langgenius/dify
1
0
Fork 0
mirror of https://github.com/langgenius/dify.git synced 2026-07-21 03:55:25 -04:00
Code Issues 775 Packages Projects Releases 10 Wiki Activity

The configured knowledge base list have different embedding model, please set reranking model - Error Message #21957

New Issue
Open
opened 2026-02-21 20:15:01 -05:00 by yindo · 5 comments
yindo commented 2026-02-21 20:15:01 -05:00
Owner
Copy Link
Copy Source

Originally created by @aref-gif on GitHub (Jan 28, 2026).

Self Checks

  • I have read the Contributing Guide and Language Policy.
  • This is only for bug report, if you would like to ask a question, please head to Discussions.
  • I have searched for existing issues search for existing issues, including closed ones.
  • I confirm that I am using English to submit this report, otherwise it will be closed.
  • 【中文用户 & Non English User】请使用英语提交，否则会被关闭 ：）
  • Please do not modify this template :) and fill in all the required fields.

Dify version

Version1.11.4

Cloud or Self Hosted

Cloud

Steps to reproduce

We are facing an issue where we cannot use more than one Knowledge Base in the same studio, and I believe the problem stems from the Knowledge Base being deactivated due to inactivity for 30 days.

The issue: The configured knowledge base list have different embedding model, please set reranking model.

✔️ Expected Behavior

We should be able to use all knowledge bases in any studio without complications.

❌ Actual Behavior

An error message appears stating that the model is incompatible, but we have verified that all procedures are correct.

Originally created by @aref-gif on GitHub (Jan 28, 2026). ### Self Checks - [x] I have read the [Contributing Guide](https://github.com/langgenius/dify/blob/main/CONTRIBUTING.md) and [Language Policy](https://github.com/langgenius/dify/issues/1542). - [x] This is only for bug report, if you would like to ask a question, please head to [Discussions](https://github.com/langgenius/dify/discussions/categories/general). - [x] I have searched for existing issues [search for existing issues](https://github.com/langgenius/dify/issues), including closed ones. - [x] I confirm that I am using English to submit this report, otherwise it will be closed. - [x] 【中文用户 & Non English User】请使用英语提交，否则会被关闭 ：） - [x] Please do not modify this template :) and fill in all the required fields. ### Dify version Version1.11.4 ### Cloud or Self Hosted Cloud ### Steps to reproduce We are facing an issue where we cannot use more than one Knowledge Base in the same studio, and I believe the problem stems from the Knowledge Base being deactivated due to inactivity for 30 days. The issue: The configured knowledge base list have different embedding model, please set reranking model. ### ✔️ Expected Behavior We should be able to use all knowledge bases in any studio without complications. ### ❌ Actual Behavior An error message appears stating that the model is incompatible, but we have verified that all procedures are correct.
yindo added the 🐞 bugcloud labels 2026-02-21 20:15:01 -05:00
yindo commented 2026-02-21 20:15:02 -05:00
Author
Owner
Copy Link
Copy Source

@dosubot[bot] commented on GitHub (Jan 28, 2026):

Hi @aref-gif! I'm Dosu and I’m helping the dify team.

The error happens because Dify requires all knowledge bases used together to have the same embedding model and provider. If they differ, you must explicitly set a reranking model in your retrieval node settings—otherwise, retrieval fails with this error. This is enforced in both the backend and UI for data integrity and result compatibility (code reference).

A reranking model reorders or scores the combined results from multiple knowledge bases to ensure relevance, even if their embeddings are not directly comparable. You configure this by selecting both a reranking provider and model name in the retrieval settings. Make sure these settings are saved and that the reranking model is available in your environment (example).

To resolve the error, you have two options:

  1. Align all knowledge bases to use the same embedding model and provider (requires re-embedding documents).
  2. Set a reranking model in the retrieval node: choose a reranking provider and model (like Cohere or BAAI/bge-reranker-v2-m3), and ensure reranking_enable is true. This enables Dify to merge and rerank results across different embedding models (discussion).

If you recently changed embedding models or reactivated a knowledge base, allow time for re-embedding to complete—during this period, retrieval may fail or show errors, but this is unrelated to the reranking requirement (details).

If seamless multi-KB usage is critical, keep an eye on future Dify updates, as this area is actively being improved.

