Add health check endpoints and automatic language detection
Health Check Features (Item 12): - Added /health endpoint for basic health monitoring - Added /health/detailed for comprehensive component status - Added /health/ready for Kubernetes readiness probes - Added /health/live for liveness checks - Frontend health monitoring with auto-recovery - Clear stuck requests after 60 seconds - Visual health warnings when service is degraded - Monitoring script for external health checks Automatic Language Detection (Item 13): - Added "Auto-detect" option in source language dropdown - Whisper automatically detects language when auto-detect is selected - Shows detected language in UI after transcription - Updates language selector with detected language - Caches transcriptions with correct detected language 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
parent
829e8c3978
commit
0c9186e57e
149
app.py
149
app.py
@ -402,9 +402,11 @@ def transcribe():
|
||||
audio_file.save(temp_path)
|
||||
|
||||
try:
|
||||
# Check if we should auto-detect language
|
||||
auto_detect = source_lang == 'auto' or source_lang == ''
|
||||
|
||||
# Use Whisper for transcription with GPU optimizations
|
||||
transcribe_options = {
|
||||
"language": LANGUAGE_TO_CODE.get(source_lang, None),
|
||||
"task": "transcribe",
|
||||
"temperature": 0, # Disable temperature sampling for faster inference
|
||||
"best_of": 1, # Disable beam search for faster inference
|
||||
@ -416,6 +418,10 @@ def transcribe():
|
||||
"no_speech_threshold": 0.6
|
||||
}
|
||||
|
||||
# Only set language if not auto-detecting
|
||||
if not auto_detect:
|
||||
transcribe_options["language"] = LANGUAGE_TO_CODE.get(source_lang, None)
|
||||
|
||||
# Clear GPU cache before transcription
|
||||
if device.type == 'cuda':
|
||||
torch.cuda.empty_cache()
|
||||
@ -429,6 +435,19 @@ def transcribe():
|
||||
|
||||
transcribed_text = result["text"]
|
||||
|
||||
# Get detected language if auto-detection was used
|
||||
detected_language = None
|
||||
if auto_detect and 'language' in result:
|
||||
# Convert language code back to full name
|
||||
detected_code = result['language']
|
||||
for lang_name, lang_code in LANGUAGE_TO_CODE.items():
|
||||
if lang_code == detected_code:
|
||||
detected_language = lang_name
|
||||
break
|
||||
|
||||
# Log detected language
|
||||
logger.info(f"Auto-detected language: {detected_language} ({detected_code})")
|
||||
|
||||
# Send notification if push is enabled
|
||||
if len(push_subscriptions) > 0:
|
||||
send_push_notification(
|
||||
@ -437,10 +456,16 @@ def transcribe():
|
||||
tag="transcription-complete"
|
||||
)
|
||||
|
||||
return jsonify({
|
||||
response = {
|
||||
'success': True,
|
||||
'text': transcribed_text
|
||||
})
|
||||
}
|
||||
|
||||
# Include detected language if auto-detection was used
|
||||
if detected_language:
|
||||
response['detected_language'] = detected_language
|
||||
|
||||
return jsonify(response)
|
||||
except Exception as e:
|
||||
logger.error(f"Transcription error: {str(e)}")
|
||||
return jsonify({'error': f'Transcription failed: {str(e)}'}), 500
|
||||
@ -598,5 +623,123 @@ def get_audio(filename):
|
||||
logger.error(f"Audio retrieval error: {str(e)}")
|
||||
return jsonify({'error': f'Audio retrieval failed: {str(e)}'}), 500
|
||||
|
||||
# Health check endpoints for monitoring
|
||||
@app.route('/health', methods=['GET'])
|
||||
def health_check():
|
||||
"""Basic health check endpoint"""
|
||||
return jsonify({
|
||||
'status': 'healthy',
|
||||
'timestamp': time.time(),
|
||||
'service': 'voice-translator'
|
||||
})
|
||||
|
||||
@app.route('/health/detailed', methods=['GET'])
|
||||
def detailed_health_check():
|
||||
"""Detailed health check with component status"""
|
||||
health_status = {
|
||||
'status': 'healthy',
|
||||
'timestamp': time.time(),
|
||||
'components': {
|
||||
'whisper': {'status': 'unknown'},
|
||||
'ollama': {'status': 'unknown'},
|
||||
