Our library introduces three key enhancements:
  • Response Model: Specify a data model to be returned by LLM to simplify your code.
  • Validation: Automatically validate response generated by LLM before you start using it.
  • Max Retries: Automated retry attempts for invalid responses.

​
A Glimpse into Instructor’s Capabilities

With Instructor, your code becomes more efficient and readable. Here’s a quick peek.

​
Understanding the workflow

Let’s see how we can leverage it to make use of instructor

​
Step 1: Define the data model

Create a data model to define the structure of the data you want to extract. This model will map directly to the information in the prompt. 
<?php
class UserDetail {
    public string $name;
    public int $age;
}

​
Step 2: Extract

Use the StructuredOutput::create() method to send a prompt and extract the data into the target object. The responseModel parameter specifies the model to use for extraction. 
/** @var UserDetail */
$user = (new StructuredOutput)->with(
    messages: [["role": "user", "content": "Extract Jason is 25 years old"]],
    responseModel: UserDetail::class,
    model: "gpt-3.5-turbo",
)->get();

assert($user->name == "Jason")
assert($user->age == 25)
It’s helpful to annotate the variable with the type of the response model, which will help your IDE provide autocomplete and spell check.

​
Understanding Validation

Validation can also be plugged into the same data model. If the response triggers any validation rules Instructor will raise a validation error.

​
Self Correcting on Validation Error

Here, the LeadReport model is passed as the $responseModel, and $maxRetries is set to 2. It means that if the extracted data does not match the model, Instructor will re-ask the model 2 times before giving up. 
use Cognesy\Instructor\StructuredOutput;
use Symfony\Component\Validator\Constraints as Assert;

class UserDetails
{
    public string $name;
    #[Assert\Email]
    public string $email;
}

$user = (new StructuredOutput)->with(
    messages: [['role' => 'user', 'content' => "you can reply to me via jason@gmailcom -- Jason"]],
    responseModel: UserDetails::class,
    maxRetries: 2
)->get();

assert($user->email === "jason@gmail.com");
!!! note “More about Validation” Check out Jason’s blog post Good LLM validation is just good validation

​
Custom Validators

Instructor uses Symfony validation component to validate extracted data. You can use #[Assert/Callback] annotation to build fully customized validation logic. See Symfony docs for more details on how to use Callback constraint. 
use Cognesy\Instructor\StructuredOutput;
use Symfony\Component\Validator\Constraints as Assert;
use Symfony\Component\Validator\Context\ExecutionContextInterface;
    
class UserDetails
{
    public string $name;
    public int $age;
    
    #[Assert\Callback]
    public function validateName(ExecutionContextInterface $context, mixed $payload) {
        if ($this->name !== strtoupper($this->name)) {
            $context->buildViolation("Name must be in uppercase.")
                ->atPath('name')
                ->setInvalidValue($this->name)
                ->addViolation();
        }
    }
}
    
$user = (new StructuredOutput)->with(
    messages: [['role' => 'user', 'content' => 'jason is 25 years old']],
    responseModel: UserDetails::class,
    maxRetries: 2
)->get();

assert($user->name === "JASON");