To reply, just mention @dosu.

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

Chat with Dosu Open in Cursor Join Discord Share on X

@dosubot[bot] commented on GitHub (Jan 28, 2026): <!-- Greeting --> Hi @aref-gif! I'm [Dosu](https://go.dosu.dev/dosubot) and I’m helping the dify team. <!-- Answer --> The error happens because Dify requires all knowledge bases used together to have the same embedding model and provider. If they differ, you must explicitly set a reranking model in your retrieval node settings—otherwise, retrieval fails with this error. This is enforced in both the backend and UI for data integrity and result compatibility ([code reference](https://github.com/langgenius/dify/blob/e98c1adfbf8b94adede7cb6abd98ee8ca222ddf9/api/core/rag/retrieval/dataset_retrieval.py#L68-L1502)). A reranking model reorders or scores the combined results from multiple knowledge bases to ensure relevance, even if their embeddings are not directly comparable. You configure this by selecting both a reranking provider and model name in the retrieval settings. Make sure these settings are saved and that the reranking model is available in your environment ([example](https://github.com/langgenius/dify/pull/30549)). To resolve the error, you have two options: 1. Align all knowledge bases to use the same embedding model and provider (requires re-embedding documents). 2. Set a reranking model in the retrieval node: choose a reranking provider and model (like Cohere or BAAI/bge-reranker-v2-m3), and ensure `reranking_enable` is true. This enables Dify to merge and rerank results across different embedding models ([discussion](https://github.com/langgenius/dify/discussions/31571)). If you recently changed embedding models or reactivated a knowledge base, allow time for re-embedding to complete—during this period, retrieval may fail or show errors, but this is unrelated to the reranking requirement ([details](https://github.com/langgenius/dify/issues/19866#issuecomment-2888493548)). If seamless multi-KB usage is critical, keep an eye on future Dify updates, as this area is actively being improved. <!-- Dosu Comment Footer --> *To reply, just mention [@dosu](https://go.dosu.dev/dosubot).* --- <sup>How did I do? [Good](https://app.dosu.dev/response-feedback/01d7c6ce-c129-4b4b-b349-d816558e5e35?feedback_type=great_response) | [Irrelevant](https://app.dosu.dev/response-feedback/01d7c6ce-c129-4b4b-b349-d816558e5e35?feedback_type=irrelevant_answer) | [Incorrect](https://app.dosu.dev/response-feedback/01d7c6ce-c129-4b4b-b349-d816558e5e35?feedback_type=incorrect_sources) | [Verbose](https://app.dosu.dev/response-feedback/01d7c6ce-c129-4b4b-b349-d816558e5e35?feedback_type=too_verbose) | [Hallucination](https://app.dosu.dev/response-feedback/01d7c6ce-c129-4b4b-b349-d816558e5e35?feedback_type=hallucination) | [Report 🐛](https://app.dosu.dev/response-feedback/01d7c6ce-c129-4b4b-b349-d816558e5e35?feedback_type=bug_report) | [Other](https://app.dosu.dev/response-feedback/01d7c6ce-c129-4b4b-b349-d816558e5e35?feedback_type=other)</sup> [![Chat with Dosu](https://dosu.dev/dosu-chat-badge.svg)](https://app.dosu.dev/6088a4f4-073a-43f3-b3d4-4f06fb0c05a0/ask?utm_source=github)&nbsp;[![Open in Cursor](https://dosu.dev/dosu-in-cursor.svg)](https://cursor.com/link/prompt?text=The%20error%20happens%20because%20Dify%20requires%20all%20knowledge%20bases%20used%20together%20to%20have%20the%20same%20embedding%20model%20and%20provider.%20If%20they%20differ%2C%20you%20must%20explicitly%20set%20a%20reranking%20model%20in%20your%20retrieval%20node%20settings%E2%80%94otherwise%2C%20retrieval%20fails%20with%20this%20error.%20This%20is%20enforced%20in%20both%20the%20backend%20and%20UI%20for%20data%20integrity%20and%20result%20compatibility%20%28%5Bcode%20reference%5D%28https%3A//github.com/langgenius/dify/blob/e98c1adfbf8b94adede7cb6abd98ee8ca222ddf9/api/core/rag/retrieval/dataset_retrieval.py%23L68-L1502%29%29.%0A%0AA%20reranking%20model%20reorders%20or%20scores%20the%20combined%20results%20from%20multiple%20knowledge%20bases%20to%20ensure%20relevance%2C%20even%20if%20their%20embeddings%20are%20not%20directly%20comparable.%20You%20configure%20this%20by%20selecting%20both%20a%20reranking%20provider%20and%20model%20name%20in%20the%20retrieval%20settings.%20Make%20sure%20these%20settings%20are%20saved%20and%20that%20the%20reranking%20model%20is%20available%20in%20your%20environment%20%28%5Bexample%5D%28https%3A//github.com/langgenius/dify/pull/30549%29%29.%0A%0ATo%20resolve%20the%20error%2C%20you%20have%20two%20options%3A%0A1.%20Align%20all%20knowledge%20bases%20to%20use%20the%20same%20embedding%20model%20and%20provider%20%28requires%20re-embedding%20documents%29.%0A2.%20Set%20a%20reranking%20model%20in%20the%20retrieval%20node%3A%20choose%20a%20reranking%20provider%20and%20model%20%28like%20Cohere%20or%20BAAI/bge-reranker-v2-m3%29%2C%20and%20ensure%20%60reranking_enable%60%20is%20true.%20This%20enables%20Dify%20to%20merge%20and%20rerank%20results%20across%20different%20embedding%20models%20%28%5Bdiscussion%5D%28https%3A//github.com/langgenius/dify/discussions/31571%29%29.%0A%0AIf%20you%20recently%20changed%20embedding%20models%20or%20reactivated%20a%20knowledge%20base%2C%20allow%20time%20for%20re-embedding%20to%20complete%E2%80%94during%20this%20period%2C%20retrieval%20may%20fail%20or%20show%20errors%2C%20but%20this%20is%20unrelated%20to%20the%20reranking%20requirement%20%28%5Bdetails%5D%28https%3A//github.com/langgenius/dify/issues/19866%23issuecomment-2888493548%29%29.%0A%0AIf%20seamless%20multi-KB%20usage%20is%20critical%2C%20keep%20an%20eye%20on%20future%20Dify%20updates%2C%20as%20this%20area%20is%20actively%20being%20improved.)&nbsp;[![Join Discord](https://img.shields.io/badge/join-5865F2?logo=discord&logoColor=white&label=)](https://go.dosu.dev/discord-bot)&nbsp;[![Share on X](https://img.shields.io/badge/X-share-black)](https://twitter.com/intent/tweet?text=%40dosu_ai%20helped%20me%20solve%20this%20issue!&url=https%3A//github.com/langgenius/dify/issues/31665)
yindo commented 2026-02-21 20:15:02 -05:00
Author
Owner
Copy Link
Copy Source

