Vision
Low‑latency speech pipeline (STT → LLM → TTS) with terminology glossaries, diarization, and privacy mode. Target: fluent bilingual presence in calls and workshops.
Milestones (30/60/90)
- 30d: Glossary CRUD; SSML presets; hotkeys; basic diarization.
- 60d: Streaming UX; dynamic model routing; offline cache; security hardening.
- 90d: Team spaces; phrasebooks; export/share; analytics.
KPIs
- Round‑trip latency p95; ASR WER; TTS MOS proxy; user retention.
Stack
Azure STT/TTSFastAPIElectron/DeskPrompt policiesOllama/GPT