How to sequentially assign multiple LLM outputs to variables in a workflow?

[diracsea]

In the current workflow, the LLM outputs multiple parameters, but the input for the next node can only accept the LLM’s full text output as a single variable. Calling the large model again with a parameter simulator is too time-consuming. Besides a parameter extractor, is there any faster way to obtain these multiple output parameters? Currently, the temporary solution is to add a code node after the LLM to parse the output using Python, based on the Dify JSON response standard format. However, writing a separate Python script for parsing each output is too complex.

[cheng]

@diracsea

The LLM outputs multiple parameters, but the actual input for the next node can only take the LLM text as the entire variable input. Calling the large model again for a parameter simulator is too time-consuming. Besides a parameter extractor, is there any faster way to get the multiple output parameters? The current temporary solution is to add a code node after the LLM and parse it with Python, based on the Dify JSON response standard format. However, writing a separate Python script for each output parsing is too complex.

Using Python for parsing is the fastest method…