PluginInvokeError

[zfdsapoiu]

Self Checks

• I have read the Contributing Guide and Language Policy.
• This is only for bug report, if you would like to ask a question, please head to Discussions.
• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report, otherwise it will be closed.
• 【中文用户 & Non English User】请使用英语提交，否则会被关闭 ：）
• Please do not modify this template :) and fill in all the required fields.

Dify version

cloud

Cloud or Self Hosted

Cloud

Steps to reproduce

Hi I got a problem and I can't fix it.

I just want to test the video capability in dify

✔️ Expected Behavior

When I try image input or I transfor the video to audio, everything is fine

❌ Actual Behavior

But when I try to input a video to a LLM module(I tried Gemini several models and qwen vl models) and all give such error:

req_id: f3ca8fc150 PluginInvokeError: {"args":null,"error_type":"PluginDaemonInnerError","message":"encountered an error: invalid character '\u003c' looking for beginning of value status: 502 Bad Gateway original response: \u003chtml\u003e"}

[dosubot]

Hi @zfdsapoiu! I'm Dosu and I’m helping the dify team.

This error happens because when you send a video file to Gemini or Qwen VL models in Dify Cloud, the backend plugin daemon returns an HTML error page (502 Bad Gateway) instead of the expected JSON. This usually means the plugin or model backend can't handle video input, even though image and audio inputs work fine. The error message about "invalid character '<'" is a direct result of trying to parse this HTML as JSON, which fails and triggers the PluginInvokeError you see.

This is a known issue with Gemini and Qwen VL plugins, especially when the plugin expects a certain response format or endpoint and gets something unexpected (like an error page) instead. For Gemini, issues often relate to using the wrong endpoint or unsupported modalities, and similar errors have been reported by others when using video input or certain model versions. Workarounds that sometimes help for Gemini include switching to non-preview models, using the us-central1 region, or trying Gemini 2.5 flash-lite, but there is no guarantee these will resolve video input issues specifically. Full support for video input may require future plugin updates or backend changes citation 1 citation 2.

Right now, Dify's backend does convert video files into the correct prompt format for supported models, but if the model or plugin doesn't support video, you'll get this kind of server-side failure citation 3 citation 4. There is no explicit handling for video output in the workflow, only for images.

If you need to process video, converting it to audio or extracting frames as images (as you've already done) is the most reliable workaround for now. For true video support, keep an eye out for plugin updates or announcements about expanded multimodal capabilities.

To reply, just mention @dosu.

---

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