v0.2.0

Changed

• Breaking: the default port is now 8255 (TALK on a phone keypad), not
  8000. Port 8000 is the default of many other servers. To keep the old port,
  set VM_PORT=8000 in .env and pass port=8000 to VoiceManagerClient.

v0.1.0

First public release. Developed privately since March 2026.

• Streaming STT and TTS over one WebSocket (/ws/stream), full duplex: interim
  and final transcripts, batch tts and streaming tts_stream, tts_cancel
  for barge-in, turns and flush. The protocol is documented in
  docs/client-integration.md, so a client can be written in any language.
• Python client library voicemanager.client.
• Pluggable engines. Default: Nemotron-3.5 ASR and Breeze-TTS 2 in an
  audio.cpp sidecar. Also faster-whisper, NeMo Nemotron and Qwen3-ASR for STT,
  and OpenVoice v2 (voice cloning) and VibeVoice for TTS.
• Users and bearer tokens in a SQLite store, each token with an optional voice.
  The voicemanager command runs the server and manages users and tokens.
• Web UI without JavaScript, where users manage their tokens and password.
• Preset voices in voices/designed/, and sample systemd units in
  contrib/systemd/.