I’m working on a workflow where I’m using Gmail Trigger → Text Classifier → Gemini (Google AI) model. I can see the Gemini node runs fine and returns output, but I’m stuck on how to use that output data in the next node.
For example, in my case I’d like to pass the Gemini response into another node (like Function, Set, or Router), but I’m not sure which JSON path/expression to use, or the best way to reference the Gemini output.
What’s the correct way to access the Gemini node’s response and make it available in following nodes? Any sample expressions or examples would be super helpful.
What’s your use case for needing that data? You can’t reference sub-nodes via expressions so you’d need to build the object manually, but that would be fairly straightforward as true will always be whatever the name of the logic branch you’re on.
I’m not sure you actually need it, if you explain what you need it for I can probably suggest a better solution
I’d like to add something, as I worked with something very similar..
The “Text Classifier” is essentially a specialized version of the “Basic LLM Chain” with a custom Prompt, custom Categories (branch outputs), and other options and the most important thing is the Structured Output, where you can define your own JSON Schema.
To replicate the “Text Classifier” while retaining the flexibility to design your own JSON Schema for use in your workflow, simply inspect and use the Text Classifier system prompt (or create your own) and the JSON Schema (or create your own), Then just use the Basic LLM Chain node with that prompt and with Structured Output enabled with that Schema..
The results are the same, it’s all about the prompt and the JSON schema.
