Make AI Decisions

These actions answer structured questions about a given state, for use cases such as routing, triage, intent detection and scoring. They return typed answers with probabilities instead of free text.

Make AI Decisions Actions

Decide

Answers one or more typed questions about a state and, optionally, a conversation history. Event metadata fields applicable for this action are as follows:

Fields (table)

Field	Definition	Example	Default
Input Element	Json path for the input in request event payload	input	-
Output Element	Json path for the output in response event payload	output	-

Fields (JSON Schema)

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "type": "object",
  "properties": {
    "eventMeta": {
      "type": "object",
      "properties": {
        "inputElement": {
          "type": "string",
          "description": "Json path for the input in request event payload",
          "examples": ["input"],
          "default": null
        },
        "outputElement": {
          "type": "string",
          "description": "Json path for the output in response event payload",
          "examples": ["output"],
          "default": null
        }
      }
    }
  }
}

With event metadata parameters as:

Parameters (table)

Parameter	Definition	Example	Default
Input Pattern	JMESpath pattern to apply on input element for getting state, questions and history	{state: ticket, questions: q, history: h}	-

Parameters (JSON Schema)

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "type": "object",
  "properties": {
    "eventMeta": {
      "type": "object",
      "properties": {
        "parameters": {
          "type": "object",
          "properties": {
            "inputPattern": {
              "type": "string",
              "description": "JMESpath pattern to apply on input element for getting state, questions and history",
              "examples": ["{state: ticket, questions: q, history: h}"],
              "default": null
            }
          }
        }
      }
    }
  }
}

The input element (after the input pattern is applied) must be in {state, questions, history} format:

• state: the content to decide on. It can be a text value or an object. Object fields are passed to the model as named state values.

• questions (required): an object keyed by question name. Each question has a type, instructions and, depending on the type, criteria.

• history (optional): an array of {role, content} messages, where role is user or assistant.

Either state or history must be provided.

Question Type	Definition	Criteria
choice	Selects one option from a set (classification)	Object of {option: description} (description can be null), or an array of option names. At least 2 options
score	Rates on an ordered scale	Array of levels from lowest to highest
noul	Answers as yes, no or unknown	Not used

instructions is usually text. Structured instructions are also accepted and passed to the model as JSON text.

{
  "state": "I was charged twice for my order and want my money back today.",
  "questions": {
    "department": {"type": "choice", "instructions": "Which team should handle this?", "criteria": {"billing": "payments and invoices", "support": null, "sales": null}},
    "urgency":    {"type": "score",  "instructions": "How urgent is this request?", "criteria": ["low", "medium", "high"]},
    "refund":     {"type": "noul",   "instructions": "Is the customer asking for a refund?"}
  }
}

Answers are written to answers in the output element, keyed by question name. Model details are written to usage as {model, input_tokens}.

{
  "answers": {
    "department": {"type": "choice", "choice": "billing", "probabilities": {"billing": 0.91, "support": 0.07, "sales": 0.02}, "confidence": 0.91},
    "urgency":    {"type": "score", "score": 0.86, "level": "high", "probabilities": {"low": 0.03, "medium": 0.18, "high": 0.79}},
    "refund":     {"type": "noul", "noul": true}
  },
  "usage": {"model": "deberta-v3-base-mnli-fever-anli", "input_tokens": 142}
}

probabilities and confidence are returned only when the model provides them. For score questions, level is the most likely level.