[fix] 修复了sensevoice STT重复下载模型文件和缺失依赖项的问题

[xiewoc]

修复了重复下载模型文件都问题，并更改了模型文件储存位置，便于移除

解决了缺失依赖项的问题，自动检查并通过PipInstaller安装

并且加入了FFmpeg依赖缺失的检测和提示

// 这些在旧版的guide里面有体现，但是新版删除了这些

Modifications / 改动点

astrbot/core/provider/source/sensevoice_selfhost_source.py

实现功能同上

• This is NOT a breaking change. / 这不是一个破坏性变更。

Screenshots or Test Results / 运行截图或测试结果

能用了，新版模型加载如图

（图不知道为什么传不上来）

---

Checklist / 检查清单

• 😊 If there are new features added in the PR, I have discussed it with the authors through issues/emails, etc.
  / 如果 PR 中有新加入的功能，已经通过 Issue / 邮件等方式和作者讨论过。

• 👀 My changes have been well-tested, and "Verification Steps" and "Screenshots" have been provided above.
  / 我的更改经过了良好的测试，并已在上方提供了“验证步骤”和“运行截图”。

• 🤓 I have ensured that no new dependencies are introduced, OR if new dependencies are introduced, they have been added to the appropriate locations in requirements.txt and pyproject.toml.
  / 我确保没有引入新依赖库，或者引入了新依赖库的同时将其添加到 requirements.txt 和 pyproject.toml 文件相应位置。

• 😮 My changes do not introduce malicious code.
  / 我的更改没有引入恶意代码。

Summary by Sourcery

Improve SenseVoice self-hosted STT setup by making dependency handling automatic and model storage persistent.

New Features:

• Add automatic detection and installation of missing SenseVoice Python dependencies.
• Add FFmpeg availability checks with actionable error messages before model initialization.

Bug Fixes:

• Prevent repeated SenseVoice model downloads by persisting and reusing the model in the AstrBot data directory.
• Fix SenseVoice loading and inference when optional dependencies are unavailable at module import time.

Enhancements:

• Load SenseVoice integrations lazily and reuse cached modules during provider operations.
• Improve model initialization, inference validation, and error reporting.

[sourcery-ai]

Hey - I've found 3 issues

Prompt for AI Agents

Please address the comments from this code review:

## Individual Comments

### Comment 1
<location path="astrbot/core/provider/sources/sensevoice_selfhosted_source.py" line_range="85" />
<code_context>
+            loop = asyncio.get_event_loop()
+            if loop.is_running():
+                logger.warning("检测到运行中的事件循环，自动安装可能失败。建议手动安装依赖。")
+            asyncio.run(_install_dependencies())
+        except Exception as e:
+            logger.error(f"自动安装依赖失败: {e}")
</code_context>
<issue_to_address>
**issue (bug_risk):** `asyncio.run(_install_dependencies())` raises `RuntimeError` whenever SenseVoice is initialized through the normal async provider lifecycle, because `initialize()` already runs inside an active event loop. The exception is caught and reported as dependency-installation failure, so automatic installation never works when dependencies are missing.

**Triggers:** When any SenseVoice dependency is not installed before provider initialization.

**Suggested fix:** Await `_install_dependencies()` directly from an async initialization path instead of calling `asyncio.run()`.
</issue_to_address>

### Comment 2
<location path="astrbot/core/provider/sources/sensevoice_selfhosted_source.py" line_range="130" />
<code_context>
+            loop = asyncio.get_running_loop()
+            await loop.run_in_executor(
+                None,
+                lambda: snapshot_download("iic/SenseVoiceSmall", local_dir=self.model_path),
+            )

</code_context>
<issue_to_address>
**issue (bug_risk):** The configured `stt_model` is ignored: the provider stores it through `set_model(...)`, but initialization always downloads `iic/SenseVoiceSmall` and loads the fixed `self.model_path` instead of using `self.model_name`. Configuring another ModelScope model therefore still downloads and loads SenseVoiceSmall.

