[Platform] Add support for Cartesia

docs/components/platform.rst

@@ -102,6 +102,9 @@ Supported Models & Platforms
     * `LM Studio Catalog`_ and `HuggingFace`_ Models  with `LM Studio`_ as Platform.
     * All models provided by `HuggingFace`_ can be listed with a command in the examples folder,
       and also filtered, e.g. ``php examples/huggingface/_model-listing.php --provider=hf-inference --task=object-detection``
+* **Voice Models**
+    * `Cartesia TTS` with `Cartesia`_ as Platform
+    * `Cartesia STT` with `Cartesia`_ as Platform
 
 Options
 -------
@@ -463,6 +466,7 @@ Code Examples
 .. _`Anthropic's Claude`: https://www.anthropic.com/claude
 .. _`Anthropic`: https://www.anthropic.com/
 .. _`AWS Bedrock`: https://aws.amazon.com/bedrock/
+.. _`Cartesia`: https://cartesia.ai/sonic
 .. _`Meta's Llama`: https://www.llama.com/
 .. _`Ollama`: https://ollama.com/
 .. _`Replicate`: https://replicate.com/

examples/.env

@@ -52,6 +52,10 @@ OPENROUTER_KEY=
 ELEVEN_LABS_URL=https://api.elevenlabs.io/v1
 ELEVEN_LABS_API_KEY=
 
+# For using Cartesia
+CARTESIA_API_KEY=
+CARTESIA_API_VERSION=2025-04-16
+
 # For using SerpApi (tool)
 SERP_API_KEY=
 

examples/cartesia/README.md

@@ -0,0 +1,9 @@
+# Cartesia Examples
+
+One use case of Cartesia is to convert text to speech, which creates audio files from text input.
+
+To run the examples, you can use additional tools like (mpg123)[https://www.mpg123.de/]:
+
+```bash
+php cartesia/text-to-speech.php | mpg123 -
+```

examples/cartesia/speech-to-text.php

@@ -0,0 +1,25 @@
+<?php
+
+/*
+ * This file is part of the Symfony package.
+ *
+ * (c) Fabien Potencier <fabien@symfony.com>
+ *
+ * For the full copyright and license information, please view the LICENSE
+ * file that was distributed with this source code.
+ */
+
+use Symfony\AI\Platform\Bridge\Cartesia\PlatformFactory;
+use Symfony\AI\Platform\Message\Content\Audio;
+
+require_once dirname(__DIR__).'/bootstrap.php';
+
+$platform = PlatformFactory::create(
+    apiKey: env('CARTESIA_API_KEY'),
+    version: env('CARTESIA_API_VERSION'),
+    httpClient: http_client(),
+);
+
+$result = $platform->invoke('ink-whisper', Audio::fromFile(dirname(__DIR__, 2).'/fixtures/audio.mp3'));
+
+echo $result->asText().\PHP_EOL;

examples/cartesia/text-to-speech.php

@@ -0,0 +1,32 @@
+<?php
+
+/*
+ * This file is part of the Symfony package.
+ *
+ * (c) Fabien Potencier <fabien@symfony.com>
+ *
+ * For the full copyright and license information, please view the LICENSE
+ * file that was distributed with this source code.
+ */
+
+use Symfony\AI\Platform\Bridge\Cartesia\PlatformFactory;
+use Symfony\AI\Platform\Message\Content\Text;
+
+require_once dirname(__DIR__).'/bootstrap.php';
+
+$platform = PlatformFactory::create(
+    apiKey: env('CARTESIA_API_KEY'),
+    version: env('CARTESIA_API_VERSION'),
+    httpClient: http_client(),
+);
+
+$result = $platform->invoke('sonic-3', new Text('Hello world'), [
+    'voice' => '6ccbfb76-1fc6-48f7-b71d-91ac6298247b', // Tessa (https://play.cartesia.ai/voices/6ccbfb76-1fc6-48f7-b71d-91ac6298247b)
+    'output_format' => [
+        'container' => 'mp3',
+        'sample_rate' => 48000,
+        'bit_rate' => 192000,
+    ],
+]);
+
+echo $result->asBinary().\PHP_EOL;

src/ai-bundle/config/options.php

@@ -54,6 +54,16 @@
                             ->end()
                         ->end()
                     ->end()
+                    ->arrayNode('cartesia')
+                        ->children()
+                            ->stringNode('api_key')->isRequired()->end()
+                            ->stringNode('version')->isRequired()->end()
+                            ->stringNode('http_client')
+                            ->defaultValue('http_client')
+                                ->info('Service ID of the HTTP client to use')
+                            ->end()
+                        ->end()
+                    ->end()
                     ->arrayNode('eleven_labs')
                         ->children()
                             ->stringNode('host')->end()

