langgenius / dify

Dify is an open-source LLM app development platform. Dify's intuitive interface combines AI workflow, RAG pipeline, agent capabilities, model management, observability features and more, letting you quickly go from prototype to production.
https://dify.ai
Other
51.56k stars 7.45k forks source link

The configuration of the same model was updated on the workbench, but the dialog configuration did not take effect. #10613

Closed dadastory closed 1 day ago

dadastory commented 1 day ago

Self Checks

Dify version

0.11.1

Cloud or Self Hosted

Self Hosted (Docker)

Steps to reproduce

I tried to modify a certain model parameter on the workbench, such as gpt-4o, modify parameters such as temperature or top k, max token, etc. After publishing and updating, the configuration in the dialog did not take effect, and I could only create a new dialog. At this time, configure will take effect

✔️ Expected Behavior

After the configuration of the same model is modified or the model is replaced, the historical conversations can be modified and updated synchronously to ensure a better user experience.

❌ Actual Behavior

This problem is very unfriendly to users who have preset parameter variables, because they must re-add a new conversation and repeatedly fill in the topic variables of the previous conversation, and modifications to some problematic conversation parameter configurations cannot take effect immediately.

crazywoola commented 1 day ago

I tried to modify a certain model parameter on the workbench, such as gpt-4o, modify parameters such as temperature or top k, max token, etc. After publishing and updating, the configuration in the dialog did not take effect, and I could only create a new dialog. At this time, configure will take effect

Yes the configuration only takes effect for the new conversation. In another word, this is an expected behavior, see this as well https://github.com/langgenius/dify/issues/10599#issuecomment-2470553368