RAGFlow Workspace Code Commit ID and Image Version

Oct 13, 2025 by thesweetdesigns.com 1 views

Introduction

We have encountered a critical issue where the chat model fails to intervene when using the "Table" mode to parse Excel files. This results in direct output of data without relevant responses to user queries. In this report, we will outline the steps to reproduce the issue, provide expected behavior, and offer additional information to aid in debugging and resolution.

Is There an Existing Issue for the Same Bug?

After conducting a thorough search, we have confirmed that there is no existing issue for this specific bug.

RAGFlow Workspace Code Commit ID and Image Version

The RAGFlow workspace code commit ID is g555c7067, and the image version is g555c7067 (v0.17.0-29).

Actual Behavior

When using the "Table" mode to parse Excel spreadsheets, we have observed that the chat model fails to intervene. The output is direct data, and the responses are not relevant to the questions asked. For example, when inquiring about "broccoli," the output was "xishengcai" (which is incorrect).

Example Use Case

To illustrate this issue, we have included a series of images demonstrating the problem:

Incorrect Response: When asked about "broccoli," the output was "xishengcai" (which is incorrect).
Search Test: We have confirmed that "broccoli" exists in the data, and the "search test" can accurately locate it.
Comparison with General Mode: For comparison, we tried parsing the same Excel file using the "General" mode, with all other settings (including the prompt) kept the same. The response is shown in the figure below:

Expected Behavior

We expect the chat model to intervene and provide responses of the same high quality as in the 'General' mode.

Steps to Reproduce

To reproduce this issue, follow these steps:

Create a Knowledge Base: Configure the "Table" mode.
Import Excel Spreadsheet: Parse the Excel file.
Create a Chatbot: Select the knowledge base and the deepseek-r1 model.
Ask a Question: Inquire about a topic, such as "broccoli."

Additional Information

We have not received any responses to this issue.

Conclusion

Q: What is the issue with the chat model in "Table" mode for Excel file parsing?

A: The chat model fails to intervene when using the "Table" mode to parse Excel files. This results in direct output of data without relevant responses to user queries.

Q: What are the symptoms of this issue?

A: The symptoms of this issue include:

Direct output of data without relevant responses to user queries
Incorrect responses to user queries (e.g., "xishengcai" instead of "broccoli")
Failure of the chat model to intervene in the response generation process

Q: What are the steps to reproduce this issue?

A: To reproduce this issue, follow these steps:

Create a Knowledge Base: Configure the "Table" mode.
Import Excel Spreadsheet: Parse the Excel file.
Create a Chatbot: Select the knowledge base and the deepseek-r1 model.
Ask a Question: Inquire about a topic, such as "broccoli."

Q: What is the expected behavior of the chat model in "Table" mode?

A: The expected behavior of the chat model in "Table" mode is to intervene and provide responses of the same high quality as in the 'General' mode.

Q: What are the differences between "Table" mode and "General" mode?

A: The main difference between "Table" mode and "General" mode is the parsing method used to process the Excel file. "Table" mode uses a table-based approach, while "General" mode uses a more general-purpose approach.

Q: Why does the chat model fail to intervene in "Table" mode?

A: The exact reason for the chat model's failure to intervene in "Table" mode is not yet clear. However, it is suspected that the issue may be related to the character matching process, where only the character "西" (xi) is matched, even though the prompt emphasizes the need for an exact match.

Q: How can this issue be resolved?

A: To resolve this issue, we recommend the following:

Investigate the character matching process and ensure that it is working correctly.
Review the parsing method used in "Table" mode and consider alternative approaches.
Test the chat model's intervention in "Table" mode with a variety of user queries and Excel files.

Q: What is the impact of this issue on users?

A: The impact of this issue on users is significant, as it results in direct output of data without relevant responses to user queries. This can lead to frustration and disappointment, particularly if users are relying on the chat model for accurate and helpful responses.

Q: How can users work around this issue?

A: Users can work around this issue by:

Using the "General" mode instead of "Table" mode.
Modifying their prompts to emphasize the need for an exact match.
Providing additional context or information to help the chat model understand the user's query.