How to continue a test run without restart a new conversation?

[WilliamChen-luckbob]

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.

1. Is this request related to a challenge you're experiencing? Tell me about your story.

There is a scenario where I need to have a step-by-step dialogue with an agent.

Different LLMs are selected via a decision-maker for processing, and the process has strict sequentiality.

In such cases, during testing, I need to repeatedly test a specific branch starting from a certain node in the dialogue.

Does Dify support such an operation? Currently, every time I finish debugging the prompt, I have to restart the entire complete dialogue flow, which is a huge waste of tokens.

Do you have any proper test methods for this situation?

2. Additional context or comments

No response

[dosubot[bot]]

Dify does not currently provide a user-facing feature to continue a test run from a specific node in a dialogue flow without restarting the entire conversation. Multiple users have requested this feature for workflow debugging and to avoid wasting tokens, but it is not yet implemented in the UI or as an official workflow feature [issue 14115].

Under the hood, Dify's workflow engine does support pausing and resuming execution from specific nodes by saving and restoring the runtime state, including outputs and paused nodes [graph_engine.py, graph_runtime_state.py]. However, this capability is not exposed for manual use through the UI or public APIs.

For now, the best available options are:

• Use node-level debugging and variable inspection to test and debug individual nodes, but this does not allow resuming the entire flow from an arbitrary point.
• If you are comfortable with backend development, you could potentially leverage the workflow engine's state serialization and restoration features to implement a custom solution for partial test runs, but this would require significant engineering effort and is not officially documented or supported.

If step-by-step or branch-specific testing is critical for your workflow, you may want to follow or contribute to the related feature requests in the repository [issue 14115, issue 12141].

To reply, just mention @dosu.

---

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