The document parsing is stuck and not moving.

[zephyershen]

I want to know if there are any differences in the update content of v0.18.0 between the v0.18.0 full and v0.18.0 slim versions? I'm currently using the v0.18.0 full version to parse documents, but it keeps getting stuck.

[dosubot[bot]]

Hello @zephyershen! I'm Dosu, your friendly bot here to help you solve bugs, answer questions, and assist with contributions while you wait for a human maintainer. Let's get started!

The differences between the v0.18.0 full and v0.18.0 slim versions are mainly in the inclusion of built-in embedding models and Python packages. The full version includes these models and allows downloading additional ones through the RAGFlow UI, resulting in a larger image size of approximately 9 GB. In contrast, the slim version does not include any embedding models or Python packages and is about 2 GB in size. The slim version is designed to be lightweight, suitable for environments with limited resources, while the full version provides a more comprehensive setup for embedding and deep document tasks ^[1][2][3].

If your document parsing is stuck, you might want to check the following troubleshooting steps:

1. Restart the parsing process by clicking the red cross beside the 'parsing status' bar.
2. Check the log of your RAGFlow server to ensure it is running properly:

   docker logs -f ragflow-server

3. Verify if the task_executor.py process exists.
4. Ensure your RAGFlow server can access hf-mirror.com or huggingface.com, as internet connectivity is required for document parsing.
5. If the parsing stalls near completion without any error, it might be due to insufficient RAM. Increase your memory allocation by adjusting the MEM_LIMIT value in docker/.env and restart your RAGFlow server:

   docker compose stop
   docker compose up -d

6. Check the RAGFlow logs for any errors:

   tail -f ragflow/docker/ragflow-logs/*.log

These steps should help diagnose and resolve the issue with document parsing ^[1][4][5].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Bug Report | Other}

[zephyershen]

He will eventually succeed in parsing, but it's extremely slow!!!!!!!!!!!!!! Super slow!!!!!!!!!! @dosu

[dosubot[bot]]

To speed up the parsing process, consider these recommendations:

1. Use GPU: Utilize a GPU to reduce embedding time, which can significantly speed up the process.

2. Disable RAPTOR: Turn off the "Use RAPTOR to enhance retrieval" option in your knowledge base configuration, as it can be time-consuming ^[1].

3. Avoid Knowledge Graph Extraction: If not necessary, disable knowledge graph extraction (GraphRAG), as it is time-consuming.

4. Disable Auto-keyword and Auto-question: These features depend on the LLM and can slow down the process. Disable them in your knowledge base configuration.

5. Choose Naive Document Parser: If your document is a plain text PDF and does not require GPU-intensive processes like OCR, TSR, or DLA, select "Naive" over "DeepDoc" or other large model options in the Document parser dropdown. This will substantially reduce document parsing time ^[1].

Implementing these changes should help reduce the time taken for the data processing operation. If you continue to experience issues, it might be helpful to loop in a maintainer for further assistance.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Bug Report | Other}

[zephyershen]

I would like to know if these updates are also available in the v0.18.0 slim version.
I used v0.18.0 slim to parse documents very quickly, but when I switched to v0.18.0 full, the parsing speed became very slow, so I urgently need to know whether the updated v0.18.0 slim version can fully enjoy these features. @dosu

[zephyershen]

What does the following error mean? Is it the key factor that affects my slow document parsing speed? @dosu

==> ragflow-logs/task_executor_541ed847b197_0.log <==
2025-05-15 15:30:23,398 ERROR 19 set_progress(e5965d94315011f0b033e6274861cc01), progress: -1, progress_msg: 15:30:23 Page(1100000001): [Canceled], got exception
Traceback (most recent call last):
File "/ragflow/rag/svr/task_executor.py", line 176, in set_progress
raise TaskCanceledException(msg)
TaskCanceledException: 15:30:23 Page(1100000001): [Canceled]

[dosubot[bot]]

The error message indicates that a TaskCanceledException was raised in the set_progress method of the task_executor.py file. This exception suggests that the task was canceled, which could be a key factor affecting your document parsing speed. Task cancellations can interrupt the parsing process, leading to incomplete or stalled operations.

To address this, you might want to investigate why the task was canceled. Check for any conditions or configurations that might trigger task cancellations, such as resource limitations or specific settings in your RAGFlow setup. Additionally, ensure that your system has sufficient resources allocated and that there are no external factors causing the task to be canceled.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Bug Report | Other}

[zephyershen]

So, except for the lack of built-in embedding model, there is no functional difference between v0.18.0 full and v0.18.0 slim, right? That is, the accuracy of the reference knowledge base and the ability to deeply analyze PDF images are both available?

[zephyershen]

@dosu

[dosubot[bot]]

The main difference between the v0.18.0 full and slim versions is the absence of built-in embedding models in the slim version. This affects the efficiency and accuracy of tasks related to the reference knowledge base, as embedding models are crucial for generating vector representations of text used in retrieval-augmented generation tasks. The slim version may rely on external or user-provided embedding models, which could lead to variations in performance depending on their quality and compatibility ^[1][2][3].

Regarding the ability to deeply analyze PDF images, I couldn't find specific information on whether the slim version supports the same level of deep analysis as the full version. The full version is designed to provide comprehensive capabilities for embedding and document tasks, which might include advanced PDF analysis features. If deep analysis of PDF images is critical for your use case, you might want to verify this aspect further.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Bug Report | Other}

[ZhenhangTung]

@zephyershen Could you share the Docker logs?
If the issue persists, feel free to open a GitHub issue here — we'd be happy to take a closer look!

[zephyershen]

OK,I may upload a log in the afternoon，Thank u

[ZhenhangTung]

How is it going?