The Inference class is the primary interface for making requests to LLM providers in Polyglot. It provides methods for creating and sending requests, managing connections, and processing responses.

​
Creating an Inference Instance

<?php
use Cognesy\Polyglot\LLM\Inference;

// Create a basic inference instance with default settings
$inference = new Inference();

// Create an inference instance with a specific connection
$inference = new Inference()->withConnection('openai');

// Use the static method for simple, one-off requests
$response = Inference::text('What is the capital of France?');

​
Key Methods

The Inference class provides several key methods:

  • create(): Sends a request to the LLM provider
  • withConnection(): Specifies which connection to use
  • withConfig(): Sets a custom configuration
  • withHttpClient(): Specifies a custom HTTP client
  • withDebug(): Enables or disables debug mode
  • withCachedContext(): Sets cached context for improved performance