Add stream support for Text To Speech

src/Contracts/Resources/AudioContract.php

@@ -2,6 +2,7 @@
 
 namespace OpenAI\Contracts\Resources;
 
+use OpenAI\Responses\Audio\SpeechStreamResponse;
 use OpenAI\Responses\Audio\TranscriptionResponse;
 use OpenAI\Responses\Audio\TranslationResponse;
 
@@ -16,6 +17,15 @@ interface AudioContract
      */
     public function speech(array $parameters): string;
 
+    /**
+     * Generates streamed audio from the input text.
+     *
+     * @see https://platform.openai.com/docs/api-reference/audio/createSpeech
+     *
+     * @param  array<string, mixed>  $parameters
+     */
+    public function speechStreamed(array $parameters): SpeechStreamResponse;
+
     /**
      * Transcribes audio into the input language.
      *

src/Contracts/ResponseStreamContract.php

@@ -0,0 +1,18 @@
+<?php
+
+declare(strict_types=1);
+
+namespace OpenAI\Contracts;
+
+use IteratorAggregate;
+
+/**
+ * @template T
+ *
+ * @extends IteratorAggregate<int, T>
+ *
+ * @internal
+ */
+interface ResponseStreamContract extends IteratorAggregate
+{
+}

src/Resources/Audio.php

@@ -5,6 +5,7 @@
 namespace OpenAI\Resources;
 
 use OpenAI\Contracts\Resources\AudioContract;
+use OpenAI\Responses\Audio\SpeechStreamResponse;
 use OpenAI\Responses\Audio\TranscriptionResponse;
 use OpenAI\Responses\Audio\TranslationResponse;
 use OpenAI\ValueObjects\Transporter\Payload;
@@ -28,6 +29,22 @@ public function speech(array $parameters): string
         return $this->transporter->requestContent($payload);
     }
 
+    /**
+     * Generates streamed audio from the input text.
+     *
+     * @see https://platform.openai.com/docs/api-reference/audio/createSpeech
+     *
+     * @param  array<string, mixed>  $parameters
+     */
+    public function speechStreamed(array $parameters): SpeechStreamResponse
+    {
+        $payload = Payload::create('audio/speech', $parameters);
+
+        $response = $this->transporter->requestStream($payload);
+
+        return new SpeechStreamResponse($response);
+    }
+
     /**
      * Transcribes audio into the input language.
      *

src/Responses/Audio/SpeechStreamResponse.php

@@ -0,0 +1,53 @@
+<?php
+
+namespace OpenAI\Responses\Audio;
+
+use Generator;
+use Http\Discovery\Psr17Factory;
+use OpenAI\Contracts\ResponseHasMetaInformationContract;
+use OpenAI\Contracts\ResponseStreamContract;
+use OpenAI\Responses\Meta\MetaInformation;
+use Psr\Http\Message\ResponseInterface;
+
+/**
+ * @implements ResponseStreamContract<string>
+ */
+final class SpeechStreamResponse implements ResponseHasMetaInformationContract, ResponseStreamContract
+{
+    public function __construct(
+        private readonly ResponseInterface $response,
+    ) {
+        //
+    }
+
+    /**
+     * {@inheritDoc}
+     */
+    public function getIterator(): Generator
+    {
+        while (! $this->response->getBody()->eof()) {
+            yield $this->response->getBody()->read(1024);
+        }
+    }
+
+    public function meta(): MetaInformation
+    {
+        // @phpstan-ignore-next-line
+        return MetaInformation::from($this->response->getHeaders());
+    }
+
+    public static function fake(string $content = null, MetaInformation $meta = null): static
+    {
+        $psr17Factory = new Psr17Factory();
+        $response = $psr17Factory->createResponse()
+            ->withBody($psr17Factory->createStream($content ?? (string) file_get_contents(__DIR__.'/../../Testing/Responses/Fixtures/Audio/speech-streamed.mp3')));
+
+        if ($meta instanceof \OpenAI\Responses\Meta\MetaInformation) {
+            foreach ($meta->toArray() as $key => $value) {
+                $response = $response->withHeader($key, (string) $value);
+            }
+        }
+
+        return new self($response);
+    }
+}

