mirror of
https://github.com/langgenius/dify-official-plugins.git
synced 2026-07-21 17:45:23 -04:00
The volcengine_maas plugin version cannot be downgraded. Downgrading will cause Dify to freeze/crash. #864
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @bilibilipala on GitHub (Dec 10, 2025).
Self Checks
Dify version
1.10.1
Plugin version
0.0.37
Cloud or Self Hosted
Self Hosted (Docker)
Steps to reproduce
The version of Dify I am using is 1.10.1. I want to downgrade the Volcengine plugin to version 0.0.25, because the latest version of the plugin always causes the Doubao large model to experience hallucinations. However, after installing the 0.0.25 plugin, Dify becomes unresponsive/frozen.
✔️ Error log
2025/12/11 02:38:02 run.go:147: [INFO]plugin langgenius/volcengine_maas:0.0.25 started
2025/12/11 02:38:03 run.go:135: [ERROR]plugin langgenius/volcengine_maas:0.0.25 exited with error: exit status 1
age/cwd/langgenius/volcengine_maas-0.0.25@69d479387c274399cdf967854fa2eab8b2ece3c1b28c702724402913639dcc24/.venv/lib/python3.12/site-packages/volcenginesdkarkruntime/_client.py", line 15, in
File "/app/storage/cwd/langgenius/volcengine_maas-0.0.25@69d479387c274399cdf967854fa2eab8b2ece3c1b28c702724402913639dcc24/.venv/lib/python3.12/site-packages/volcenginesdkarkruntime/_models.py", line 31, in
File "/app/storage/cwd/langgenius/volcengine_maas-0.0.25@69d479387c274399cdf967854fa2eab8b2ece3c1b28c702724402913639dcc24/.venv/lib/python3.12/site-packages/volcenginesdkarkruntime/_utils/__init__.py", line 1, in
File "/app/storage/cwd/langgenius/volcengine_maas-0.0.25@69d479387c274399cdf967854fa2eab8b2ece3c1b28c702724402913639dcc24/.venv/lib/python3.12/site-packages/volcenginesdkarkruntime/_utils/_utils.py", line 23, in
ModuleNotFoundError: No module named 'sniffio'
2025/12/11 02:38:03 run.go:185: [INFO]plugin langgenius/volcengine_maas:0.0.25 stopped
@dosubot[bot] commented on GitHub (Dec 10, 2025):
Hi @bilibilipala! I'm Dosu and I’m helping the dify-official-plugins team.
Downgrading the volcengine_maas plugin to version 0.0.25 on Dify 1.10.1 causes a crash because the plugin's Python environment is missing the required
sniffiodependency. This happens even thoughsniffiois listed in the plugin's requirements, because each plugin runs in its own isolated virtual environment inside Docker, and dependencies sometimes aren't installed as expected during downgrades or manual installs. Installingsniffioglobally in the container won't help—each plugin needs it inside its own.venvenvironment.Workaround:
.venvdirectory:pipis missing, install it:This should resolve the missing module error and unfreeze Dify.
Reference 1 | Reference 2 | Reference 3
Permanent fix:
Add
sniffio(and any other missing dependencies) to the plugin'srequirements.txt, then rebuild or restart the plugin so dependencies are installed automatically in the future.Reference
This issue is specific to Dify 1.10.1 and does not appear in earlier versions like 1.9.2, likely due to changes in how dependencies are handled in newer Dify/plugin daemon versions. Keep an eye on plugin updates and changelogs for a more robust fix upstream.
To reply, just mention @dosu.
How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other
@nadazero commented on GitHub (Dec 11, 2025):
新版本解决了缺少库的bug ModuleNotFoundError: No module named 'sniffio', 你这一降级等于又把问题重新复现了。需要手动在虚拟环境安装库了。
@nadazero commented on GitHub (Dec 11, 2025):
另外大模型幻觉和插件版本有关系?不知道这个结论是怎么得出的
@bilibilipala commented on GitHub (Dec 11, 2025):
出于某些原因,我们正在测试从dify 1.4.0 升级至1.10.0,我们的AI应用在0.0.25的火山引擎插件上工作一切正常,但升级到0.0.37之后一切都变了,在同样的输出参数下,豆包模型几乎每次都会重复已输出的段落很多次,这极大影响了模型输出效率,我试着调整存在惩罚和频率惩罚参数,但这并未奏效,最终我只好降级到现有可以使用的稳定版本,所以我就遇到的这个问题,我也很想知道为什么大模型幻觉会与插件版本有关系,但我降级完成后,一切又都正常了
@nadazero commented on GitHub (Dec 11, 2025):
应该是插件版本变更中代码影响某些参数的变化导致输出发生了变化,但具体太细节了的确不好排查。
先降级应付着吧。
或者换个思路,重复已输出的段落很多次这种问题以前也遇到过,本质也和模型能力有关,换个更强大的模型试试。
@bilibilipala commented on GitHub (Dec 11, 2025):
感谢您的回复,由于更强大的模型也会导致更慢的响应速率,我们的应用对速率有一定的要求,所以暂时没办法更换,但今天我们也测试了千问的高速率大模型,测试效果不错,后期考虑也接入进来做下冗余
@dosubot[bot] commented on GitHub (Dec 27, 2025):
Hi, @bilibilipala. I'm Dosu, and I'm helping the dify-official-plugins team manage their backlog and am marking this issue as stale.
Issue Summary:
Next Steps:
Thank you for your understanding and contribution!