vLLM

¶Embeddings

The embeddings method allows you to generate vector representations of text inputs. These embeddings are numerical representations capturing the semantic meaning of the input text and are commonly used for tasks like semantic search, clustering, or as input features for machine learning models.

Note

This example demonstrates how to use the embeddings endpoint of a vLLM server with an embedding model such as BAAI/bge-m3.

¶Example

This example generates embeddings for the input sentence "What is the best French cheese?" using the BAAI/bge-m3 embedding model.

use Partitech\PhpMistral\Clients\Vllm\VllmClient;
use Partitech\PhpMistral\MistralClientException;

$apiKey = getenv('VLLM_API_KEY');
$urlApiEmbedding = getenv('VLLM_API_EMBEDDING_URL');
$embeddingModel = getenv('VLLM_API_EMBEDDING_MODEL');

$client = new VllmClient(apiKey: (string) $apiKey, url: $urlApiEmbedding);

$inputs = [];

for ($i = 0; $i < 1; $i++) {
    $inputs[] = "What is the best French cheese?";
}

try {
    $embeddingsBatchResponse = $client->embeddings($inputs, $embeddingModel);
    print_r($embeddingsBatchResponse);
} catch (MistralClientException $e) {
    echo $e->getMessage();
    exit(1);
}

¶Example output

Array
(
    [id] => embd-270102c01ce54eb0bfef4103bad34106
    [object] => list
    [created] => 1742312694
    [model] => BAAI/bge-m3
    [data] => Array
        (
            [0] => Array
                (
                    [index] => 0
                    [object] => embedding
                    [embedding] => Array
                        (
                            [0] => -0.018157958984375
                            [1] => 0.011955261230469
                            [2] => -0.012351989746094
                            ...
                            [1023] => 0.0035037994384766
                        )
                )
        )
    [usage] => Array
        (
            [prompt_tokens] => 9
            [completion_tokens] => 0
            [total_tokens] => 9
        )
)

Tip

The embedding vector is a fixed-length array of floating-point numbers that represents the semantic content of the input text. You can compute similarity between two texts by comparing their vectors, typically using cosine similarity.

Important

Ensure the embedding model (e.g., BAAI/bge-m3) is available and properly configured on your vLLM server. The embeddings endpoint requires a specific model trained for embedding tasks.

Note

The embeddings() method always processes inputs in batch mode: even a single input will be returned as part of an array.