src/Responses/StreamResponse.php

@@ -3,8 +3,8 @@
 namespace OpenAI\Responses;
 
 use Generator;
-use IteratorAggregate;
 use OpenAI\Contracts\ResponseHasMetaInformationContract;
+use OpenAI\Contracts\ResponseStreamContract;
 use OpenAI\Exceptions\ErrorException;
 use OpenAI\Responses\Meta\MetaInformation;
 use Psr\Http\Message\ResponseInterface;
@@ -13,9 +13,9 @@
 /**
  * @template TResponse
  *
- * @implements IteratorAggregate<int, TResponse>
+ * @implements ResponseStreamContract<TResponse>
  */
-final class StreamResponse implements IteratorAggregate, ResponseHasMetaInformationContract
+final class StreamResponse implements ResponseHasMetaInformationContract, ResponseStreamContract
 {
     /**
      * Creates a new Stream Response instance.

src/Testing/ClientFake.php

@@ -4,6 +4,7 @@
 
 use OpenAI\Contracts\ClientContract;
 use OpenAI\Contracts\ResponseContract;
+use OpenAI\Contracts\ResponseStreamContract;
 use OpenAI\Responses\StreamResponse;
 use OpenAI\Testing\Requests\TestRequest;
 use OpenAI\Testing\Resources\AssistantsTestResource;
@@ -113,7 +114,7 @@ private function resourcesOf(string $type): array
         return array_filter($this->requests, fn (TestRequest $request): bool => $request->resource() === $type);
     }
 
-    public function record(TestRequest $request): ResponseContract|StreamResponse|string
+    public function record(TestRequest $request): ResponseContract|ResponseStreamContract|string
     {
         $this->requests[] = $request;
 

src/Testing/Resources/AudioTestResource.php

@@ -4,6 +4,7 @@
 
 use OpenAI\Contracts\Resources\AudioContract;
 use OpenAI\Resources\Audio;
+use OpenAI\Responses\Audio\SpeechStreamResponse;
 use OpenAI\Responses\Audio\TranscriptionResponse;
 use OpenAI\Responses\Audio\TranslationResponse;
 use OpenAI\Testing\Resources\Concerns\Testable;
@@ -22,6 +23,11 @@ public function speech(array $parameters): string
         return $this->record(__FUNCTION__, func_get_args());
     }
 
+    public function speechStreamed(array $parameters): SpeechStreamResponse
+    {
+        return $this->record(__FUNCTION__, func_get_args());
+    }
+
     public function transcribe(array $parameters): TranscriptionResponse
     {
         return $this->record(__FUNCTION__, func_get_args());

src/Testing/Resources/Concerns/Testable.php

@@ -3,7 +3,7 @@
 namespace OpenAI\Testing\Resources\Concerns;
 
 use OpenAI\Contracts\ResponseContract;
-use OpenAI\Responses\StreamResponse;
+use OpenAI\Contracts\ResponseStreamContract;
 use OpenAI\Testing\ClientFake;
 use OpenAI\Testing\Requests\TestRequest;
 
@@ -18,7 +18,7 @@ abstract protected function resource(): string;
     /**
      * @param  array<string, mixed>  $args
      */
-    protected function record(string $method, array $args = []): ResponseContract|StreamResponse|string
+    protected function record(string $method, array $args = []): ResponseContract|ResponseStreamContract|string
     {
         return $this->fake->record(new TestRequest($this->resource(), $method, $args));
     }

tests/Arch.php

@@ -29,6 +29,7 @@
 ]);
 
 test('responses')->expect('OpenAI\Responses')->toOnlyUse([
+    'Http\Discovery\Psr17Factory',
     'OpenAI\Enums',
     'OpenAI\Exceptions\ErrorException',
     'OpenAI\Contracts',

