0.4.3

Full Changelog: 0.4.2...0.4.3

0.4.2

Full Changelog: 0.4.1...0.4.2

0.4.1

What's Changed

• chore(deps): bump tower-http from 0.6.11 to 0.7.0 in /cortex-stt by @dependabot[bot] in #31
• chore(deps): bump rubato from 2.0.0 to 5.0.0 in /cortex-stt by @dependabot[bot] in #33
• chore(deps): bump the web-deps group across 1 directory with 13 updates by @dependabot[bot] in #37
• chore(deps): bump the actions group across 1 directory with 3 updates by @dependabot[bot] in #38
• chore(deps): bump the rust-deps group across 1 directory with 12 updates by @dependabot[bot] in #39
• chore: unyank chacha20 and disambiguate the two CI workflows by @parkghost in #40

New Contributors

• @parkghost made their first contribution in #40

Full Changelog: 0.4.0...0.4.1

0.4.0

Full Changelog: 0.3.2...0.4.0

0.3.2

Full Changelog: 0.3.1...0.3.2

0.3.1

Full Changelog: 0.3.0...0.3.1

0.3.0

0.3.0 — single transcribe.cpp runtime

The engine layer is rebuilt on transcribe.cpp (GGUF/ggml): one runtime for every model family — Whisper, Parakeet, SenseVoice, Qwen3-ASR, Canary, Moonshine, and more — with real incremental streaming over a single WebSocket endpoint.

⚠️ Breaking changes — read before updating

• All previously downloaded models are removed on first start. 0.2.x model files (.bin whisper / ONNX directories) cannot be loaded by the new runtime and are deleted automatically. Re-download your models from the catalog after updating (Admin UI → Models).
• Model ids changed to upstream catalog slugs (e.g. whisper-tiny-int8 → whisper-tiny, sense-voice-int8 → SenseVoiceSmall). Home Assistant STT entities are rebuilt under the new ids — voice pipelines must re-select their STT entity, and anything wrapping an old entity (e.g. STT Corrector) will raise a repair issue; use its fix flow to re-point at the new entity.
• A stored default model that no longer exists leaves /health reporting starting until you download a model and set a new default.
• SSE transcription variant removed — streaming is WebSocket-only (GET /api/transcribe/stream). Sync JSON and async jobs are unchanged. API clients using the old Accept: text/event-stream variant of POST /api/transcribe must migrate.
• Wire format changes: /api/models now carries quant info (quants[], default_quant, downloaded_quant); history segments is structured (no segments_json).

Update together with the cortex-stt integration ≥ 0.4.0 (WS streaming + new model wire format) and, if you use it, stt-corrector ≥ 0.7.0.

Highlights

• Multi-model on one runtime: full Handy catalog (60+ models, 3–6 quants each), pick a quant at download time, one quant per model on disk
• Real streaming: feed/finalize incremental decoding with partial transcripts; non-streaming models transparently fall back to buffered mode (same wire contract)
• Capture-device attribution + audio quality metrics: history records which Assist satellite recorded each utterance (with the HA integration) plus RMS/peak/clipping levels — find the microphone that transcribes poorly
• Live admin UI: engine load state (lazy loads, idle offloads) and history stream over SSE; per-device history filters
• History facets, invalid-filter 400s, retention unchanged (records + audio survive the update; audio now stored as Ogg Opus for new records)

Full Changelog: 0.2.0...0.3.0

0.2.0

What's Changed

• chore(deps): bump actions/checkout from 6 to 6.0.2 in the actions group by @dependabot[bot] in #24
• chore(deps): bump the web-deps group in /cortex-stt/web with 5 updates by @dependabot[bot] in #25
• chore(deps): bump the rust-deps group in /cortex-stt with 4 updates by @dependabot[bot] in #26
• chore(deps): bump gethostname from 0.5.0 to 1.1.0 in /cortex-stt by @dependabot[bot] in #27
• chore(deps): bump tikv-jemallocator from 0.6.1 to 0.7.0 in /cortex-stt by @dependabot[bot] in #28

Full Changelog: 0.1.7...0.2.0

0.1.7

Full Changelog: 0.1.6...0.1.7

0.1.6

Full Changelog: 0.1.5...0.1.6