@aref-gif commented on GitHub (Jan 28, 2026):

I'm sure that all knowledge bases use the same embedding model and provider.

So we need another solution

@aref-gif commented on GitHub (Jan 28, 2026): I'm sure that all knowledge bases use the same embedding model and provider. So we need another solution
yindo commented 2026-02-21 20:15:02 -05:00
Author
Owner
Copy Link
Copy Source

@dosubot[bot] commented on GitHub (Jan 28, 2026):

If all your knowledge bases use the same embedding model and provider but you still see this error, it's likely due to stale or inconsistent metadata in the backend or workflow graph. This can happen if a previous update didn't fully sync, or if a knowledge base was deactivated/reactivated and the metadata wasn't refreshed. Here’s what you can try:

  1. Open each knowledge base, re-select and save the embedding model and provider—even if they appear correct—to force a metadata refresh.
  2. Re-save or re-publish your workflow/studio to update the workflow graph with the latest knowledge base metadata.
  3. Remove and re-add the affected knowledge bases to the studio, then re-save the workflow. This forces a full metadata and workflow sync.
  4. Clear your browser cache or use incognito mode to rule out frontend caching issues.
  5. Check for any orphaned or deleted provider references in your workflow or knowledge base configuration.
  6. If the issue persists, re-upload files or recreate the knowledge bases to trigger backend re-indexing and metadata refresh (not ideal, but it can resolve hidden sync issues)
  7. If none of these steps work, there may be a deeper backend data inconsistency or bug. Collect backend logs for further investigation.

