Add Gunicorn support for production in Vosk service and update requirements

vosk/vosk_service/gunicorn_config.py

@@ -0,0 +1,43 @@
+# Gunicorn configuration for Vosk service with 192-core optimization
+
+# Server socket
+bind = "0.0.0.0:5000"
+backlog = 2048
+
+# Worker processes
+workers = 48  # 192/4 for optimal worker count
+worker_class = "sync"
+worker_connections = 1000
+max_requests = 1000
+max_requests_jitter = 50
+preload_app = True
+
+# Timeout settings
+timeout = 300
+keepalive = 2
+graceful_timeout = 30
+
+# Logging
+accesslog = "-"
+errorlog = "-"
+loglevel = "info"
+access_log_format = '%(h)s %(l)s %(u)s %(t)s "%(r)s" %(s)s %(b)s "%(f)s" "%(a)s" %(D)s'
+
+# Process naming
+proc_name = "vosk-service"
+
+# Security
+limit_request_line = 4094
+limit_request_fields = 100
+limit_request_field_size = 8190
+
+# Performance
+worker_tmp_dir = "/dev/shm"
+worker_exit_on_app_exit = True
+
+# Memory management
+max_requests_jitter = 50
+preload_app = True
+
+# Worker lifecycle
+worker_abort_on_app_exit = True