Change the embedding model when retrieving after finishing embedding knowledge bases.

[mengze1320]

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 (我已阅读并同意 Language Policy).
• [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:)
• Please do not modify this template :) and fill in all the required fields.

Content

Can I change the embedding model when I retrieving the knowledge base that has been embedded by another embedding model?
I only find the entrance of changing rerank model.

[dosubot[bot]]

You can't change the embedding model just for retrieval after a knowledge base has already been embedded with a different model in Dify. Retrieval always uses the embeddings generated by the current embedding model set for the knowledge base. If you want to use a new embedding model, you need to change it in the knowledge base settings—this will automatically trigger re-embedding of all documents with the new model. The rerank model is the only retrieval-time option you can swap independently; embedding models are tightly coupled to the stored vectors for mathematical reasons and enforced by the system for data integrity. If you change the embedding model, give the system 30–60 minutes for all chunks and metadata to update before running retrieval tests, as the re-embedding process runs in the background and may not show progress indicators. For more details, see this discussion and code logic.

To reply, just mention @dosu.

---

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