Text

Text is the simplest and most portable output format. When you do not set responseFormat, Polyglot asks the provider for a normal text response. Every provider supports this shape, making it the safest default for any use case.

Basic Usage

A minimal text request requires nothing more than a message:

<?php

use Cognesy\Messages\Messages;
use Cognesy\Polyglot\Inference\Inference;

$text = Inference::using('openai')
    ->withMessages(Messages::fromString('What is the single responsibility principle?'))
    ->get();

The get() method returns the raw string content from the model's response. There is no JSON parsing, no schema validation -- just the text the model produced.

When to Use Text Mode

Plain text is ideal for:

• Simple question answering
• Creative content generation (stories, poems, copy)
• Conversational interactions and chat
• Summaries and paraphrasing
• Any use case where structured data is not required

Working Across Providers

Text mode works consistently across all providers, making it the most portable option. You can swap providers without changing anything else about your request:

<?php

use Cognesy\Messages\Messages;
use Cognesy\Polyglot\Inference\Inference;

// Using OpenAI
$response = Inference::using('openai')
    ->withMessages(Messages::fromString('Write a short poem about the ocean.'))
    ->get();

// Using Anthropic -- same API, same result shape
$response = Inference::using('anthropic')
    ->withMessages(Messages::fromString('Write a short poem about the ocean.'))
    ->get();

Using the with() Method

You can also use the with() method to pass messages alongside other parameters. Since text is the default, there is no need to specify a response format:

<?php

use Cognesy\Messages\Messages;
use Cognesy\Polyglot\Inference\Inference;

$response = Inference::using('openai')
    ->with(
        messages: Messages::fromString('Explain the SOLID principles in one paragraph.'),
        options: ['temperature' => 0.3],
    )
    ->get();

Streaming Text Responses

For long-form content, you may want to stream the response so your application can display output as it arrives:

<?php

use Cognesy\Messages\Messages;
use Cognesy\Polyglot\Inference\Inference;

$stream = Inference::using('openai')
    ->withMessages(Messages::fromString('Write a short essay about renewable energy.'))
    ->stream();

foreach ($stream->deltas() as $delta) {
    echo $delta->contentDelta;
}

Each delta contains a contentDelta with the next chunk of text from the model. Streaming works with all providers that support it.

Accessing the Full Response

If you need metadata beyond the raw text -- such as token usage or the finish reason -- use the response() method instead of get():

<?php

use Cognesy\Messages\Messages;
use Cognesy\Polyglot\Inference\Inference;

$response = Inference::using('openai')
    ->withMessages(Messages::fromString('What is photosynthesis?'))
    ->response();

$text = $response->content();
$usage = $response->usage();
$reason = $response->finishReason();