'tts': {'status': 'unknown'},
|
||||
'gpu': {'status': 'unknown'}
|
||||
},
|
||||
'metrics': {}
|
||||
}
|
||||
|
||||
# Check Whisper model
|
||||
try:
|
||||
if whisper_model is not None:
|
||||
health_status['components']['whisper']['status'] = 'healthy'
|
||||
health_status['components']['whisper']['model_size'] = MODEL_SIZE
|
||||
else:
|
||||
health_status['components']['whisper']['status'] = 'unhealthy'
|
||||
health_status['status'] = 'degraded'
|
||||
except Exception as e:
|
||||
health_status['components']['whisper']['status'] = 'unhealthy'
|
||||
health_status['components']['whisper']['error'] = str(e)
|
||||
health_status['status'] = 'unhealthy'
|
||||
|
||||
# Check GPU availability
|
||||
try:
|
||||
if torch.cuda.is_available():
|
||||
health_status['components']['gpu']['status'] = 'healthy'
|
||||
health_status['components']['gpu']['device'] = torch.cuda.get_device_name(0)
|
||||
health_status['components']['gpu']['memory_allocated'] = f"{torch.cuda.memory_allocated(0) / 1024**2:.2f} MB"
|
||||
health_status['components']['gpu']['memory_reserved'] = f"{torch.cuda.memory_reserved(0) / 1024**2:.2f} MB"
|
||||
elif torch.backends.mps.is_available():
|
||||
health_status['components']['gpu']['status'] = 'healthy'
|
||||
health_status['components']['gpu']['device'] = 'Apple Silicon GPU'
|
||||
else:
|
||||
health_status['components']['gpu']['status'] = 'not_available'
|
||||
health_status['components']['gpu']['device'] = 'CPU'
|
||||
except Exception as e:
|
||||
health_status['components']['gpu']['status'] = 'error'
|
||||
health_status['components']['gpu']['error'] = str(e)
|
||||
|
||||
# Check Ollama connection
|
||||
try:
|
||||
ollama_models = ollama.list()
|
||||
health_status['components']['ollama']['status'] = 'healthy'
|
||||
health_status['components']['ollama']['available_models'] = len(ollama_models.get('models', []))
|
||||
except Exception as e:
|
||||
health_status['components']['ollama']['status'] = 'unhealthy'
|
||||
health_status['components']['ollama']['error'] = str(e)
|
||||
health_status['status'] = 'degraded'
|
||||
|
||||
# Check TTS server
|
||||
try:
|
||||
tts_response = requests.get(app.config['TTS_SERVER'].replace('/v1/audio/speech', '/health'), timeout=5)
|
||||
if tts_response.status_code == 200:
|
||||
health_status['components']['tts']['status'] = 'healthy'
|
||||
health_status['components']['tts']['server_url'] = app.config['TTS_SERVER']
|
||||
else:
|
||||
health_status['components']['tts']['status'] = 'unhealthy'
|
||||
health_status['components']['tts']['http_status'] = tts_response.status_code
|
||||
health_status['status'] = 'degraded'
|
||||
except Exception as e:
|
||||
health_status['components']['tts']['status'] = 'unhealthy'
|
||||
health_status['components']['tts']['error'] = str(e)
|
||||
health_status['status'] = 'degraded'
|
||||
|
||||
# Add system metrics
|
||||
health_status['metrics']['uptime'] = time.time() - app.start_time if hasattr(app, 'start_time') else 0
|
||||
health_status['metrics']['request_count'] = getattr(app, 'request_count', 0)
|
||||
|
||||
# Set appropriate HTTP status code
|
||||
http_status = 200 if health_status['status'] == 'healthy' else 503 if health_status['status'] == 'unhealthy' else 200
|
||||
|
||||
return jsonify(health_status), http_status
|
||||
|
||||
@app.route('/health/ready', methods=['GET'])
|
||||
def readiness_check():
|
||||
"""Readiness probe - checks if service is ready to accept traffic"""
|
||||
try:
|
||||
# Check if all critical components are loaded
|
||||
if whisper_model is None:
|
||||
return jsonify({'status': 'not_ready', 'reason': 'Whisper model not loaded'}), 503
|
||||
|
||||
# Check Ollama connection
|
||||
ollama.list()
|
||||
|
||||
return jsonify({'status': 'ready', 'timestamp': time.time()})
|
||||
except Exception as e:
|
||||
return jsonify({'status': 'not_ready', 'reason': str(e)}), 503
|
||||
|
||||
@app.route('/health/live', methods=['GET'])
|
||||
def liveness_check():
