Help！"Run Failed: 'unknown'"!

[DiaJB]

Self Checks

• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
• [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:)
• Please do not modify this template :) and fill in all the required fields.

1. Is this request related to a challenge you're experiencing? Tell me about your story.

In chatflow, when an LLM node continuously outputs a long piece of text, its streaming output process can be observed on the frontend. However, once it finishes outputting, the content disappears and an error is reported: "Run Failed: 'unknown'".My Dify instance is deployed via Docker. I have tried modifying certain parameters in .envand docker-compose.yaml—including those related to text length limits and TIMEOUT settings—but none of these adjustments resolved the issue.

2. Additional context or comments

No response

[dosubot[bot]]

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

The "Run Failed: 'unknown'" error after long LLM outputs in chatflow is almost always caused by an exception during streaming or post-processing in the backend. If the exception message is empty or generic, the UI just shows "unknown" as the error. This is not directly related to text length or timeout settings—those changes typically won't fix it.

Common causes include: plugin or model backend returning an unexpected response (like None instead of an iterable), response parsing errors, or the streaming HTTP connection being closed prematurely (e.g., ChunkedEncodingError) source. Some plugins or models (like DeepSeek V3.1 with tongyi, Qwen with openai_api_compatible) have known issues with output loss or incomplete streaming due to how they handle streamed output source.

Here’s what you can do:

• Enable DEBUG and ENABLE_REQUEST_LOGGING in your .env for Dify, plugin daemon, and model server to capture detailed logs.
• Check your Docker and proxy/server settings: increase timeouts (GUNICORN_TIMEOUT, HTTP_REQUEST_MAX_READ_TIMEOUT), and make sure NGINX has proxy_buffering off; and proxy_http_version 1.1; source.
• If possible, try disabling streaming for the problematic model/plugin (set enable_stream or stream to false in the plugin config or code) to see if non-streaming mode avoids the error.
• Try a different LLM provider/model to see if the issue is model- or plugin-specific.
• If you’re using a plugin, check for updates or known bugs in its output handling logic—sometimes patching the plugin to use append instead of replace for streamed output fixes output loss source.
• For some users, setting the stream_mode_delimiter to \n in the config solved similar issues with empty or failed outputs source.

If the error persists, collect the backend logs at the time of failure—look for exceptions in the LLM node streaming or post-processing path. This will help pinpoint the root cause and whether it’s a plugin/model bug or an infrastructure issue.

To reply, just mention @dosu.

---

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