v0.1.19

Highlights

• OpenAI-compatible realtime transcription backend (#49, #50) — new openai-compatible-realtime backend that streams audio to any external WebSocket ASR server speaking the OpenAI Realtime transcription protocol, with Lemonade as the first supported profile. Bring your own local, GPU-optimized whisper.cpp server and get phrase-at-a-time transcripts via server VAD. It shares one protocol engine with the existing openai-realtime backend, which also gained a more robust stop/finalize path (no more ~30s hang or false timeout when a turn completes right before you stop). Replaceable interim partials are buffered internally; only completed phrases are typed. Configure with backend = "openai-compatible-realtime" (see docs/configuration.md). Thanks @pettijohn for the PR.

Install / Upgrade

curl -sSL https://y0sif.github.io/whisrs/install.sh | bash

# Or:
cargo install whisrs --locked
yay -S whisrs-git

Full Changelog: v0.1.18...v0.1.19