|
||||
"""Liveness probe - basic check to see if process is alive"""
|
||||
return jsonify({'status': 'alive', 'timestamp': time.time()})
|
||||
|
||||
# Initialize app start time for metrics
|
||||
app.start_time = time.time()
|
||||
app.request_count = 0
|
||||
|
||||
# Middleware to count requests
|
||||
@app.before_request
|
||||
def before_request():
|
||||
app.request_count = getattr(app, 'request_count', 0) + 1
|
||||
|
||||
if __name__ == '__main__':
|
||||
app.run(host='0.0.0.0', port=5005, debug=True)
|
||||
|
||||
91
health-monitor.py
Executable file
91
health-monitor.py
Executable file
@ -0,0 +1,91 @@
|
||||
#!/usr/bin/env python3
|
||||
"""
|
||||
Health monitoring script for Talk2Me application
|
||||
Usage: python health-monitor.py [--detailed] [--interval SECONDS]
|
||||
"""
|
||||
|
||||
import requests
|
||||
import time
|
||||
import argparse
|
||||
import json
|
||||
from datetime import datetime
|
||||
|
||||
def check_health(url, detailed=False):
|
||||
"""Check health of the Talk2Me service"""
|
||||
endpoint = f"{url}/health/detailed" if detailed else f"{url}/health"
|
||||
|
||||
try:
|
||||
response = requests.get(endpoint, timeout=5)
|
||||
data = response.json()
|
||||
|
||||
if detailed:
|
||||
print(f"\n=== Health Check at {datetime.now().strftime('%Y-%m-%d %H:%M:%S')} ===")
|
||||
print(f"Overall Status: {data['status'].upper()}")
|
||||
print("\nComponent Status:")
|
||||
for component, status in data['components'].items():
|
||||
status_icon = "✅" if status.get('status') == 'healthy' else "❌"
|
||||
print(f" {status_icon} {component}: {status.get('status', 'unknown')}")
|
||||
if 'error' in status:
|
||||
print(f" Error: {status['error']}")
|
||||
if 'device' in status:
|
||||
print(f" Device: {status['device']}")
|
||||
if 'model_size' in status:
|
||||
print(f" Model: {status['model_size']}")
|
||||
|
||||
if 'metrics' in data:
|
||||
print("\nMetrics:")
|
||||
uptime = data['metrics'].get('uptime', 0)
|
||||
hours = int(uptime // 3600)
|
||||
minutes = int((uptime % 3600) // 60)
|
||||
print(f" Uptime: {hours}h {minutes}m")
|
||||
print(f" Request Count: {data['metrics'].get('request_count', 0)}")
|
||||
else:
|
||||
status_icon = "✅" if response.status_code == 200 else "❌"
|
||||
print(f"{status_icon} {datetime.now().strftime('%H:%M:%S')} - Status: {data.get('status', 'unknown')}")
|
||||
|
||||
return response.status_code == 200
|
||||
|
||||
except requests.exceptions.ConnectionError:
|
||||
print(f"❌ {datetime.now().strftime('%H:%M:%S')} - Connection failed")
|
||||
return False
|
||||
except requests.exceptions.Timeout:
|
||||
print(f"❌ {datetime.now().strftime('%H:%M:%S')} - Request timeout")
|
||||
return False
|
||||
except Exception as e:
|
||||
print(f"❌ {datetime.now().strftime('%H:%M:%S')} - Error: {str(e)}")
|
||||
return False
|
||||
|
||||
def main():
|
||||
parser = argparse.ArgumentParser(description='Monitor Talk2Me service health')
|
||||
parser.add_argument('--url', default='http://localhost:5005', help='Service URL')
|
||||
parser.add_argument('--detailed', action='store_true', help='Show detailed health info')
|
||||
parser.add_argument('--interval', type=int, default=30, help='Check interval in seconds')
|
||||
parser.add_argument('--once', action='store_true', help='Run once and exit')
|
||||
|
||||
args = parser.parse_args()
|
||||
|
||||
print(f"Monitoring {args.url}")
|
||||
print("Press Ctrl+C to stop\n")
|
||||
|
||||
consecutive_failures = 0
|
||||
|
||||
try:
|
||||
while True:
|
||||
success = check_health(args.url, args.detailed)
|
||||
|
||||
if not success:
|
||||
consecutive_failures += 1
|
||||
if consecutive_failures >= 3:
|
||||
print(f"\n⚠️ ALERT: Service has been down for {consecutive_failures} consecutive checks!")
|
||||
else:
|
||||
consecutive_failures = 0
|
||||
|
||||
if args.once:
|
||||
break
|
||||
|
||||
time.sleep(args.interval)
|
||||
except KeyboardInterrupt:
|
||||
print("\n\nMonitoring stopped.")