tests/Fixtures/Audio.php

@@ -158,3 +158,11 @@ function audioFileContent(): string
 {
     return file_get_contents(__DIR__.'/audio.mp3');
 }
+
+/**
+ * @return resource
+ */
+function speechStream()
+{
+    return fopen(__DIR__.'/Streams/Speech.mp3', 'r');
+}

tests/Resources/Audio.php

@@ -1,5 +1,8 @@
 <?php
 
+use GuzzleHttp\Psr7\Response;
+use GuzzleHttp\Psr7\Stream;
+use OpenAI\Responses\Audio\SpeechStreamResponse;
 use OpenAI\Responses\Audio\TranscriptionResponse;
 use OpenAI\Responses\Audio\TranscriptionResponseSegment;
 use OpenAI\Responses\Audio\TranslationResponse;
@@ -209,3 +212,29 @@
         ->toBeString()
         ->toBe(audioFileContent());
 });
+
+test('text to speech streamed', function () {
+    $response = new Response(
+        body: new Stream(speechStream()),
+        headers: metaHeaders(),
+    );
+
+    $client = mockStreamClient('POST', 'audio/speech', [
+        'model' => 'tts-1',
+        'input' => 'Hello, how are you?',
+        'voice' => 'alloy',
+    ], $response, 'requestContent');
+
+    $result = $client->audio()->speechStreamed([
+        'model' => 'tts-1',
+        'input' => 'Hello, how are you?',
+        'voice' => 'alloy',
+    ]);
+
+    expect($result)
+        ->toBeInstanceOf(SpeechStreamResponse::class)
+        ->toBeInstanceOf(IteratorAggregate::class);
+
+    expect($result->getIterator())
+        ->toBeInstanceOf(Iterator::class);
+});

tests/Responses/Audio/SpeechStreamResponse.php

@@ -0,0 +1,36 @@
+<?php
+
+use GuzzleHttp\Psr7\Response;
+use GuzzleHttp\Psr7\Stream;
+use OpenAI\Responses\Audio\SpeechStreamResponse;
+use OpenAI\Responses\Meta\MetaInformation;
+
+test('from response', function () {
+    $response = new Response(
+        body: new Stream(speechStream()),
+        headers: metaHeaders(),
+    );
+
+    $speech = new SpeechStreamResponse($response);
+
+    expect($speech)
+        ->toBeInstanceOf(SpeechStreamResponse::class)
+        ->getIterator()->toBeInstanceOf(Iterator::class)
+        ->meta()->toBeInstanceOf(MetaInformation::class);
+});
+
+test('fake', function () {
+    $response = SpeechStreamResponse::fake();
+
+    foreach ($response as $chunk) {
+        expect($chunk)->toBeString();
+    }
+});
+
+test('fake with override', function () {
+    $response = SpeechStreamResponse::fake('fake audio');
+
+    foreach ($response as $chunk) {
+        expect($chunk)->toBe('fake audio');
+    }
+});

tests/Testing/Resources/AudioTestResource.php

@@ -1,6 +1,7 @@
 <?php
 
 use OpenAI\Resources\Audio;
+use OpenAI\Responses\Audio\SpeechStreamResponse;
 use OpenAI\Responses\Audio\TranscriptionResponse;
 use OpenAI\Responses\Audio\TranslationResponse;
 use OpenAI\Testing\ClientFake;
@@ -26,6 +27,27 @@
     });
 });
 
+it('records a streamed speech request', function () {
+    $fake = new ClientFake([
+        SpeechStreamResponse::fake(),
+    ]);
+
+    $fake->audio()->speechStreamed([
+        'model' => 'tts-1',
+        'input' => 'Hello, how are you?',
+        'voice' => 'alloy',
+    ]);
+
+    $fake->assertSent(Audio::class, function ($method, $parameters) {
+        return $method === 'speechStreamed' &&
+            $parameters === [
+                'model' => 'tts-1',
+                'input' => 'Hello, how are you?',
+                'voice' => 'alloy',
+            ];
+    });
+});
+
 it('records an audio transcription request', function () {
     $fake = new ClientFake([
         TranscriptionResponse::fake(),