Customizing AI models

By default, transformers use the AI provider and model configured in your config/url-ai-transformer.php file. However, you can customize which AI models to use on a per-transformer basis, or even skip AI entirely.

This package uses Prism under the hood for all AI interactions. Prism provides a unified interface for working with different AI providers like OpenAI, Anthropic, Google Gemini, and more. For detailed information about available providers, models, and configuration options, check out the Prism documentation.

Here's an example where we use various options like temperature and max tokens.

class PreciseTransformer extends Transformer
{
    public function transform(): void
    {
        $response = Prism::text()
            ->using(Config::aiProvider(), Config::aiModel())
            ->withPrompt($this->getPrompt())
            ->withMaxTokens(1000)
            ->withTemperature(0.1) // Low temperature for precise, consistent output
            ->asText();

        $this->transformationResult->result = $response->text;
    }
}

class CreativeTransformer extends Transformer
{
    public function transform(): void
    {
        $response = Prism::text()
            ->using(Config::aiProvider(), Config::aiModel())
            ->withPrompt($this->getPrompt())
            ->withMaxTokens(2000)
            ->withTemperature(0.8) // Higher temperature for creative output
            ->asText();

        $this->transformationResult->result = $response->text;
    }
}

##Transformers without AI

Not all transformers need to use AI. You can create transformers that process content using traditional methods:

class WordCountTransformer extends Transformer
{
    public function transform(): void
    {
        $wordCount = str_word_count(strip_tags($this->urlContent));
        
        $this->transformationResult->result =  $wordCount;
    }
}