Reverse-Flash-Kamen

Even after adding a variable aggregation node, the streaming output still cannot be used normally.

So, does that mean I have to give up streaming output if I want to use parallelism in the workflow? @dosu

It's not that using if-else nodes causes the streaming output to switch to blocking mode. Now, as long as I use parallelism, this issue will be reproduced. @dosu

I merged complex parallel operations into the workflow and called the workflow in Chatflow, which allows normal use of streaming output. I found that the current version supports passing parameters...

My new problem is with the Variable Aggregator node – if you want to use streaming output, you must not use this node in Chatflow.

https://github.com/langgenius/dify/discussions/20413

> 你可以参考这个 [#19324](https://github.com/langgenius/dify/issues/19324) The Dify team has supported these modifications in v1.4.0, but I still can't use streaming output currently.

> [@Yingjie-Zhao](https://github.com/Yingjie-Zhao) this issue [#19891](https://github.com/langgenius/dify/issues/19891) might offer a workaround, whereby you "package" complex behavior that includes parallelism as its own workflow (think of it as a "function" in traditional programming),...

> > [@Yingjie-Zhao](https://github.com/Yingjie-Zhao) this issue [#19891](https://github.com/langgenius/dify/issues/19891) might offer a workaround, whereby you "package" complex behavior that includes parallelism as its own workflow (think of it as a "function" in traditional...

Maybe you used an API that does not support streaming output, as I know some APIs' default model is blocking.