These steps address the most common causes of this error when embedding models actually match, based on how Dify stores and syncs knowledge base metadata and workflow configuration (source).

To reply, just mention @dosu.

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

Chat with Dosu Open in Cursor Join Discord Share on X

@dosubot[bot] commented on GitHub (Jan 28, 2026): <!-- Answer --> If all your knowledge bases use the same embedding model and provider but you still see this error, it's likely due to stale or inconsistent metadata in the backend or workflow graph. This can happen if a previous update didn't fully sync, or if a knowledge base was deactivated/reactivated and the metadata wasn't refreshed. Here’s what you can try: 1. Open each knowledge base, re-select and save the embedding model and provider—even if they appear correct—to force a metadata refresh. 2. Re-save or re-publish your workflow/studio to update the workflow graph with the latest knowledge base metadata. 3. Remove and re-add the affected knowledge bases to the studio, then re-save the workflow. This forces a full metadata and workflow sync. 4. Clear your browser cache or use incognito mode to rule out frontend caching issues. 5. Check for any orphaned or deleted provider references in your workflow or knowledge base configuration. 6. If the issue persists, re-upload files or recreate the knowledge bases to trigger backend re-indexing and metadata refresh (not ideal, but it can resolve hidden sync issues) 7. If none of these steps work, there may be a deeper backend data inconsistency or bug. Collect backend logs for further investigation. These steps address the most common causes of this error when embedding models actually match, based on how Dify stores and syncs knowledge base metadata and workflow configuration ([source](https://github.com/langgenius/dify/blob/24ebe2f5c6f31c5be430531ca1a02f83af35c51e/api/services/dataset_service.py#L355-L777)). <!-- Dosu Comment Footer --> *To reply, just mention [@dosu](https://go.dosu.dev/dosubot).* --- <sup>How did I do? [Good](https://app.dosu.dev/response-feedback/3777a28a-3f6e-430b-96be-54313b7d8672?feedback_type=great_response) | [Irrelevant](https://app.dosu.dev/response-feedback/3777a28a-3f6e-430b-96be-54313b7d8672?feedback_type=irrelevant_answer) | [Incorrect](https://app.dosu.dev/response-feedback/3777a28a-3f6e-430b-96be-54313b7d8672?feedback_type=incorrect_sources) | [Verbose](https://app.dosu.dev/response-feedback/3777a28a-3f6e-430b-96be-54313b7d8672?feedback_type=too_verbose) | [Hallucination](https://app.dosu.dev/response-feedback/3777a28a-3f6e-430b-96be-54313b7d8672?feedback_type=hallucination) | [Report 🐛](https://app.dosu.dev/response-feedback/3777a28a-3f6e-430b-96be-54313b7d8672?feedback_type=bug_report) | [Other](https://app.dosu.dev/response-feedback/3777a28a-3f6e-430b-96be-54313b7d8672?feedback_type=other)</sup> [![Chat with Dosu](https://dosu.dev/dosu-chat-badge.svg)](https://app.dosu.dev/6088a4f4-073a-43f3-b3d4-4f06fb0c05a0/ask?utm_source=github)&nbsp;[![Open in Cursor](https://dosu.dev/dosu-in-cursor.svg)](https://cursor.com/link/prompt?text=If%20all%20your%20knowledge%20bases%20use%20the%20same%20embedding%20model%20and%20provider%20but%20you%20still%20see%20this%20error%2C%20it%27s%20likely%20due%20to%20stale%20or%20inconsistent%20metadata%20in%20the%20backend%20or%20workflow%20graph.%20This%20can%20happen%20if%20a%20previous%20update%20didn%27t%20fully%20sync%2C%20or%20if%20a%20knowledge%20base%20was%20deactivated/reactivated%20and%20the%20metadata%20wasn%27t%20refreshed.%20Here%E2%80%99s%20what%20you%20can%20try%3A%0A%0A1.%20Open%20each%20knowledge%20base%2C%20re-select%20and%20save%20the%20embedding%20model%20and%20provider%E2%80%94even%20if%20they%20appear%20correct%E2%80%94to%20force%20a%20metadata%20refresh.%0A2.%20Re-save%20or%20re-publish%20your%20workflow/studio%20to%20update%20the%20workflow%20graph%20with%20the%20latest%20knowledge%20base%20metadata.%0A3.%20Remove%20and%20re-add%20the%20affected%20knowledge%20bases%20to%20the%20studio%2C%20then%20re-save%20the%20workflow.%20This%20forces%20a%20full%20metadata%20and%20workflow%20sync.%0A4.%20Clear%20your%20browser%20cache%20or%20use%20incognito%20mode%20to%20rule%20out%20frontend%20caching%20issues.%0A5.%20Check%20for%20any%20orphaned%20or%20deleted%20provider%20references%20in%20your%20workflow%20or%20knowledge%20base%20configuration.%0A6.%20If%20the%20issue%20persists%2C%20re-upload%20files%20or%20recreate%20the%20knowledge%20bases%20to%20trigger%20backend%20re-indexing%20and%20metadata%20refresh%20%28not%20ideal%2C%20but%20it%20can%20resolve%20hidden%20sync%20issues%29%20%20%0A7.%20If%20none%20of%20these%20steps%20work%2C%20there%20may%20be%20a%20deeper%20backend%20data%20inconsistency%20or%20bug.%20Collect%20backend%20logs%20for%20further%20investigation.%0A%0AThese%20steps%20address%20the%20most%20common%20causes%20of%20this%20error%20when%20embedding%20models%20actually%20match%2C%20based%20on%20how%20Dify%20stores%20and%20syncs%20knowledge%20base%20metadata%20and%20workflow%20configuration%20%28%5Bsource%5D%28https%3A//github.com/langgenius/dify/blob/24ebe2f5c6f31c5be430531ca1a02f83af35c51e/api/services/dataset_service.py%23L355-L777%29%29.)&nbsp;[![Join Discord](https://img.shields.io/badge/join-5865F2?logo=discord&logoColor=white&label=)](https://go.dosu.dev/discord-bot)&nbsp;[![Share on X](https://img.shields.io/badge/X-share-black)](https://twitter.com/intent/tweet?text=%40dosu_ai%20helped%20me%20solve%20this%20issue!&url=https%3A//github.com/langgenius/dify/issues/31665)
yindo commented 2026-02-21 20:15:02 -05:00
Author
Owner
Copy Link
Copy Source

