La Plateforme

¶Embeddings

This section demonstrates how to generate embeddings for an array of sentences using the PhpMistral library with the Mistral Platform.

Embeddings are numerical representations of text, often used for tasks like semantic search, clustering, or similarity comparison.

Tip

By default, the mistral-embed model is used for generating embeddings.
If you need to use a different model, simply pass its name as the second argument to the embeddings() method:

$client->embeddings(
    datas: ["What is the best French cheese?"], 
    model: 'your-custom-model'
);

¶Example

use Partitech\PhpMistral\MistralClient;
use Partitech\PhpMistral\MistralClientException;

$client = new MistralClient($apiKey);

try {
    $embeddingsBatchResponse = $client->embeddings(["What is the best French cheese?"]);
} catch (MistralClientException $e) {
    echo $e->getMessage();
    exit(1);
}

¶Response Structure

The response contains an array with the embeddings for each input sentence under the data key:

Array
(
    [id] => 5b427f9a6c6b45739eca178cec9b78a1
    [object] => list
    [data] => Array
        (
            [0] => Array
                (
                    [object] => embedding
                    [embedding] => Array
                        (
                            [0] => -0.018600463867188
                            [1] => 0.027099609375
                            ...
                            [1023] => -0.001347541809082
                        )
                    [index] => 0
                )
        )
    [model] => mistral-embed
    [usage] => Array
        (
            [prompt_tokens] => 9
            [total_tokens] => 9
            [completion_tokens] => 0
        )
)

Note

Each embedding array represents a high-dimensional vector (e.g., 1024 dimensions) corresponding to the semantic representation of the input sentence.

¶Parameters

¶Use Cases

• Semantic Search: Compare embeddings to rank documents by similarity.
• Clustering: Group similar sentences based on their vector proximity.
• Recommendation Systems: Suggest content with similar embeddings.

Important

Embedding vectors can be compared using cosine similarity or other distance metrics for various NLP tasks.