cognesy/instructor-struct

Structured data extraction in PHP, powered by LLMs - core library

v2.0.0 2026-03-13 19:49 UTC

Requires

• php: ^8.3
• cognesy/instructor-config: ^2.0
• cognesy/instructor-events: ^2.0
• cognesy/instructor-messages: ^2.0
• cognesy/instructor-pipeline: ^2.0
• cognesy/instructor-polyglot: ^2.0
• cognesy/instructor-setup: ^2.0
• cognesy/instructor-templates: ^2.0
• cognesy/instructor-utils: ^2.0
• phpdocumentor/reflection-docblock: ^5.4
• phpstan/phpdoc-parser: ^1.29
• symfony/property-access: ^7.0
• symfony/property-info: ^7.0
• symfony/serializer: ^7.0
• symfony/validator: ^7.0

Requires (Dev)

• icanhazstring/composer-unused: ^0.9.0
• jetbrains/phpstorm-attributes: ^1.2
• maglnet/composer-require-checker: ^4.16
• mockery/mockery: ^1.6
• pestphp/pest: ^2.34
• phpstan/phpstan: ^1.11
• roave/security-advisories: dev-latest
• vimeo/psalm: ^6.0

Suggests

• cognesy/instructor-schema: For schema validation and management

MIT

• Dariusz Debowczyk

• automation
• inference
• data extraction
• ai
• data processing
• openai
• llm
• anthropic
• ollama
• language models
• genai
• cohere
• structured output
• semantic processing

README

Core structured-output engine for InstructorPHP.

Use it to turn unstructured LLM responses into typed PHP data, with validation, retries, and streaming updates.

Example

<?php

use Cognesy\Instructor\StructuredOutput;
use Cognesy\Polyglot\Inference\Config\LLMConfig;

class Person {
    public string $name;
    public int $age;
}

$person = StructuredOutput::fromConfig(LLMConfig::fromArray(['driver' => 'openai']))
    ->with(
        messages: 'His name is Jason and he is 28 years old.',
        responseModel: Person::class,
    )
    ->get();

Documentation

• packages/instructor/docs/quickstart.md
• packages/instructor/docs/essentials/usage.md
• packages/instructor/docs/_meta.yaml