fix(asr): 修复 Windows 本地长语音只识别前段

[Duyi-Wang]

User description

摘要

修复 Windows 本地 ASR 在长语音场景下可能只识别前一段的问题。

主要原因是 foundry-local-whisper 使用 Whisper 系模型，单次解码窗口约 30 秒；当录音超过这个窗口时，单次把整段 WAV 交给 Foundry SDK 可能只返回前段结果。这个 PR 将 Foundry Local Whisper 的 PCM 输入按 30 秒分片转写，再合并分片文本；同时把 Windows 本地 ASR 的转写超时从固定 30 秒改为按音频时长动态增长。

未关联 issue。

修复 / 新增 / 改进

• foundry-local-whisper 录音结束后按 30 秒 PCM chunk 分片，逐段生成临时 WAV 并调用 Foundry SDK。
• 复用 Whisper batch ASR 的 chunk split / transcript join 逻辑，避免中英文、CJK 标点拼接时产生明显断裂。
• 为 Foundry / Sherpa provider 增加 buffer_duration_ms()，coordinator 可按实际录音长度计算 timeout。
• 普通听写、QA、overlay QA、历史重新转录路径都接入 Windows 本地 ASR 动态 timeout。
• 将 Windows Foundry / Sherpa 本地 ASR timeout 调整为 max(30s, ceil(audio_s) + 20s)，避免长音频撞固定 30 秒超时。
• 增加 Foundry 分片、buffer duration、Sherpa buffer duration、Windows 本地 ASR timeout 相关测试。

兼容

• 不包含：

  • 云端 ASR 协议变更。
  • 录音采集链路变更。
  • 插入 / polish 流程变更。
  • 用户配置迁移或模型配置变更。

• 对现有用户 / 本地环境 / 构建流程的影响：

  • 短语音仍使用 30 秒 timeout 保底，行为基本不变。
  • 长语音在 foundry-local-whisper 下会拆成多段本地识别，整体耗时会随音频长度增加，但可以覆盖完整录音。
  • sherpa-onnx-local 仅调整 timeout 预算，不改变识别输入和模型调用方式。

测试计划

• 命令：cargo test

• 结果：本地 Rust 单测通过，546 passed; 0 failed

• 证据路径：本地终端运行结果

• 命令：fork 上的 GitHub Actions CI

• 结果：通过

• 证据路径：https://github.com/Duyi-Wang/openless/actions/runs/28922670487

• 命令：fork 上的桌面安装包构建 Release Tauri (cross-platform)

• 结果：通过

• 证据路径：https://github.com/Duyi-Wang/openless/actions/runs/28922670561

• 命令：fork 上的 Android APK 构建

• 结果：通过

• 证据路径：https://github.com/Duyi-Wang/openless/actions/runs/28922670488

---

PR Type

Bug fix, Enhancement

---

Description

• Split Foundry Local Whisper PCM into 30s chunks for long audio

• Add dynamic timeout based on actual audio duration for both Foundry and Sherpa

• Reuse existing Whisper chunk split/join logic to avoid text gaps

• Apply dynamic timeout across dictation, QA, overlay QA, and re-transcribe paths

---

Diagram Walkthrough

flowchart LR
  A["Audio PCM"] --> B["buffer_duration_ms()"]
  B --> C{"duration > 30s?"}
  C -- Yes --> D["split_pcm_by_duration(30s)"]
  D --> E["Transcribe each chunk separately"]
  E --> F["join_transcript_chunks()"]
  F --> G["Final transcript"]
  C -- No --> H["Transcribe whole audio"]
  H --> G
  B --> I["Calculate timeout: max(30, ceil(audio_secs)+20)"]
  I --> J["Set transcribe timeout"]
  J --> E
  J --> H

Loading

File Walkthrough

	Relevant files
Enhancement	foundry_provider.rs Implement PCM chunking for Foundry Local Whisper                  openless-all/app/src-tauri/src/asr/local/foundry_provider.rs Added FOUNDRY_WHISPER_CHUNK_LIMIT_MS constant (30s) Added buffer_duration_ms() method to report buffered PCM duration Modified transcribe() to split PCM into chunks and transcribe each Added tests for chunk splitting and buffer duration +73/-17  sherpa_provider.rs Add buffer duration reporting for Sherpa provider                openless-all/app/src-tauri/src/asr/local/sherpa_provider.rs Added buffer_duration_ms() method for both Offline and Online modes Added test for buffer duration without consuming PCM +25/-0    coordinator.rs Implement dynamic transcribe timeout for local ASR              openless-all/app/src-tauri/src/coordinator.rs Replaced fixed 30s timeout with dynamic foundry_audio_transcribe_timeout(audio_secs) and sherpa_audio_transcribe_timeout(audio_secs) Added tests for new timeout functions Updated ActiveAsr::FoundryLocalWhisper and SherpaOnnxLocal branches to use dynamic timeout +43/-19
Miscellaneous	whisper.rs Make Whisper chunk utilities public                                            openless-all/app/src-tauri/src/asr/whisper.rs Made split_pcm_by_duration and join_transcript_chunks pub(crate) for reuse +2/-2
Bug fix	dictation.rs Apply dynamic timeout to dictation end session                      openless-all/app/src-tauri/src/coordinator/dictation.rs Updated Foundry and Sherpa dictation end_session to compute dynamic timeout from buffer_duration_ms Added logging for audio duration and timeout +16/-8    qa_session.rs Apply dynamic timeout to QA and overlay dictation paths    openless-all/app/src-tauri/src/coordinator/qa_session.rs Updated overlay dictation and end QA session for Foundry/Sherpa to use dynamic timeout Added logging for audio duration and timeout +22/-16

---

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🧪 PR contains tests

🔒 No security concerns identified

⚡ No major issues detected