@aref-gif commented on GitHub (Jan 29, 2026):

Thanks for the detailed suggestions. I went through all the steps mentioned.

However, the issue is still persisting and the same error continues to appear. It seems like it may be related to a deeper backend metadata inconsistency or bug. Could you please advise on the next steps or help us investigate further from the backend/logs side?

@aref-gif commented on GitHub (Jan 29, 2026): Thanks for the detailed suggestions. I went through all the steps mentioned. However, the issue is still persisting and the same error continues to appear. It seems like it may be related to a deeper backend metadata inconsistency or bug. Could you please advise on the next steps or help us investigate further from the backend/logs side?
yindo commented 2026-02-21 20:15:02 -05:00
Author
Owner
Copy Link
Copy Source

@aref-gif commented on GitHub (Feb 2, 2026):

Dear Dify Team,

I hope you are doing well.

I am writing to follow up regarding the reported issue and to check if there are any updates on the resolution progress.

We would appreciate your support in addressing this matter at your earliest convenience.

Thank you for your assistance, and I look forward to your response.

@aref-gif commented on GitHub (Feb 2, 2026): Dear Dify Team, I hope you are doing well. I am writing to follow up regarding the reported issue and to check if there are any updates on the resolution progress. We would appreciate your support in addressing this matter at your earliest convenience. Thank you for your assistance, and I look forward to your response.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
Knowledge Pipeline
Proposal
WIP
cloud
dependencies
feat:webapp
good first issue
outdated
performance
pull-request
Mirrored from GitHub Pull Request
 python:uv
refactor
size:L
stale
status: accepting prs
status: blocked
status: in discussion
stt & tts
vulnerablity
web
wont-fix
🌊 feat:workflow
🌚 invalid
🐞 bug
👻 feat:rag
💪 enhancement
📚 documentation
🔨 feat:tools
🙋‍♂️ question
🙏 help wanted
🤔 cant-reproduce
🤖 feat:agent
No Label cloud 🐞 bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: langgenius/dify#21957
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?