[Question]: RagFlow API: Document Parsed Successfully but Not Searchable in Knowledge Base

[gzmasterpulse]

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 (Language Policy).
• Non-english title submitions will be closed directly ( 非英文标题的提交将会被直接关闭 ) (Language Policy).
• Please do not modify this template :) and fill in all the required fields.

Describe your problem

After I call the RagFlow API to upload a document and the parsing succeeds, I cannot search for it in the knowledge base.In the RagFlow UI, I can view the document chunks successfully, and refreshing to rebuild the vectors in the UI also completes successfully. However, during retrieval testing and search, no knowledge base information can be matched. At first, I suspected it was due to caching, but even after several days the issue persists. I then restarted the ragflow-server Docker container, but it had no effect. Could this be a RagFlow bug or caused by something else? The RagFlow version I’m using is the latest nightly build.

[dosubot]

Hi @gzmasterpulse! I'm Dosu and I’m helping the ragflow team.

This issue has come up before, especially around recent RagFlow versions and Docker images. If you’re using the latest nightly build, make sure you’re running the full nightly Docker image—not nightly-slim—since retrieval bugs have been fixed only in the full nightly release for some setups (source, source, source).

If you’re already on the correct image, here’s a quick troubleshooting checklist:

• Double-check that the embedding model is properly configured and matches across your knowledge base and retrieval/chat pipeline (source, source).
• Make sure the system prompt for dialog/chat includes the {knowledge} variable, and that the Retrieval component is linked to the correct knowledge base (source).
• Review your similarity threshold and top_k settings—they may be too restrictive for your queries (source).
• Confirm that all document chunks are fully parsed, embedded, and indexed (the UI showing chunks doesn’t guarantee they’re in the vector store for retrieval) (source, source).
• Check for any errors or warnings in the RagFlow server logs during parsing, embedding, or indexing (source).

If all configs look correct and you’re still seeing no results, try rebuilding the knowledge base vectors again and restarting the server. If the problem persists, reviewing logs for silent failures during embedding or indexing is the next step.

Let me know if you need help with any specific config or log details!

To reply, just mention @dosu.

---

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

[gzmasterpulse]

This is the configuration I’m using, and I’ve already checked all the possible causes you mentioned—there are no errors. If I reinstall RagFlow, the first processing works without any issues, but subsequent calls or adding documents directly through the UI fail to work properly.

[KevinHuSh]

Try another query or uploading and parsing via web UI.

[gzmasterpulse]

Try another query or uploading and parsing via web UI.
Only right after installing RagFlow, the first upload through the UI works. For subsequent uploads, even though the document parsing completes, the data still cannot be added to the knowledge base for retrieval. At first, I thought this issue only occurred when uploading via the API, but later I found the same problem happens when uploading through the UI as well.

[gzmasterpulse]

[ERROR]Generate embedding error:{"code":20042,"message":"input must have less than 512 tokens","data":null}
Sometimes this kind of error is reported

[mileskui]

I have also encountered the same problem, where one knowledge base can be searched while the other cannot

[KevinHuSh]

[ERROR]Generate embedding error:{"code":20042,"message":"input must have less than 512 tokens","data":null} Sometimes this kind of error is reported

What's the embedding inference service? What about switching to another embedding service?