lsm / HyperNeo / 30723944371

83%

Build Type

push

github

Committed by web-flow

Commit Message

feat: add voice input transcription (#2279)

Add a mic button to the chat composer that records WAV audio and transcribes it
via a user-configured OpenAI-compatible Audio Transcriptions backend, inserting
the transcript at the cursor (no auto-send, no TTS, no realtime). Global settings only.

- Settings: voice endpoint, model, optional API key, allow-private-network, allow-insecure-TLS.
- Security: DNS pinning, per-hop redirect re-validation + credential stripping,
  per-client (1) + daemon-wide (4) concurrency, rate limits (6/min client, 20/min daemon)
  with TOCTOU-safe check-then-commit, 256 KB response cap with streaming cancel,
  base64 pre-validation, server-owned credential scope under a voice-credential lock.
- Recorder: AudioWorklet with ScriptProcessorNode fallback; sample cap under the 3 MB WAV limit.
- Composer: CJK-aware spacing, leading-space suppression, char cap, submit guard while recording.

Coverage Stats

11759 of 16965 branches covered (69.31%)

Branch coverage included in aggregate %.

822 of 1024 new or added lines in 14 files covered. (80.27%)

348 existing lines in 10 files now uncovered.

101206 of 119768 relevant lines covered (84.5%)

300.7 hits per line