fix: cast audio output to float32 and adjust server binding based on platform environment

chore: remove unused aiohttp and addict dependencies from project configuration

feat: replace VoxCPM2 with Bark for TTS and update project dependencies

refactor: replace ollama with custom MiniCPM5-1B pipeline and update Gradio interface to support chat history

feat: add MiniCPM5-1B model integration and migrate TTS to VoxCPM2 with TorchDynamo compilation fixes

feat: implement Kokoro TTS model integration and voice processing pipeline