Skip to main content
Embeddings map text (or multimodal data) into vectors where semantic similarity becomes numerical distance. Polyglot exposes embeddings through a thin Embeddings facade and a runtime-first infrastructure layer.

​
Embeddings facade

Embeddings is request-focused:
  • Request composition (withInputs, withModel, withOptions, with, withRequest)
  • Execution shortcuts (get, vectors, first)
  • Runtime handoff (runtime, withRuntime)
Provider/config/http/event assembly lives in EmbeddingsRuntime.

​
Supported providers

  • Azure OpenAI
  • Cohere
  • Gemini
  • Jina
  • OpenAI
Provider settings come from config presets or DSN input.

​
Basic usage

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

$result = Embeddings::using('openai')
    ->withModel('text-embedding-3-small')
    ->withInputs(['The quick brown fox'])
    ->get();

$vector = $result->first()?->values() ?? [];
echo count($vector);
// @doctest id="110b"

​
Runtime selection

Use constructor sugar for common paths: 
<?php
use Cognesy\Polyglot\Embeddings\Embeddings;

$embeddings = Embeddings::using('openai');
$embeddings = Embeddings::fromDsn('driver=openai,model=text-embedding-3-small');
// @doctest id="158e"
Inject a fully assembled runtime for advanced setup: 
<?php
use Cognesy\Http\Creation\HttpClientBuilder;
use Cognesy\Http\Drivers\Mock\MockHttpDriver;
use Cognesy\Polyglot\Embeddings\Embeddings;
use Cognesy\Polyglot\Embeddings\EmbeddingsRuntime;

$http = (new HttpClientBuilder())
    ->withDriver(new MockHttpDriver())
    ->create();

$runtime = EmbeddingsRuntime::using(
    preset: 'openai',
    httpClient: $http,
);

$embeddings = Embeddings::fromRuntime($runtime);
// @doctest id="dba5"

​
Request methods

$embeddings
    ->withInputs(['text 1', 'text 2'])
    ->withModel('text-embedding-3-large')
    ->withOptions(['dimensions' => 1536]);

$embeddings->with(
    input: ['text 1', 'text 2'],
    options: ['dimensions' => 1536],
    model: 'text-embedding-3-large',
);
// @doctest id="3985"

​
Response shortcuts

$response = $embeddings->get();
$vectors = $embeddings->vectors();
$first = $embeddings->first();
// @doctest id="70fb"

​
Multiple providers

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

$openai = Embeddings::using('openai')
    ->withModel('text-embedding-3-large')
    ->withInputs(['Document 1', 'Document 2'])
    ->vectors();

$cohere = Embeddings::using('cohere')
    ->withModel('embed-english-v3.0')
    ->withInputs(['Document 1', 'Document 2'])
    ->vectors();
// @doctest id="505b"

​
Custom runtime config

For explicit config/provider wiring, assemble runtime via EmbeddingsRuntime::fromProvider(...).

​
Driver registration

use Cognesy\Polyglot\Embeddings\Embeddings;

Embeddings::registerDriver('custom-provider', CustomEmbeddingsDriver::class);
// @doctest id="b21d"