[Bug]: Embedding index rebuild fails with 500 Internal Server Error (ReMe instance is None)

[justsavo]

QwenPaw Version

2.2.0b5 (Windows desktop installer, QwenPaw-Tauri-2.2.0b5-Windows-setup.exe)

Description

Clicking "Rebuild Memory Index" in the memory panel consistently returns:

HTTP 500 Internal Server Error

The underlying traceback shows the failure is caused by self._reme being None inside ReMe's light-memory integration — not by the embedding provider configuration.

Backtrace (from qwenpaw-desktop.log):

File "...\qwenpaw\app\routers\agents.py", line 1340, in rebuild_agent_memory_index
File "...\qwenpaw\agents\memory\reme_light_memory_manager.py", line 816, in rebuild_index
File "...\qwenpaw\agents\memory\reme_embedding.py", line 141, in rebuild_index
File "...\qwenpaw\utils\io_utils.py", line 142, in run_async_to_completion
File "...\qwenpaw\agents\memory\reme_embedding.py", line 210, in _prepare_embedding_reindex
File "...\qwenpaw\agents\memory\reme_light_memory_manager.py", line 390, in _require_embedding_rebuild
AttributeError: 'NoneType' object has no attribute 'update_component'

Related PR(s): #7441 (bumps ReMe to 0.4.1.11), #7133 (embedding reindex scoped)

Component(s) Affected

• Core / Backend (app, agents, config, providers, utils, local_models)
• Console (frontend web UI)
• Channels (DingTalk, Feishu, QQ, Discord, iMessage, etc.)
• Skills
• CLI
• Documentation (website)
• Tests
• CI/CD
• Scripts / Deploy

Environment

• QwenPaw version: 2.2.0b5
• OS: Windows 10 (AMD64)
• Install method: Windows desktop installer (.exe, Tauri)
• Python version (if applicable): backend embeds Python 3.11 (PyInstaller one-dir)

Steps to Reproduce

1. Install QwenPaw 2.2.0b5 on Windows via the setup exe.
2. Open the console UI -> Memory panel.
3. Click "Rebuild Memory Index" (or call POST /api/agents/default/memory/reindex).
4. Observe the 500 Internal Server Error.

Expected: the embedding index is rebuilt successfully (or a meaningful message).

Actual vs Expected

• Actual: 500 Internal Server Error — AttributeError: 'NoneType' object has no attribute 'update_component' at reme_light_memory_manager.py:390.
• Expected: Embedding index rebuild succeeds.

Observations (supporting the root cause)

• POST /api/agents/default/memory/reindex -> 500.
• GET /api/agents/default/memory/status -> 503 ("ReMe is not started or status reporting is unavailable").
• GET /api/agents/default/memory/runtime-status -> worker: idle, auto_memory.enabled: true, but embedding_reindex_required: true.
• Embedding config is valid and reachable (SiliconFlow OpenAI-compatible endpoint, BAAI/bge-m3, dim=1024). The failure is not a provider/network issue.
• In the installed bundle, ...\qwenpaw-backend\_internal\reme\ contains only 17 resource files (config YAML, tokenizer stopwords, steps/.../*.yaml) and 0 .py/.pyd/.pyc files. This suggests the ReMe light-memory core (@_reme instance) is not initialized at runtime, which matches self._reme = None.

Note: This is a high-confidence inference (log evidence + _reme is None) but I could not fully confirm whether reme-ai is simply missing from the bundle vs. a contract/version mismatch (_REQUIRED_REME_VERSION). Maintainer confirmation would help.

Logs / Screenshots

File "...\qwenpaw\agents\memory\reme_light_memory_manager.py", line 390, in _require_embedding_rebuild
AttributeError: 'NoneType' object has no attribute 'update_component'

Additional Notes

• Workaround: none found yet; memory index rebuild is blocked as long as self._reme is None.

[github-actions]

Welcome to QwenPaw! 🐾

Hi @justsavo, this is your 3rd issue.

We'll review your issue soon. Thank you for supporting QwenPaw!

---

Tip

⭐ If you find QwenPaw useful, please give us a Star!

Staying ahead

Star QwenPaw on GitHub and be instantly notified of new releases.

Your star helps more developers discover this project! 🐾

[samluoabc]

同样遇到这个问题
更新了qwenpaw的v2.2.0b5以后，在设置-运行配置-长期记忆-长期记忆中心里，显示：
1、无法读取 ReMe 运行状态
ReMe is not started or status reporting is unavailable - {"detail":"ReMe is not started or status reporting is unavailable"}
2、确认重建 BM25 索引？
将根据当前记忆分块重新生成关键词搜索索引。——显示记忆索引重建失败：ReMe is not started or the reindex job failed - {"detail":"ReMe is not started or the reindex job failed"}

[jinliyl]

Thank you for the detailed report and especially for including the traceback and packaged-file observations. They were very helpful in tracing the failure back to the desktop bundle.

The fix has been submitted in #7458. The root cause is that ReMe's auto-fin and daily-paper plugins are discovered through Python entry points and loaded dynamically, while the PyInstaller spec did not include their modules, manifests, or distribution metadata. ReMe initialization therefore failed and left _reme unset; the reindex path then exposed the secondary NoneType.update_component error.

The PR adds the missing plugin packaging, guards reindex when ReMe is unavailable, and adds an installed-desktop BM25 reindex check to prevent this regression from recurring.

Thanks again for reporting this so clearly!