-
|
169 |
|
170 |
-
|
171 |
-
|
172 |
-
empty_response_count += 1
|
173 |
-
if empty_response_count % 50 == 0:
|
174 |
-
log_warning(f"⚠️ Received {empty_response_count} empty responses from Google STT")
|
175 |
-
continue
|
176 |
|
177 |
-
|
178 |
-
|
179 |
-
|
180 |
-
|
181 |
-
# İlk alternatifi al
|
182 |
-
alternative = result.alternatives[0]
|
183 |
|
184 |
-
#
|
185 |
-
if
|
186 |
-
|
187 |
-
|
188 |
-
|
189 |
-
|
190 |
-
confidence=alternative.confidence if hasattr(alternative, 'confidence') and alternative.confidence else 0.0,
|
191 |
-
timestamp=datetime.now().timestamp()
|
192 |
-
)
|
193 |
|
194 |
-
|
195 |
-
|
|
|
|
|
|
|
|
|
196 |
|
197 |
-
#
|
198 |
-
if
|
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 |
async def stream_audio(self, audio_chunk: bytes) -> AsyncIterator[TranscriptionResult]:
|
225 |
"""Stream audio chunk and get transcription results"""
|
@@ -246,39 +250,51 @@ class GoogleCloudSTT(STTInterface):
|
|
246 |
raise
|
247 |
|
248 |
async def stop_streaming(self) -> Optional[TranscriptionResult]:
|
249 |
-
|
250 |
-
|
251 |
-
|
252 |
-
|
253 |
-
try:
|
254 |
-
log_info("🛑 Stopping Google STT streaming...")
|
255 |
-
|
256 |
-
self.is_streaming = False
|
257 |
-
self.stop_event.set()
|
258 |
-
|
259 |
-
# Send poison pill to queue
|
260 |
-
self.audio_queue.put(None)
|
261 |
-
|
262 |
-
# Wait for thread to finish
|
263 |
-
if self.stream_thread:
|
264 |
-
self.stream_thread.join(timeout=5.0)
|
265 |
-
|
266 |
-
# Clear queues
|
267 |
-
while not self.audio_queue.empty():
|
268 |
-
self.audio_queue.get_nowait()
|
269 |
|
270 |
-
|
271 |
-
|
272 |
-
|
273 |
-
|
274 |
-
|
275 |
-
|
276 |
-
|
277 |
-
|
278 |
-
|
279 |
-
|
280 |
-
|
281 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
282 |
|
283 |
def supports_realtime(self) -> bool:
|
284 |
"""Google Cloud STT supports real-time streaming"""
|
|
|
110 |
log_error(f"❌ Error queuing result: {e}")
|
111 |
|
112 |
def _run_stream(self):
|
113 |
+
"""Run the streaming recognition in a separate thread"""
|
114 |
+
try:
|
115 |
+
log_info("🎤 Google STT stream thread started")
|
|
|
|
|
|
|
|
|
|
|
116 |
|
117 |
+
def request_generator():
|
118 |
+
"""Generate streaming requests"""
|
119 |
+
chunk_count = 0
|
120 |
+
start_time = datetime.now()
|
121 |
+
|
122 |
+
while not self.stop_event.is_set():
|
123 |
+
try:
|
124 |
+
# 5 dakika sınırına yaklaşıyorsak stream'i sonlandır
|
125 |
+
elapsed = (datetime.now() - start_time).total_seconds()
|
126 |
+
if elapsed > 280: # 4 dakika 40 saniye - güvenli margin
|
127 |
+
log_warning(f"⚠️ Approaching 5-minute limit ({elapsed:.1f}s), ending stream gracefully")
|
128 |
+
break
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
129 |
|
130 |
+
# Get audio chunk with timeout
|
131 |
+
chunk = self.audio_queue.get(timeout=0.1)
|
132 |
+
if chunk is None: # Poison pill
|
133 |
+
log_info("📛 Poison pill received, stopping request generator")
|
134 |
+
break
|
135 |
+
chunk_count += 1
|
136 |
+
|
137 |
+
# Sadece önemli milestone'larda logla
|
138 |
+
if chunk_count == 1:
|
139 |
+
log_info(f"📤 First chunk sent to Google STT, size: {len(chunk)} bytes")
|
140 |
+
elif chunk_count % 100 == 0:
|
141 |
+
log_info(f"📤 Sent {chunk_count} chunks to Google STT (elapsed: {elapsed:.1f}s)")
|
142 |
+
|
143 |
+
yield speech.StreamingRecognizeRequest(audio_content=chunk)
|
144 |
+
except queue.Empty:
|
145 |
+
continue
|
146 |
+
except Exception as e:
|
147 |
+
log_error(f"❌ Error in request generator: {e}")
|
148 |
+
break
|
149 |
+
|
150 |
+
# Create streaming client
|
151 |
+
requests = request_generator()
|
152 |
|
153 |
+
log_info("🎤 Creating Google STT streaming client...")
