fanlusky
a7779d2b05
`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。
1.1 KiB
Shared VoiceInput
Reusable C++17 code is in src/ with public headers in include/msime/voice/. The standalone Windows host in platforms/windows/ must not enter the public targets. The host owns microphone permission, hotkeys, UI, token storage, threading, and committing text; for this product the host is server/.
Build cmake -S voice -B build-voice, then build and run CTest. Capture is optional, and core engine builds must not acquire voice dependencies. Local Whisper recognition was removed with the vendoring; do not reintroduce it without saying why the cloud path is insufficient. Public code uses metasequoia::voice, UTF-8 text and mono 16 kHz finite float PCM. Never open a microphone, send user audio or contact paid providers in automated tests; use synthetic samples and loopback HTTP fixtures.
Preserve imported history and licenses. Network code must release handles on success and failure, bound responses, check status, support cancellation and avoid logging response bodies or tokens. Audio device state is per instance. Start/stop are serialized by the host; callbacks must never call stop or destroy the capture object.