src/ai-bundle/config/services.php

@@ -28,6 +28,7 @@
 use Symfony\AI\Platform\Bridge\Anthropic\Contract\AnthropicContract;
 use Symfony\AI\Platform\Bridge\Anthropic\ModelCatalog as AnthropicModelCatalog;
 use Symfony\AI\Platform\Bridge\Anthropic\TokenOutputProcessor as AnthropicTokenOutputProcessor;
+use Symfony\AI\Platform\Bridge\Cartesia\ModelCatalog as CartesiaModelCatalog;
 use Symfony\AI\Platform\Bridge\Cerebras\ModelCatalog as CerebrasModelCatalog;
 use Symfony\AI\Platform\Bridge\DeepSeek\ModelCatalog as DeepSeekModelCatalog;
 use Symfony\AI\Platform\Bridge\DockerModelRunner\ModelCatalog as DockerModelRunnerModelCatalog;
@@ -85,6 +86,7 @@
         // model catalog
         ->set('ai.platform.model_catalog.aimlapi', AiMlApiModelCatalog::class)
         ->set('ai.platform.model_catalog.anthropic', AnthropicModelCatalog::class)
+        ->set('ai.platform.model_catalog.cartesia', CartesiaModelCatalog::class)
         ->set('ai.platform.model_catalog.cerebras', CerebrasModelCatalog::class)
         ->set('ai.platform.model_catalog.deepseek', DeepSeekModelCatalog::class)
         ->set('ai.platform.model_catalog.dockermodelrunner', DockerModelRunnerModelCatalog::class)

src/ai-bundle/src/AiBundle.php

@@ -41,6 +41,7 @@
 use Symfony\AI\Chat\MessageStoreInterface;
 use Symfony\AI\Platform\Bridge\Anthropic\PlatformFactory as AnthropicPlatformFactory;
 use Symfony\AI\Platform\Bridge\Azure\OpenAi\PlatformFactory as AzureOpenAiPlatformFactory;
+use Symfony\AI\Platform\Bridge\Cartesia\PlatformFactory as CartesiaPlatformFactory;
 use Symfony\AI\Platform\Bridge\Cerebras\PlatformFactory as CerebrasPlatformFactory;
 use Symfony\AI\Platform\Bridge\DeepSeek\PlatformFactory as DeepSeekPlatformFactory;
 use Symfony\AI\Platform\Bridge\DockerModelRunner\PlatformFactory as DockerModelRunnerPlatformFactory;
@@ -293,6 +294,26 @@ private function processPlatformConfig(string $type, array $platform, ContainerB
             return;
         }
 
+        if ('cartesia' === $type) {
+            $definition = (new Definition(Platform::class))
+                ->setFactory(CartesiaPlatformFactory::class.'::create')
+                ->setLazy(true)
+                ->addTag('proxy', ['interface' => PlatformInterface::class])
+                ->setArguments([
+                    $platform['api_key'],
+                    $platform['version'],
+                    new Reference($platform['http_client'], ContainerInterface::NULL_ON_INVALID_REFERENCE),
+                    new Reference('ai.platform.model_catalog.cartesia'),
+                    null,
+                    new Reference('event_dispatcher'),
+                ])
+                ->addTag('ai.platform', ['name' => 'cartesia']);
+
+            $container->setDefinition('ai.platform.cartesia', $definition);
+
+            return;
+        }
+
         if ('eleven_labs' === $type) {
             $platformId = 'ai.platform.eleven_labs';
             $definition = (new Definition(Platform::class))

src/ai-bundle/tests/DependencyInjection/AiBundleTest.php

@@ -2795,6 +2795,11 @@ private function getFullConfig(): array
                             'api_version' => '2024-02-15-preview',
                         ],
                     ],
+                    'cartesia' => [
+                        'api_key' => 'cartesia_key_full',
+                        'version' => '2025-04-16',
+                        'http_client' => 'http_client',
+                    ],
                     'eleven_labs' => [
                         'host' => 'https://api.elevenlabs.io/v1',
                         'api_key' => 'eleven_labs_key_full',

src/platform/CHANGELOG.md

@@ -27,6 +27,7 @@ CHANGELOG
    - AI/ML API (language models and embeddings)
    - Docker Model Runner (local model hosting)
    - Scaleway (language models like OpenAI OSS, Llama 4, Qwen 3, and more)
+   - Cartesia (voice model that supports both text-to-speech and speech-to-text)
  * Add comprehensive message system with role-based messaging:
    - `UserMessage` for user inputs with multi-modal content
    - `SystemMessage` for system instructions

src/platform/src/Bridge/Cartesia/Cartesia.php

@@ -0,0 +1,21 @@
+<?php
+
+/*
+ * This file is part of the Symfony package.
+ *
+ * (c) Fabien Potencier <fabien@symfony.com>
+ *
+ * For the full copyright and license information, please view the LICENSE
+ * file that was distributed with this source code.
+ */
+
+namespace Symfony\AI\Platform\Bridge\Cartesia;
+
+use Symfony\AI\Platform\Model;
+
+/**
+ * @author Guillaume Loulier <personal@guillaumeloulier.fr>
+ */
+final class Cartesia extends Model
+{
+}

