[Bug]v4.22.1后 Whisper (API) 无法被正常调用 STT失效

[200303]

What happened / 发生了什么

v4.22.1后 发送语音后不会正常解析成文本

把版本回退到 v4.22.1 和 vv4.22.0后能正常解析

Reproduce / 如何复现？

AstrBot版本 v4.22.2/v4.22.3
STT模型提供商 Whisper(API)
开启STT 选择模型提供商
发送语音后不会正常解析

AstrBot version, deployment method (e.g., Windows Docker Desktop deployment), provider used, and messaging platform used. / AstrBot 版本、部署方式（如 Windows Docker Desktop 部署）、使用的提供商、使用的消息平台适配器

Linux podman

OS

Linux

Logs / 报错日志

AstrBot v4.22.1 正常日志

[2026-04-11 20:23:08.744] [Core] [INFO] [core.event_bus:61]: [default] [ABot(aiocqhttp)] [会话]: [ComponentType.Record]
[2026-04-11 20:23:08.744] [Core] [DBUG] [waking_check.stage:157]: enabled_plugins_name: ['*']
[2026-04-11 20:23:08.970] [Core] [INFO] [sources.whisper_api_source:112]: Converting amr file to wav using convert_to_pcm_wav...
[2026-04-11 20:23:08.971] [Core] [DBUG] [utils.tencent_record_helper:73]: pyffmpeg 转换失败: No module named 'pyffmpeg', 尝试使用 ffmpeg 命令行进行转换
[2026-04-11 20:23:09.011] [Core] [INFO] [utils.tencent_record_helper:96]: [FFmpeg] stdout:
[2026-04-11 20:23:09.011] [Core] [DBUG] [utils.tencent_record_helper:97]: [FFmpeg] stderr: [amr @ 0x5941f60b02c0] Estimating duration from bitrate, this may be inaccurate
Input #0, amr, from '/AstrBot/data/temp/whisper_api_2edfd6a9.input':
Duration: 00:00:02.96, bitrate: 5 kb/s
Stream #0:0: Audio: amr_nb (amrnb) (samr / 0x726D6173), 8000 Hz, mono, fltp, 5 kb/s
Stream mapping:
Stream #0:0 -> #0:0 (amr_nb (amrnb) -> pcm_s16le (native))
Press [q] to stop, [?] for help
Output #0, wav, to '/AstrBot/data/temp/whisper_api_5201e777.wav':
Metadata:
ISFT : Lavf61.7.100
Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 24000 Hz, mono, s16, 384 kb/s
Metadata:
encoder : Lavc61.19.101 pcm_s16le
[out#0/wav @ 0x5941f60b1bc0] video:0KiB audio:232KiB subtitle:0KiB other streams:0KiB global headers:0KiB muxing overhead: 0.032762%
size= 233KiB time=00:00:04.96 bitrate= 384.1kbits/s speed=1.99e+03x
[2026-04-11 20:23:09.011] [Core] [INFO] [utils.tencent_record_helper:98]: [FFmpeg] return code: 0
[2026-04-11 20:23:09.113] [Core] [INFO] [preprocess_stage.stage:86]: 语音转文本结果: 能听得见我说话吗
[2026-04-11 20:23:09.113] [Core] [DBUG] [method.star_request:44]: plugin -> session_controller - handle_session_control_agent
[2026-04-11 20:23:09.113] [Core] [DBUG] [method.star_request:44]: plugin -> session_controller - handle_empty_mention
[2026-04-11 20:23:09.113] [Core] [DBUG] [method.star_request:44]: plugin -> astrbot - on_message
[2026-04-11 20:23:09.115] [Core] [DBUG] [agent_sub_stages.internal:166]: ready to request llm provider
[2026-04-11 20:23:09.115] [Core] [DBUG] [agent_sub_stages.internal:185]: acquired session lock for llm request
[2026-04-11 20:23:09.118] [Core] [DBUG] [core.astr_main_agent_resources:463]: [知识库] 使用全局配置，知识库数量: 0
[2026-04-11 20:23:09.118] [Core] [DBUG] [pipeline.context_utils:95]: hook(OnLLMRequestEvent) -> astrbot - decorate_llm_req
[2026-04-11 20:23:09.118] [Core] [DBUG] [pipeline.context_utils:95]: hook(OnLLMRequestEvent) -> astrbot-web-searcher - edit_web_search_tools
[2026-04-11 20:23:09.443] [Core] [DBUG] [runners.base:64]: Agent state transition: AgentState.IDLE -> AgentState.RUNNING
[2026-04-11 20:23:09.444] [Core] [DBUG] [runners.tool_loop_agent_runner:312]: [BefCompact] RunCtx.messages -> [2] system,user
[2026-04-11 20:23:09.444] [Core] [DBUG] [runners.tool_loop_agent_runner:312]: [AftCompact] RunCtx.messages -> [2] system,user

AstrBot v4.22.3 非正常日志

[2026-04-11 20:46:02.081] [Core] [INFO] [core.event_bus:61]: [default] [ABot(aiocqhttp)] [会话]: [ComponentType.Record]
[2026-04-11 20:46:02.082] [Core] [DBUG] [waking_check.stage:157]: enabled_plugins_name: ['*']

对 到这就没有下文了

Are you willing to submit a PR? / 你愿意提交 PR 吗？

• Yes!

Code of Conduct

• I have read and agree to abide by the project's Code of Conduct。

[dosubot]

Hi @200303! I'm Dosu and I’m helping the AstrBot team.

这个问题很可能是 v4.22.2 中 PR #6893 引入的回归 bug。

原因分析： PR #6893 在 session controller 的 empty_mention_waiter 中添加了空消息检查 [1]：

if not event.message_str or not event.message_str.strip():
    return

语音消息（Record）在到达 PreProcessStage 进行 STT 转换之前，event.message_str 是空的。这个空消息检查在 waking_check 之后、PreProcessStage 之前执行，导致语音消息被当作"空消息"直接过滤掉，STT 流程根本不会被触发。

这也解释了你的日志现象——v4.22.3 中处理到 waking_check.stage 之后就没有下文了。

可能的修复方向： 修改空消息检查，排除包含 Record 组件的消息：

has_record = any(isinstance(comp, Record) for comp in event.get_messages())
if (not event.message_str or not event.message_str.strip()) and not has_record:
    return

这样语音消息就能正常通过管道到达 PreProcessStage 进行 Whisper API 转写了。

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}