Files
fanlusky a7779d2b05 refactor(engine): 将 engine 从 submodule 内嵌为仓内一等组件
`vendor/MetasequoiaImeEngine` submodule 改为 `engine/` 目录,导入的是
product-lock.json 原本钉住的提交 c810d201,逐文件核对过 blob,因此这是一次
搬迁而不是引擎升级。引擎自己的嵌套 submodule(googlepinyinime-rev、utfcpp、
miniaudio)按各自 gitlink 一并展开,来源与提交见 engine/UPSTREAM.md。

Windows 专门化裁掉了不属于本产品的部分:whisper.cpp 与本地推理 provider、
macOS 示例宿主、整个 dictionary/ 建库流水线、上游 CI 与 release-please 配置。
词库仍按 product-lock.json 在构建时取发布产物。

引擎不再进 product-lock.json:没有 gitlink 就没有可比对的东西,锁里写一个上游
commit 等于宣称这棵树仍等于上游,而第一个 Windows 专门化改动落地时这句话就假了。
verify-contracts / verify-checkout 随之移除,verify-published 保留。

CI 与工具链:引擎源码并入根 scripts/format.sh(删掉引擎自己的那份,避免两套排除
列表分叉);引擎 voice.yml 里的 silero_vad.onnx 摘要校验移到本仓 lock job;
engine-bump-triage 工作流与其测试删除;gitsubmodule 的 dependabot 频率由 daily
改为 monthly。

README、README.en、AGENTS.md、CONTRIBUTING、docs/product-release.md 及各子目录
文档同步说明新的位置与边界。
2026-09-12 03:40:48 +08:00

1.1 KiB

Voice models

silero_vad.onnx

The voice activity detector, committed to this repository rather than downloaded. It is executed in-process, so it is pinned by digest.

Upstream snakers4/silero-vad, src/silero_vad/data/silero_vad.onnx
Version v6.2 (identical in v6.2.1); introduced by upstream commit bfdc0193023f
License MIT
Size 2327524 bytes
SHA256 1a153a22f4509e292a94e67d6f9b85e8deb25b4988682b7e174c65279d8788e3

CI re-checks that digest on every run, so replacing the file without updating this table fails the build. The check used to live in the engine's own voice.yml; it is now a step of the lock job in .github/workflows/ci.yml.

Only the imported standalone Windows host (MSIME_VOICE_WINDOWS_APP, off by default) loads this model. The supported VAD in the shipped library is the RMS-based VadSegmenter, which needs no model at all.

Whisper models

There are none. Local Whisper inference was removed when the engine was vendored into this repository -- this product transcribes through a cloud provider. See UPSTREAM.md.