Fix: skip OpenAI Whisper API for Nvidia, use tiny model by default

Fix critical issues: API key race conditions, unused pipeline stages, port mapping

Updated project files

initial commit