In chatflow, can the streaming output content from the LLM be captured and passed as a parameter to other nodes for continuous use?

[MannyEC]

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.

I need to continuously capture the streaming output and inspect it. This can be done either through a chatflow node or by writing a plugin.

2. Additional context or comments

No response

[ishita-0301]

Streaming interception would be really useful. There are a lot of cases where you want to inspect partial outputs for logging, validation, or triggering actions before the final response is complete.

I've also found that having visibility into the stream helps catch agent issues like repetitive output or getting stuck in loops. That's one reason I find projects like FailproofAI interesting since they focus on runtime reliability for AI agents: https://github.com/FailproofAI/failproofai.

[AaronZ345]

Not through a normal downstream Chatflow node. Dify does not execute Code, Template, Parameter Extractor, or other nodes once per LLM chunk. Those nodes consume the completed upstream value, so inserting one between the LLM and Answer/End turns that path into a completed-value path.

Streaming is preserved when the LLM text goes directly to Answer/End, or when the caller consumes the app's SSE response with response_mode: streaming. If you need to inspect every partial chunk, do it in the SSE client or a proxy around the Dify API. A plugin may emit a streaming variable, but that still does not make ordinary downstream nodes run continuously for each chunk.