v0.2.0

What's new

• Ollama model warmup at startup — both Whisper and Ollama models are now loaded sequentially at startup, so the first transcription is fast (previously ~20s cold start, now ~1-2s)
• Startup shows model load times for visibility

Full changelog

See CHANGELOG.md