Using the knowledge base as context, the bot will do not respond.

[louis-redotpay]

Self Checks

• [X] This is only for bug report, if you would like to ask a question, please head to Discussions.
• [X] I have searched for existing issues search for existing issues, including closed ones.
• [X] I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
• [X] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）
• [X] Please do not modify this template :) and fill in all the required fields.

Dify version

0.6.14

Cloud or Self Hosted

Cloud

Steps to reproduce

The app ID:cfabe94d-1fa7-4c20-8f71-debc27f99ac6 Test Conversation ID1: a93d66f3-eb67-4f1e-a412-66b6fce2877d Used the knowledge base, the bot did not respond. Test Conversation ID2: da3b78af-831a-4348-be53-517c6e12a3e7 Deleted the knowledge base, the bot responded. This situation only occurred on 7/16; it was working normally before that.I am not sure if it is related to upgrading to version 0.6.14.If the official team could conduct an inspection, I would be very grateful.

✔️ Expected Behavior

No response

❌ Actual Behavior

No response

[dosubot[bot]]

Hey @louis-redotpay! 👋 I'm here to help you with bug reports, answer questions, and assist you while waiting for a human maintainer. Let's tackle this together!

The Dify bot might not be responding when using the knowledge base in version 0.6.14 due to issues in the RAG (Retrieval-Augmented Generation) pipeline. When the knowledge base is deleted, the bot does not rely on this pipeline and thus responds normally. The problem could be related to the integration or retrieval process of the knowledge base documents.

Additionally, the recall mode settings might be affecting the bot's performance. The N-Choose-1 Recall mode relies heavily on the reasoning capability of the model and the clarity of the knowledge base description. If these factors are not optimal, it can lead to issues in retrieving the correct information. On the other hand, the Multi-Path Recall mode, which is recommended, does not depend on the reasoning capability of the model or the knowledge base description and can achieve higher quality recall effects when retrieving from multiple knowledge bases.

To improve the bot's performance, consider configuring the recall mode to Multi-Path Recall and ensuring that the Rerank model is properly set up. This mode searches for text content related to the user question across all knowledge bases and reorders the retrieved documents based on semantic relevance, which can help in providing more accurate responses ^[1][2][3].

To continue talking to Dosu, mention @dosu.

[crazywoola]

We are investigating this issue.

[crazywoola]

It has been fixed now.