src/platform/src/Bridge/Cartesia/CartesiaClient.php

@@ -0,0 +1,91 @@
+<?php
+
+/*
+ * This file is part of the Symfony package.
+ *
+ * (c) Fabien Potencier <fabien@symfony.com>
+ *
+ * For the full copyright and license information, please view the LICENSE
+ * file that was distributed with this source code.
+ */
+
+namespace Symfony\AI\Platform\Bridge\Cartesia;
+
+use Symfony\AI\Platform\Capability;
+use Symfony\AI\Platform\Exception\RuntimeException;
+use Symfony\AI\Platform\Model;
+use Symfony\AI\Platform\ModelClientInterface;
+use Symfony\AI\Platform\Result\RawHttpResult;
+use Symfony\AI\Platform\Result\RawResultInterface;
+use Symfony\Contracts\HttpClient\HttpClientInterface;
+
+/**
+ * @author Guillaume Loulier <personal@guillaumeloulier.fr>
+ */
+final class CartesiaClient implements ModelClientInterface
+{
+    public function __construct(
+        private readonly HttpClientInterface $httpClient,
+        #[\SensitiveParameter] private readonly string $apiKey,
+        private readonly string $version,
+    ) {
+    }
+
+    public function supports(Model $model): bool
+    {
+        return $model instanceof Cartesia;
+    }
+
+    public function request(Model $model, array|string $payload, array $options = []): RawResultInterface
+    {
+        return match (true) {
+            \in_array(Capability::TEXT_TO_SPEECH, $model->getCapabilities()) => $this->doTextToSpeech($model, $payload, $options),
+            \in_array(Capability::SPEECH_TO_TEXT, $model->getCapabilities()) => $this->doSpeechToText($model, $payload, $options),
+            default => throw new RuntimeException(\sprintf('The model "%s" is not supported.', $model->getName())),
+        };
+    }
+
+    /**
+     * @param array<string|int, mixed> $payload
+     * @param array<string, mixed>     $options
+     */
+    private function doTextToSpeech(Model $model, array|string $payload, array $options): RawHttpResult
+    {
+        return new RawHttpResult($this->httpClient->request('POST', 'https://api.cartesia.ai/tts/bytes', [
+            'auth_bearer' => $this->apiKey,
+            'headers' => [
+                'Cartesia-Version' => $this->version,
+            ],
+            'json' => [
+                ...$options,
+                'model_id' => $model->getName(),
+                'transcript' => $payload['text'],
+                'voice' => [
+                    'mode' => 'id',
+                    'id' => $options['voice'],
+                ],
+                'output_format' => $options['output_format'],
+            ],
+        ]));
+    }
+
+    /**
+     * @param array<string|int, mixed> $payload
+     * @param array<string, mixed>     $options
+     */
+    private function doSpeechToText(Model $model, array|string $payload, array $options): RawHttpResult
+    {
+        return new RawHttpResult($this->httpClient->request('POST', 'https://api.cartesia.ai/stt', [
+            'auth_bearer' => $this->apiKey,
+            'headers' => [
+                'Cartesia-Version' => $this->version,
+            ],
+            'body' => [
+                ...$options,
+                'model' => $model->getName(),
+                'file' => fopen($payload['input_audio']['path'], 'r'),
+                'timestamp_granularities[]' => 'word',
+            ],
+        ]));
+    }
+}

src/platform/src/Bridge/Cartesia/CartesiaResultConverter.php

@@ -0,0 +1,44 @@
+<?php
+
+/*
+ * This file is part of the Symfony package.
+ *
+ * (c) Fabien Potencier <fabien@symfony.com>
+ *
+ * For the full copyright and license information, please view the LICENSE
+ * file that was distributed with this source code.
+ */
+
+namespace Symfony\AI\Platform\Bridge\Cartesia;
+
+use Symfony\AI\Platform\Exception\RuntimeException;
+use Symfony\AI\Platform\Model;
+use Symfony\AI\Platform\Result\BinaryResult;
+use Symfony\AI\Platform\Result\RawResultInterface;
+use Symfony\AI\Platform\Result\ResultInterface;
+use Symfony\AI\Platform\Result\TextResult;
+use Symfony\AI\Platform\ResultConverterInterface;
+use Symfony\Contracts\HttpClient\ResponseInterface;
+
+/**
+ * @author Guillaume Loulier <personal@guillaumeloulier.fr>
+ */
+final class CartesiaResultConverter implements ResultConverterInterface
+{
+    public function supports(Model $model): bool
+    {
+        return $model instanceof Cartesia;
+    }
+
+    public function convert(RawResultInterface $result, array $options = []): ResultInterface
+    {
+        /** @var ResponseInterface $response */
+        $response = $result->getObject();
+
+        return match (true) {
+            str_contains($response->getInfo('url'), 'stt') => new TextResult($result->getData()['text']),
+            str_contains($response->getInfo('url'), 'tts') => new BinaryResult($result->getObject()->getContent(), 'audio/mpeg'),
+            default => throw new RuntimeException('Unsupported Cartesia response.'),
+        };
+    }
+}