**Triggers:** When a provider configuration specifies an `stt_model` other than `iic/SenseVoiceSmall`.

**Suggested fix:** Use `self.model_name` for the ModelScope download and derive the local model directory consistently from the configured model.
</issue_to_address>

### Comment 3
<location path="astrbot/core/provider/sources/sensevoice_selfhosted_source.py" line_range="126-131" />
<code_context>
+        SenseVoiceSmall, _, snapshot_download = _load_sense_voice_modules()
+
+        # 模型下载（同步操作放入线程池）
+        if not os.path.exists(os.path.join(self.model_path, "configuration.json")):
+            loop = asyncio.get_running_loop()
+            await loop.run_in_executor(
+                None,
+                lambda: snapshot_download("iic/SenseVoiceSmall", local_dir=self.model_path),
+            )

-        # 将模型加载放到线程池中执行
</code_context>
<issue_to_address>
**issue (bug_risk):** The existence of `configuration.json` is treated as proof that the model download completed. If a previous download was interrupted after creating that file, initialization skips `snapshot_download` and passes the incomplete directory to `SenseVoiceSmall`, causing model loading to fail instead of resuming or repairing the download.

**Triggers:** When the model directory contains `configuration.json` from a partial or corrupted download.

**Suggested fix:** Validate the complete expected model file set, or let `snapshot_download` resume/validate the existing local directory before loading.

```suggestion
        loop = asyncio.get_running_loop()
        await loop.run_in_executor(
            None,
            lambda: snapshot_download("iic/SenseVoiceSmall", local_dir=self.model_path),
        )
```
</issue_to_address>

Sourcery assessment

Needs a human reviewer. 3 findings to address first, and if the lazy installation or model-path logic is wrong, initialization can fail or install incompatible packages and leave downloaded model files or environment changes behind after a revert. These effects are bounded and repairable by removing the model/cache or restoring the dependency environment, but they are not undone automatically by reverting the code.

Blocking findings: astrbot/core/provider/sources/sensevoice_selfhosted_source.py:85, astrbot/core/provider/sources/sensevoice_selfhosted_source.py:130, astrbot/core/provider/sources/sensevoice_selfhosted_source.py:131

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}

[sourcery-ai]

issue (bug_risk): asyncio.run(_install_dependencies()) raises RuntimeError whenever SenseVoice is initialized through the normal async provider lifecycle, because initialize() already runs inside an active event loop. The exception is caught and reported as dependency-installation failure, so automatic installation never works when dependencies are missing.

Triggers: When any SenseVoice dependency is not installed before provider initialization.

Suggested fix: Await _install_dependencies() directly from an async initialization path instead of calling asyncio.run().

[sourcery-ai]

issue (bug_risk): The configured stt_model is ignored: the provider stores it through set_model(...), but initialization always downloads iic/SenseVoiceSmall and loads the fixed self.model_path instead of using self.model_name. Configuring another ModelScope model therefore still downloads and loads SenseVoiceSmall.

Triggers: When a provider configuration specifies an stt_model other than iic/SenseVoiceSmall.

Suggested fix: Use self.model_name for the ModelScope download and derive the local model directory consistently from the configured model.

[sourcery-ai]

issue (bug_risk): The existence of configuration.json is treated as proof that the model download completed. If a previous download was interrupted after creating that file, initialization skips snapshot_download and passes the incomplete directory to SenseVoiceSmall, causing model loading to fail instead of resuming or repairing the download.

Triggers: When the model directory contains configuration.json from a partial or corrupted download.

Suggested fix: Validate the complete expected model file set, or let snapshot_download resume/validate the existing local directory before loading.

Suggested change

	if not os.path.exists(os.path.join(self.model_path, "configuration.json")):
	loop = asyncio.get_running_loop()
	await loop.run_in_executor(
	None,
	lambda: snapshot_download("iic/SenseVoiceSmall", local_dir=self.model_path),
	)
	loop = asyncio.get_running_loop()
	await loop.run_in_executor(
	None,
	lambda: snapshot_download("iic/SenseVoiceSmall", local_dir=self.model_path),
	)