Summarise approval

Condense an approval request — purchase, hiring, access, policy exception — into the decision being asked, the rationale provided, and the risk signals an approver should weigh.

Use this when the input is explicitly an approval request, not a generic message. The output shape is tuned for "should I approve this?" not "what does this say?"

POSThttps://aiengine.velgent.com/api/v1/summarise/approval

Headers

Same as Summarise text.

Request body

Field	Type	Required	Default	Description
title	string	Yes	—	Short title of the approval request. 1–500 characters.
description	string	Yes	—	Full request description, including any context the requester provided. 1–32,000 characters.
category	"purchase" \| "hiring" \| "access" \| "policy_exception" \| "other"	No	"other"	Type of approval. Helps the model interpret the request.
amount	object	No	—	For purchases: { value: number, currency: ISO 4217 code }. Echoed in the response.
metadata	object	No	—	Free-form key/value pairs echoed on the response.

Response

Field	Type	Required	Default	Description
id	string	Yes	—	Unique identifier.
decision_ask	string	Yes	—	One sentence stating exactly what is being asked of the approver.
rationale	string	Yes	—	A short summary of the reasoning the requester provided.
risk_signals	array of string	Yes	—	Specific risks an approver should consider. May be empty.
intent	object	Yes	—	Same shape as Summarise text.
usage	object	Yes	—	Token counts.
created_at	string (ISO 8601)	Yes	—	UTC timestamp.

Examples

Example response

{
  "id":           "suma_01HZQ7K8YV9X3R2M5N6P7QABCD",
  "decision_ask": "Approve $14,400/year for 8 additional Datadog Pro seats for the platform team.",
  "rationale":    "Current seats are exhausted, blocking on-call engineers from dashboards. VP Eng has already approved.",
  "risk_signals": [
    "Recurring annual cost; renews automatically unless cancelled.",
    "No alternative tooling evaluation cited."
  ],
  "intent": {
    "action_needed":  true,
    "due_date":       null,
    "urgency_score":  68,
    "sensitivity":    "medium"
  },
  "usage":      { "input_tokens": 71, "output_tokens": 92, "total_tokens": 163 },
  "created_at": "2026-05-10T17:42:11Z"
}

Errors

See Errors.