Can the datasets in the model-config interface be dynamically assigned values?

[qwe104]

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 the datasets in the model-config interface be dynamically assigned values? It is necessary to dynamically set which knowledge base to use for retrieval. That is, it is already clearly known which knowledge base to use for the retrieval.

[dosubot[bot]]

Hi @qwe104! I'm Dosu and I’m helping the dify team.

Yes, datasets in the model-config interface can be dynamically assigned values. The Dify frontend uses components like DatasetConfig and ParamsConfig to let you select, update, and manage datasets interactively. These components update the dataset configuration in real time using functions such as setDatasetConfigs, handleSave, and handleMetadataModelChange, which means you can programmatically or interactively set which knowledge base (dataset) to use for retrieval at runtime. The backend then uses this configuration to determine which datasets to query for retrieval, supporting both single and multiple dataset strategies, and applies any metadata filtering or reranking settings you specify. This allows you to dynamically control which knowledge base is used for each retrieval operation, as long as you provide the correct dataset IDs and config in your model configuration payloads.
See code reference

To reply, just mention @dosu.

---

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

[qwe104]

Thank you for your response. I would like to dynamically pass the knowledge base to be retrieved within the Agent API.

[dosubot[bot]]

Yes, you can dynamically specify which knowledge base (dataset) to use for retrieval in the Agent API. This is handled by passing the relevant dataset IDs and retrieval configuration at runtime to the backend, specifically to the DatasetRetrieval class. When you call the Agent API, include the desired dataset IDs in the config (such as in the dataset_configs or DatasetEntity), and the system will use those for retrieval. This works for both single and multiple dataset strategies, and you can also apply dynamic metadata filtering or reranking as needed. The retrieval logic supports runtime-driven selection, so you can switch knowledge bases per request without redeploying or statically configuring the app. For more details, see how DatasetRetrieval and related components handle this in the code: example, workflow node example.

If you need a code snippet or more details on the API payload, let me know!

To reply, just mention @dosu.

---

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

[qwe104]

Thank you for your response. Ultimately, I intended to pass the knowledge base ID within the Agent API. I did pass the knowledge base ID, but it didn't have any effect.

[qwe104]

I'm not sure if the knowledge base ID will be able to be transmitted within the Agent API in the future.