Skip to content

Main Navigation Home Quick Start Changelog API

Appearance

Sidebar Navigation

How agents work

What ADK is

SoDK (a mental model)

Quickstart

Playground

Developer Tools

Overview

ADK Assembly MCP

ESLint Rules

The Real-Model Matrix

The Loop

Overview

Turn Runner

Overview

Inside one turn

Gates and non-goals

LLM Dispatch

Overview

Executor seam

ADK-side facts

Signalling and bounds

Errors during dispatch

Primitives

Overview

Tokenizable

Identity

Message

Media

Memory

Retrievable

Thought

ToolCall

Tools

Overview

What a Tool is

Registry and lifecycle

bindContext and describe()

Trust and safety

Advanced details

Artifacts

Overview

SpooledArtifact shape

forgeTools(ctx) in depth

Extending

Events

Pipelines

Overview

Composition

Throws

Abort

What each pipeline owns

Turn-scoped

Dispatch-scoped

Scoped Stashes

Gates

Overview

Canonical applications

Lifecycle

Durability and integration

Trust Tiers

Overview

Envelopes

Threat Analysis

Persistence

Threat Analysis

Identity and Reasoning

Threat Analysis

Media

Threat Analysis

Failure

Overview

Exception Reference

Budgets

Token Thrift

Behavioral Rails

Read the Wire

Featured Batteries

LLM Batteries

Overview

The Shared Contract

OpenAI Chat Completions

OpenAI Responses

Anthropic Messages

Gemini generateContent

AWS Bedrock Converse

Ollama

WebLLM

LiteRT-LM

Transformers.js

Claude Code CLI

Vector Storage

Overview

The Query Builder & Filters

Schema & Migrations

Consistency & Capabilities

Encoders: vectors or text

Retrievable Glue

The Adapter Matrix

Writing an Adapter

Media Pipeline

Overview

Plans & Self-Repair

The Pipe DSL

Text, Data & Patches

The Capability Model

Engine Dispatch

The Bundled Fleet

Generating Media

Agent Tools

BYO Runtime & Custom Engines

Skills

Overview

Context channels

Sources

Executable capability

Tool results

Middleware and placement

Authoring SKILL.md

Orchestration

Overview

The Lifecycle

The Plan IR

Predicates & Cells

Execution State

The Prose

Approval

Agent Tools

The Store Contract

Dev Tools

Overview

The Workspace

The Steps

Engines & Capabilities

In-Place Fixers

Agent Tools

Context

Overview

Token Thrift

Compact

Specialist batteries

Overview

Speech-to-Text (STT)

OCR

Image Captioning

Sandbox

Overview

run_shell_command

Workspace tools

Media executor

JavaScript

Isolation

Overview

Browser

Node

Recipes

Generation

Overview

OpenAI

Gemini

Transformers.js

Local Diffusion

Recipes

Validation

Overview

API Surface Scope

Operating Modes

Atomic Behaviors

Rule Types

Family Recipes

Gemini Sentinels

Advisories & Placeholders

Writing a Profile

SearXNG Search

Web Retrieval Glue

Scrapper

Assembly

Overview

Minimal Agent Assembly

Bring your own LLM

Bring your own storage

Bring your own tools

Bring your own retrieval

Bring your own memory

Runtime Loading

Wiring the Pipelines

Listening to the Assembly

Tools batteries

Artifact batteries

LLM batteries

Storage batteries

Embeddings batteries

Specialist batteries

Isolation batteries

Sandbox battery

Generation batteries

TTS batteries

Orchestration battery

Showcase

Overview

The Ask ADK Agent

Building the On-Device Batteries

Building the Dev Tools Battery

Punching Above Its Weights

The Agent, In Full

Glossary

Full API

Changelog

On this page

⏱1 min read · 22 words

Are you an LLM? You can read better optimized documentation at /api/batteries/orchestration/cells/structured.md for this page in Markdown format

batteries/orchestration/cells/structured

Functions

Function	Description
createStructuredCell	The `structured` predicate evaluator.

Pager

Next pageHow agents work

Copyright © 2025-present New Horizon Technology LTD