TypeSafe

Mode: json · Models: 1

Jev evaluates content against named yes/no, choice, or score questions and returns structured answers.

Models

id	Name	Input type
typesafe-evaluate	Jev	t2t

SDK

JSON models use ai.run(). They are not supported by generate(), generateText(), or the SDK's async lifecycle methods.

import { createClient, Models } from '@picsart/ai-sdk'

const ai = createClient({ apiKey: process.env.PICSART_API_KEY! })
const { result } = await ai.run(Models.TypesafeEvaluate, {
  state: 'The weather is lovely and everyone is having a great time.',
  questions: {
    tone: {
      type: 'choice',
      instructions: 'Pick the tone.',
      options: ['positive', 'neutral', 'negative'],
    },
  },
})
console.log(result)

run() returns { result, status, usage, id } without media extraction or automatic Drive saving.

Parameters

Parameters from the SDK 6.29.0 catalog.

typesafe-evaluate — Jev

Input type: t2t; output mode: json.

Param	Type	Values
state	text	required — content to evaluate
questions	object	required — caller-named questions, each with type and instructions

Question type is noul (yes/no), choice, or score. Type-specific fields, such as choice options, are passed through to the evaluator.

Availability

This reference describes the SDK catalog. Do not assume that gen-ai generate, picsart_generate, or the Playground supports structured evaluation merely because the model is listed here. Inspect the interface's model discovery and supported execution methods before use.