Skip to main content

​
The Inference Class

The Inference class is the main entry point for LLM interactions. It encapsulates the complexities of different providers behind a unified interface. 
namespace Cognesy\Polyglot\Inference;

use Cognesy\Polyglot\Inference\Contracts\CanCreateInference;
use Cognesy\Polyglot\Inference\Data\InferenceRequest;

class Inference {
    public function __construct(?CanCreateInference $runtime = null) { ... }

    // Runtime selection
    public static function using(string $preset): self { ... }
    public static function fromDsn(string $dsn): self { ... }
    public static function fromRuntime(CanCreateInference $runtime): self { ... }
    public function withRuntime(CanCreateInference $runtime): self { ... }
    public function runtime(): CanCreateInference { ... }

    // Request building and execution
    public function with(...): self { ... }
    public function withRequest(InferenceRequest $request): self { ... }
    public function create(?InferenceRequest $request = null): PendingInference { ... }

    // Convenience response accessors
    public function get(): string { ... }
    public function response(): InferenceResponse { ... }
    public function stream(): InferenceStream { ... }
}
// @doctest id="8132"
The Inference class follows a fluent interface pattern for request building; infrastructure is assembled in InferenceRuntime.

​
The Embeddings Class

Similarly, the Embeddings class provides a unified interface for generating embeddings: 
namespace Cognesy\Polyglot\Embeddings;

use Cognesy\Polyglot\Embeddings\Contracts\CanCreateEmbeddings;
use Cognesy\Polyglot\Embeddings\Data\EmbeddingsRequest;

class Embeddings {
    public function __construct(?CanCreateEmbeddings $runtime = null) { ... }

    // Runtime selection
    public static function using(string $preset): self { ... }
    public static function fromDsn(string $dsn): self { ... }
    public static function fromRuntime(CanCreateEmbeddings $runtime): self { ... }
    public function withRuntime(CanCreateEmbeddings $runtime): self { ... }
    public function runtime(): CanCreateEmbeddings { ... }

    // Request building and execution
    public function with(...): self { ... }
    public function withRequest(EmbeddingsRequest $request): self { ... }
    public function create(?EmbeddingsRequest $request = null): PendingEmbeddings { ... }

    // Convenience response accessors
    public function get(): EmbeddingsResponse { ... }
    public function first(): ?Vector { ... }
    public function vectors(): array { ... }
}
// @doctest id="0fc9"
Similarity-search helper methods are provided by EmbedUtils, not by the Embeddings facade: 
use Cognesy\Polyglot\Embeddings\Utils\EmbedUtils;

$matches = EmbedUtils::findSimilar(
    embeddings: Embeddings::using('openai')->runtime(),
    query: $query,
    documents: $documents,
    topK: 5,
);
// @doctest id="2fb2"