Document extraction tool node stuck in running state

Discussion
7

You’ve already helped narrow down the scope very clearly:

  • Official cloud
  • Document Extractor
  • PDF works normally
  • Word (even empty documents) consistently causes a freeze

This basically indicates a new issue with the ‘Word parsing pipeline’ on the cloud side, rather than a problem with your account or workflow configuration.

Currently, there are two things you can do: one is a temporary workaround, and the other is to provide complete information to facilitate official investigation.

1. The Most Stable Temporary Solution

Before the official fix, I suggest you:

  1. First, convert Word to PDF or TXT locally

    • Save Word as .pdf / .txt
    • Upload PDF / TXT in the Document Extractor, instead of the original .docx
    • You’ve already verified that PDF can be processed normally, which can ensure your operations aren’t blocked.

  2. If you have a small amount of content, you can also:

    • Directly copy the Word text content and paste it into ‘User Input’, instead of uploading a file (suitable for when there isn’t much content).

2. I Suggest You Add a “For Official Review” Message to Your Post

You just said:

PDF format works, but Word documents don’t. Hope the officials fix it soon.

You can write it a bit ‘more like a bug reproduction description’, for example, by replying with (you can copy and modify this):

Environment: Official Dify Cloud
Reproduction Steps:

  1. Create the simplest workflow: Start → Document Extractor (upload file) → Output text
  2. Upload an empty or any content Word (.docx) file
    Expected: Node execution completes, outputting the extracted text
    Actual: The Document Extractor node remains “running” indefinitely, without error
    Comparison: Under the same workflow, uploading a small PDF file completes normally

This description makes it easier for official staff to ‘reproduce it directly’ and is more likely to be treated as a platform bug.

3. How to Follow Up

  • Since this is a cloud-side issue, you cannot see the backend logs and can only wait for the official operations/development team to investigate.
  • You can continue to use the “Word → PDF/TXT then upload” method as a temporary measure.
  • If an official staff member replies to the post later with a fix progress or timeline, you can also @ me, and I can help you confirm if your existing workflow configuration needs further adjustment.

If you’d like, I can also help you refine a “bilingual Chinese-English” reproduction description, which would be convenient for you to submit to support or use on other channels later.