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
@ -107,6 +107,9 @@ function initApp(): void {
|
||||
// Initialize queue status updates
|
||||
initQueueStatus();
|
||||
|
||||
// Start health monitoring
|
||||
startHealthMonitoring();
|
||||
|
||||
// Update TTS server URL and API key
|
||||
updateTtsServer.addEventListener('click', function() {
|
||||
const newUrl = ttsServerUrl.value.trim();
|
||||
@ -153,6 +156,11 @@ function initApp(): void {
|
||||
|
||||
// Event listeners for language selection
|
||||
sourceLanguage.addEventListener('change', function() {
|
||||
// Skip conflict check for auto-detect
|
||||
if (sourceLanguage.value === 'auto') {
|
||||
return;
|
||||
}
|
||||
|
||||
if (targetLanguage.value === sourceLanguage.value) {
|
||||
for (let i = 0; i < targetLanguage.options.length; i++) {
|
||||
if (targetLanguage.options[i].value !== sourceLanguage.value) {
|
||||
@ -383,10 +391,24 @@ function initApp(): void {
|
||||
|
||||
if (data.success && data.text) {
|
||||
currentSourceText = data.text;
|
||||
|
||||
// Handle auto-detected language
|
||||
if (data.detected_language && sourceLanguage.value === 'auto') {
|
||||
// Update the source language selector
|
||||
sourceLanguage.value = data.detected_language;
|
||||
|
||||
// Show detected language info
|
||||
sourceText.innerHTML = `<p class="fade-in">${data.text}</p>
|
||||
<small class="text-muted">Detected language: ${data.detected_language}</small>`;
|
||||
|
||||
statusIndicator.textContent = `Transcription complete (${data.detected_language} detected)`;
|
||||
} else {
|
||||
sourceText.innerHTML = `<p class="fade-in">${data.text}</p>`;
|
||||
statusIndicator.textContent = 'Transcription complete';
|
||||
}
|
||||
|
||||
playSource.disabled = false;
|
||||
translateBtn.disabled = false;
|
||||
statusIndicator.textContent = 'Transcription complete';
|
||||
statusIndicator.classList.remove('processing');
|
||||
statusIndicator.classList.add('success');
|
||||
setTimeout(() => statusIndicator.classList.remove('success'), 2000);
|
||||
@ -394,7 +416,7 @@ function initApp(): void {
|
||||
// Cache the transcription in IndexedDB
|
||||
saveToIndexedDB('transcriptions', {
|
||||
text: data.text,
|
||||
language: sourceLanguage.value,
|
||||
language: data.detected_language || sourceLanguage.value,
|
||||
timestamp: new Date().toISOString()
|
||||
} as TranscriptionRecord);
|
||||
} else {
|
||||
@ -753,6 +775,99 @@ function initApp(): void {
|
||||
// Initial update
|
||||
updateQueueDisplay();
|
||||
}
|
||||
|
||||
// Health monitoring and auto-recovery
|
||||
function startHealthMonitoring(): void {
|
||||
let consecutiveFailures = 0;
|
||||
const maxConsecutiveFailures = 3;
|
||||
|
||||
async function checkHealth(): Promise<void> {
|
||||
try {
|
||||
const response = await fetch('/health', {
|
||||
method: 'GET',
|
||||
signal: AbortSignal.timeout(5000) // 5 second timeout
|
||||
});
|
||||
|
||||
if (response.ok) {
|
||||
consecutiveFailures = 0;
|
||||
|
||||
// Remove any health warning if shown
|
||||
const healthWarning = document.getElementById('healthWarning');
|
||||
if (healthWarning) {
|
||||
healthWarning.style.display = 'none';
|
||||
}
|
||||
} else {
|
||||
handleHealthCheckFailure();
|
||||
}
|
||||
} catch (error) {
|
||||
handleHealthCheckFailure();
|
||||
}
|
||||
}
|
||||
|
||||
function handleHealthCheckFailure(): void {
|
||||
consecutiveFailures++;
|
||||
console.warn(`Health check failed (${consecutiveFailures}/${maxConsecutiveFailures})`);
|
||||
|
||||
if (consecutiveFailures >= maxConsecutiveFailures) {
|
||||
showHealthWarning();
|
||||
|
||||
// Attempt auto-recovery
|
||||
attemptAutoRecovery();
|
||||
}
|
||||
}
|
||||
|
||||
function showHealthWarning(): void {
|
||||
let healthWarning = document.getElementById('healthWarning');
|
||||
if (!healthWarning) {
|
||||
healthWarning = document.createElement('div');
|
||||
healthWarning.id = 'healthWarning';
|
||||
healthWarning.className = 'alert alert-warning alert-dismissible fade show position-fixed top-0 start-50 translate-middle-x mt-3';
|
||||
healthWarning.style.zIndex = '9999';
|
||||
healthWarning.innerHTML = `
|
||||
<i class="fas fa-exclamation-triangle"></i> Service health check failed.
