Workflow node reference
Classifier
Sort messages into categories automatically. Exact configuration, outputs, execution behavior, errors, limits, and examples for the Classifier workflow node.
What this guide covers
- Category: AI.
- Use it to map text into one of your named categories and continue through the matching dynamic path.
- 3 configuration fields and 3 documented outputs.
- Classification is probabilistic and category quality depends on instructions/examples.
What this node does
Sort messages into categories automatically
Use it to map text into one of your named categories and continue through the matching dynamic path.
When not to use it
Do not use it when an exact keyword or deterministic condition is sufficient.
Prerequisites
- At least two clear categories
- Descriptions and representative examples
- Input text
- A fallback/human path
At least two clear categories. Descriptions and representative examples. Input text. A fallback/human path
Inputs and configuration fields
| Field | Key | Control / type | Requirement | Default | Accepted values and behavior |
|---|---|---|---|---|---|
Message to Classify | input | variable-input | Required when shown | {{context.message.content}} | The message to sort into categories. Value accepted by the displayed control. Variables supported |
Categories | categories | category-builder | Required when shown | No default | Define the categories to sort messages into. Value accepted by the displayed control |
Additional Context | instructions | textarea | Optional | No default | Help the AI understand your business context. This is for a restaurant booking service... |
Configure the node
- Drag Classifier from AI in the node palette.
- Select the node to open its configuration drawer.
- Complete every required field and any conditional field revealed by your choices.
- Insert variables only from the current trigger or an earlier node on the same path.
- Connect every required output handle to the intended next step.
- Save and run Workflow Test with representative data before activation.
Outputs
Later nodes can use these values only after this node has run successfully on the same execution path.
| Variable | Type | Meaning |
|---|---|---|
{{nodes.classifier.category_id}} | string | ID of the matched category |
{{nodes.classifier.category_name}} | string | Name of the matched category |
{{nodes.classifier.confidence}} | number | How confident the AI is (0-1) |
Execution behavior
- Classifies input using configured categories.
- Emits category ID/name, confidence, and reasoning fields defined by the node.
- Creates dynamic handles for categories and runs for real during Test.
AI nodes run for real during Workflow Test and can consume billable AI usage. Do not assume an automatic retry unless the execution record explicitly shows one.
Example use case
Classify an inbound message as sales, support, billing, or human_help.
Message Received → Classifier → AI Response / Human HandoffCommon errors
| Symptom | Resolution |
|---|---|
Wrong category wins | Make categories mutually distinguishable and add examples. |
A category path is missing | Connect every generated category handle. |
Confidence is too low | Route uncertain results to Human Handoff. |
Limitations
Classification is probabilistic and category quality depends on instructions/examples.
It is not a substitute for authorization or compliance checks.
Classification is probabilistic and category quality depends on instructions/examples. It is not a substitute for authorization or compliance checks.