|
|
|
|
|
154 |
|
155 |
+
try:
|
156 |
+
responses = self.client.streaming_recognize(self.streaming_config, requests)
|
157 |
+
log_info("✅ Google STT streaming client created")
|
158 |
|
159 |
+
# Process responses
|
160 |
+
response_count = 0
|
161 |
+
empty_response_count = 0
|
162 |
|
163 |
+
for response in responses:
|
164 |
+
response_count += 1
|
|
|
|
|
|
|
|
|
165 |
|
166 |
+
if self.stop_event.is_set():
|
167 |
+
log_info("🛑 Stop event detected, breaking response loop")
|
168 |
+
break
|
|
|
|
|
|
|
169 |
|
170 |
+
# Boş response'ları say ama loglama
|
171 |
+
if not response.results:
|
172 |
+
empty_response_count += 1
|
173 |
+
if empty_response_count % 50 == 0:
|
174 |
+
log_warning(f"⚠️ Received {empty_response_count} empty responses from Google STT")
|
175 |
+
continue
|
|
|
|
|
|
|
176 |
|
177 |
+
for result in response.results:
|
178 |
+
if not result.alternatives:
|
179 |
+
continue
|
180 |
+
|
181 |
+
# İlk alternatifi al
|
182 |
+
alternative = result.alternatives[0]
|
183 |
|
184 |
+
# Sadece anlamlı text'leri işle
|
185 |
+
if alternative.transcript.strip():
|
186 |
+
# Create transcription result
|
187 |
+
transcription = TranscriptionResult(
|
188 |
+
text=alternative.transcript,
|
189 |
+
is_final=result.is_final,
|
190 |
+
confidence=alternative.confidence if hasattr(alternative, 'confidence') and alternative.confidence else 0.0,
|
191 |
+
timestamp=datetime.now().timestamp()
|
192 |
+
)
|
193 |
+
|
194 |
+
# Put result in queue
|
195 |
+
self._put_result(transcription)
|
196 |
+
|
197 |
+
# SADECE final result'ları logla
|
198 |
+
if result.is_final:
|
199 |
+
log_info(f"🎯 GOOGLE STT FINAL: '{alternative.transcript}'")
|
200 |
|
201 |
+
log_info(f"📊 Google STT stream ended. Total responses: {response_count}, Empty: {empty_response_count}")
|
202 |
|
203 |
+
except Exception as e:
|
204 |
+
error_msg = str(e)
|
205 |
+
|
206 |
+
# Detaylı hata mesajları
|
207 |
+
if "Exceeded maximum allowed stream duration" in error_msg:
|
208 |
+
log_warning("⚠️ Stream duration limit exceeded (5 minutes). This is expected for long sessions.")
|
209 |
+
elif "Bad language code" in error_msg:
|
210 |
+
log_error(f"❌ Invalid language code in STT config. Check locale settings.")
|
211 |
+
elif "invalid_argument" in error_msg:
|
212 |
+
log_error(f"❌ Invalid STT configuration. Check encoding and sample rate.")
|
213 |
+
elif "Deadline Exceeded" in error_msg:
|
214 |
+
log_error(f"❌ Google STT timeout - possibly network issue or slow connection")
|
215 |
+
elif "503" in error_msg or "Service Unavailable" in error_msg:
|
216 |
+
log_error(f"❌ Google STT service temporarily unavailable. Will retry...")
|
217 |
+
else:
|
218 |
+
log_error(f"❌ Google STT stream error: {error_msg}")
|
219 |
|
220 |
+
except Exception as e:
|
221 |
+
import traceback
|
222 |
+
log_error(f"❌ Fatal error in STT stream thread", error=str(e), traceback=traceback.format_exc())
|
223 |
+
finally:
|
224 |
+
log_info("🎤 Google STT stream thread ended")
|
225 |
+
# Thread bittiğinde streaming flag'ini kapat
|
226 |
+
self.is_streaming = False
|
227 |
|
228 |
async def stream_audio(self, audio_chunk: bytes) -> AsyncIterator[TranscriptionResult]:
|
229 |
"""Stream audio chunk and get transcription results"""
|
|
|
250 |
raise
|
251 |
|
252 |
async def stop_streaming(self) -> Optional[TranscriptionResult]:
|
253 |
+
"""Stop streaming and get final result"""
|
254 |
+
if not self.is_streaming:
|
255 |
+
return None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
256 |
|
257 |
+
try:
|
258 |
+
log_info("🛑 Stopping Google STT streaming...")
|
259 |
+
|
260 |
+
self.is_streaming = False
|
261 |
+
self.stop_event.set()
|
262 |
+
|
263 |
+
# Send poison pill to queue
|
264 |
+
self.audio_queue.put(None)
|
265 |
+
|
266 |
+
# Wait for thread to finish with longer timeout
|
267 |
+
if self.stream_thread and self.stream_thread.is_alive():
|
268 |
+
self.stream_thread.join(timeout=10.0)
|
269 |
+
if self.stream_thread.is_alive():
|
270 |
+
log_warning("⚠️ STT thread did not stop gracefully")
|
271 |
+
|
272 |
+
# Clear queues
|
273 |
+
while not self.audio_queue.empty():
|
274 |
+
try:
|
275 |
+
self.audio_queue.get_nowait()
|
276 |
+
except:
|
277 |
+
pass
|
278 |
+
|
279 |
+
final_result = None
|
280 |
+
while not self.responses_queue.empty():
|
281 |
+
try:
|
282 |
+
result = self.responses_queue.get_nowait()
|
283 |
+
if result.is_final:
|
284 |
+
final_result = result
|
285 |
+
except:
|
286 |
+
pass
|
287 |
+
|
288 |
+
# Reset thread reference
|
289 |
+
self.stream_thread = None
|
290 |
+
|
291 |
+
log_info("✅ Google STT streaming stopped")
|
292 |
+
return final_result
|
293 |
+
|
294 |
+
except Exception as e:
|
295 |
+
log_error(f"❌ Failed to stop Google STT streaming", error=str(e))
|
296 |
+
self.stream_thread = None
|
297 |
+
return None
|
298 |
|
299 |
def supports_realtime(self) -> bool:
|
300 |
"""Google Cloud STT supports real-time streaming"""
|