Embeddings are a key component of many LLM-based solutions and are used to represent text (or multimodal data) with numbers capturing their meaning and relationships.

​
Embeddings class

Embeddings class offers access to embeddings APIs and convenient methods to find top K vectors or documents most similar to provided query.

​
Supported providers

Embeddings class supports following embeddings providers:

  • Azure
  • Cohere
  • Gemini
  • Jina
  • Mistral
  • OpenAI

Embeddings providers access details can be found and modified via /config/embed.php.

​
Usage example

In this example, we use OpenAI embeddings provider to generate embeddings for a given list of documents (only one in this case).

The result is a list of embeddings vectors (one per document).

<?php
use Cognesy\Polyglot\Embeddings\Embeddings;

$docs = ['Computer vision models are used to analyze images and videos.'];

$embedding = (new Embeddings)
    ->withConnection('openai')
    ->create(input: $docs)
    ->all();
?>
InferenceHttp