Classifier Node#
Reads the input and picks one of the classes you defined, then continues only along the branch connected to that class. Each class can connect to its own next node.The Classifier node is in the Logic / Transform group of the node list.
The model reads the input and picks one of the Classes. Only the next node connected to the chosen class is executed.
You can create 1–20 classes and drag to reorder them.
Model#
Choose the model to use for classification in Model.
Only models enabled for classification are listed. Custom input is not available.
If Vision appears next to a model name, that model can also classify images.
The model list may differ by region and over time.
Classifying images with Vision#
Selecting a model that supports Vision shows the Use Vision option.
With Use Vision on, the model also looks at files uploaded in the chat when classifying. Excel and CSV formats are excluded from vision processing.
Turning on Vision incurs image token costs. Documents are billed per page, so the cost grows with the volume.
Choose the variable that holds the content to classify in Input variable. Example: query
Classification instructions (optional) holds the criteria the model should follow when classifying.
Classes#
Use Add class to create a class, then enter a Name (required) and a Description (optional).
The Classes section shows 1–20 classes.
The Other class is shown in the list by default. At least one class other than Other is required.
Each class can connect to its own next node.
Set what happens when the input matches no class.
Send to Other (default): sends the input to the Other class.
Send to a chosen class: pick a Target class, and input that matches no class goes to that class.
Output variables#
The classifier result can be used in other nodes through Insert variable.
Selected class number: the number of the class the model picked
Selected class name: the name of the class the model picked
Is Other: whether the input matched no class and went to Other (true/false)
The execution log records the Selected class, the model, and usage (tokens).
Examples#
Split customer inquiries into shipping delay / defect or damage / wrong or missing item / exchange or return / refund or payment / other, and connect a different response node to each
Classify an uploaded image as dog / cat / bird / other and generate an answer for that class
Put the classification result (the Selected class name value) into an LLM node prompt so only the answer for the chosen class is generated
Modified at 2026-11-18 09:04:11