|
||||
Some features may be unavailable.
|
||||
<button type="button" class="btn-close" data-bs-dismiss="alert"></button>
|
||||
`;
|
||||
document.body.appendChild(healthWarning);
|
||||
}
|
||||
healthWarning.style.display = 'block';
|
||||
}
|
||||
|
||||
async function attemptAutoRecovery(): Promise<void> {
|
||||
console.log('Attempting auto-recovery...');
|
||||
|
||||
// Clear any stuck requests in the queue
|
||||
const queue = RequestQueueManager.getInstance();
|
||||
queue.clearStuckRequests();
|
||||
|
||||
// Re-check TTS server
|
||||
checkTtsServer();
|
||||
|
||||
// Try to reload service worker if available
|
||||
if ('serviceWorker' in navigator) {
|
||||
try {
|
||||
const registration = await navigator.serviceWorker.getRegistration();
|
||||
if (registration) {
|
||||
await registration.update();
|
||||
console.log('Service worker updated');
|
||||
}
|
||||
} catch (error) {
|
||||
console.error('Failed to update service worker:', error);
|
||||
}
|
||||
}
|
||||
|
||||
// Reset failure counter after recovery attempt
|
||||
setTimeout(() => {
|
||||
consecutiveFailures = 0;
|
||||
}, 30000); // Wait 30 seconds before resetting
|
||||
}
|
||||
|
||||
// Check health every 30 seconds
|
||||
setInterval(checkHealth, 30000);
|
||||
|
||||
// Initial health check after 5 seconds
|
||||
setTimeout(checkHealth, 5000);
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
|
||||
@ -217,6 +217,31 @@ export class RequestQueueManager {
|
||||
this.queue = [];
|
||||
}
|
||||
|
||||
// Clear stuck requests (requests older than 60 seconds)
|
||||
clearStuckRequests(): void {
|
||||
const now = Date.now();
|
||||
const stuckThreshold = 60000; // 60 seconds
|
||||
|
||||
// Clear stuck active requests
|
||||
this.activeRequests.forEach((request, id) => {
|
||||
if (now - request.timestamp > stuckThreshold) {
|
||||
console.warn(`Clearing stuck active request: ${request.type}`);
|
||||
request.reject(new Error('Request timeout - cleared by recovery'));
|
||||
this.activeRequests.delete(id);
|
||||
}
|
||||
});
|
||||
|
||||
// Clear old queued requests
|
||||
this.queue = this.queue.filter(request => {
|
||||
if (now - request.timestamp > stuckThreshold) {
|
||||
console.warn(`Clearing stuck queued request: ${request.type}`);
|
||||
request.reject(new Error('Request timeout - cleared by recovery'));
|
||||
return false;
|
||||
}
|
||||
return true;
|
||||
});
|
||||
}
|
||||
|
||||
// Update settings
|
||||
updateSettings(settings: {
|
||||
maxConcurrent?: number;
|
||||
|
||||
@ -4,6 +4,7 @@ export interface TranscriptionResponse {
|
||||
success: boolean;
|
||||
text?: string;
|
||||
error?: string;
|
||||
detected_language?: string;
|
||||
}
|
||||
|
||||
export interface TranslationResponse {
|
||||
|
||||
@ -133,6 +133,7 @@
|
||||
</div>
|
||||
<div class="card-body">
|
||||
<select id="sourceLanguage" class="form-select language-select mb-3">
|
||||
<option value="auto">Auto-detect</option>
|
||||
{% for language in languages %}
|
||||
<option value="{{ language }}">{{ language }}</option>
|
||||
{% endfor %}
|
||||
|
||||
Loading…
Reference in New Issue
Block a user