feat: implement real-time text-to-speech highlighting and document management within chat interface

perf: swap to lightweight ms-marco-MiniLM reranker for lower cpu latency

feat: implement full-stack authentication, chat interface, document management, and voice service integration

feat: implement SpeechModelManager to handle STT and TTS model initialization and warmup

feat: implement voice transcription service with FFmpeg audio processing and API routing

feat: add VoiceService with audio transcription using FFmpeg and Whisper model integration

feat: implement speech-to-text and text-to-speech services using Whisper and Kokoro with a background model manager.

Initial clean commit