Introduction to Zai

Zai is a utility library for structured LLM operations. Instead of writing raw prompts, you call typed methods like extract(), check(), summarize(), and text() that handle prompting, parsing, and validation for you.

In the ADK, Zai is available via adk.zai. It automatically uses your agent’s configured zai model from defaultModels in agent.config.ts.

import { adk } from '@botpress/runtime'

const zai = adk.zai

What you can do

Method	What it does	Page
extract()	Pull structured data from unstructured text	Extract structured data
check()	Verify a Boolean condition	Classify, validate and filter
label()	Categorize content with multiple labels	Classify, validate and filter
filter()	Filter an array by a condition	Classify, validate and filter
sort()	Sort items using natural language criteria	Classify, validate and filter
rate()	Rate items on a 1-5 scale	Classify, validate and filter
group()	Group items into categories	Classify, validate and filter
text()	Generate text from a prompt	Generate text and summaries
rewrite()	Transform text based on instructions	Generate text and summaries
summarize()	Summarize long content	Generate text and summaries
answer()	Answer questions with citations	Generate text and summaries
patch()	Make surgical edits to files	Generate text and summaries

Quick example

import { adk, z } from '@botpress/runtime'

const zai = adk.zai

const product = await zai.extract(
  'Blueberries are $3.99 and are in stock.',
  z.object({
    name: z.string(),
    price: z.number(),
    inStock: z.boolean(),
  })
)
// { name: "blueberries", price: 3.99, inStock: true }

Model configuration

Zai uses the zai model from defaultModels in agent.config.ts:

defaultModels: {
  autonomous: "cerebras:gpt-oss-120b",
  zai: "cerebras:gpt-oss-120b",
},

You can also change the model from the dev console under Settings > LLM Config.

Method categories