feat(tts/gemini): Added Gemini TTS support

src/Providers/Gemini/Gemini.php

@@ -5,30 +5,33 @@
 namespace Prism\Prism\Providers\Gemini;
 
 use Generator;
-use Illuminate\Http\Client\PendingRequest;
-use Illuminate\Http\Client\RequestException;
-use Prism\Prism\Concerns\InitializesClient;
 use Prism\Prism\Contracts\Message;
-use Prism\Prism\Embeddings\Request as EmbeddingRequest;
-use Prism\Prism\Embeddings\Response as EmbeddingResponse;
+use Prism\Prism\Providers\Provider;
+use Prism\Prism\Audio\AudioResponse as TextToSpeechResponse;
+use Illuminate\Http\Client\PendingRequest;
+use Prism\Prism\Audio\TextToSpeechRequest;
 use Prism\Prism\Exceptions\PrismException;
-use Prism\Prism\Exceptions\PrismProviderOverloadedException;
-use Prism\Prism\Exceptions\PrismRateLimitedException;
+use Prism\Prism\Concerns\InitializesClient;
+use Illuminate\Http\Client\RequestException;
+use Prism\Prism\Text\Request as TextRequest;
+use Prism\Prism\Text\Response as TextResponse;
+use Prism\Prism\Providers\Gemini\Handlers\Text;
 use Prism\Prism\Images\Request as ImagesRequest;
-use Prism\Prism\Images\Response as ImagesResponse;
+use Prism\Prism\Providers\Gemini\Handlers\Audio;
 use Prism\Prism\Providers\Gemini\Handlers\Cache;
-use Prism\Prism\Providers\Gemini\Handlers\Embeddings;
 use Prism\Prism\Providers\Gemini\Handlers\Images;
 use Prism\Prism\Providers\Gemini\Handlers\Stream;
+use Prism\Prism\Images\Response as ImagesResponse;
+use Prism\Prism\ValueObjects\Messages\SystemMessage;
+use Prism\Prism\Exceptions\PrismRateLimitedException;
+use Prism\Prism\Providers\Gemini\Handlers\Embeddings;
 use Prism\Prism\Providers\Gemini\Handlers\Structured;
-use Prism\Prism\Providers\Gemini\Handlers\Text;
-use Prism\Prism\Providers\Gemini\ValueObjects\GeminiCachedObject;
-use Prism\Prism\Providers\Provider;
+use Prism\Prism\Embeddings\Request as EmbeddingRequest;
 use Prism\Prism\Structured\Request as StructuredRequest;
+use Prism\Prism\Embeddings\Response as EmbeddingResponse;
 use Prism\Prism\Structured\Response as StructuredResponse;
-use Prism\Prism\Text\Request as TextRequest;
-use Prism\Prism\Text\Response as TextResponse;
-use Prism\Prism\ValueObjects\Messages\SystemMessage;
+use Prism\Prism\Exceptions\PrismProviderOverloadedException;
+use Prism\Prism\Providers\Gemini\ValueObjects\GeminiCachedObject;
 
 class Gemini extends Provider
 {
@@ -83,6 +86,17 @@ public function images(ImagesRequest $request): ImagesResponse
         return $handler->handle($request);
     }
 
+    #[\Override]
+    public function textToSpeech(TextToSpeechRequest $request): TextToSpeechResponse
+    {
+        $handler = new Audio($this->client(
+            $request->clientOptions(),
+            $request->clientRetry()
+        ));
+
+        return $handler->handleTextToSpeech($request);
+    }
+
     #[\Override]
     public function stream(TextRequest $request): Generator
     {

src/Providers/Gemini/Handlers/Audio.php

@@ -0,0 +1,39 @@
+<?php 
+
+namespace Prism\Prism\Providers\Gemini\Handlers;
+
+use Exception;
+use Prism\Prism\Audio\AudioResponse;
+use Illuminate\Http\Client\PendingRequest;
+use Prism\Prism\Audio\TextToSpeechRequest;
+use Prism\Prism\ValueObjects\GeneratedAudio;
+use Prism\Prism\Providers\Gemini\Concerns\ValidatesResponse;
+
+class Audio
+{
+    use ValidatesResponse;
+
+    public function __construct(protected PendingRequest $client) {}
+
+    public function handleTextToSpeech(TextToSpeechRequest $request): AudioResponse
+    {
+        $mapper = new TextToSpeechRequestMapper($request);
+
+        $response = $this->client->post("{$request->model()}:generateContent", $mapper->toPayload());
+
+        if (! $response->successful()) {
+            throw new Exception('Failed to generate audio: '.$response->body());
+        }
+
+        $data = $response->json();
+
+        $base64Audio = $data['candidates'][0]['content']['parts'][0]['inlineData']['data']
+            ??  throw new Exception('No audio data returned from TTS API');
+
+        return new AudioResponse(
+            audio: new GeneratedAudio(
+                base64: $base64Audio,
+            ),
+        );
+    }
+}

src/Providers/Gemini/Handlers/TextToSpeechRequestMapper.php

@@ -0,0 +1,122 @@
+<?php
+
+namespace Prism\Prism\Providers\Gemini\Handlers;
+
+use Illuminate\Support\Arr;
+use Prism\Prism\Enums\Provider;
+use Prism\Prism\Audio\TextToSpeechRequest;
+
+class TextToSpeechRequestMapper
+{
+    public function __construct(
+        public readonly TextToSpeechRequest $request
+    ) {}
+
+    /**
+     * @return array<string, mixed>
+     */
+    public function toPayload(): array
+    {
+        $providerOptions = $this->request->providerOptions();
+
+        $contents = [
+            'parts' => [
+                [
+                    'text' => $this->request->input(),
+                ],
+            ],
+        ];
+
+        $baseData = [
+            'model' => $this->request->model(),
+            'contents' => [$contents],
+        ];
+
+        $speechConfig = $this->buildSpeechConfig($providerOptions);
+
+        $generationConfig = Arr::whereNotNull([
+            'responseModalities' => $providerOptions['responseModalities'] ?? ['AUDIO'],
+            'speechConfig' => $speechConfig !== [] ? $speechConfig : null,
+        ]);
+
+        $supportedOptions = Arr::whereNotNull([
+            'generationConfig' => $generationConfig !== [] ? $generationConfig : null,
+        ]);
+
+        return array_merge( 
+            $baseData,
+            $supportedOptions,
+        );
+    }
+
+
+    /**
+     * @param array<string, mixed> $providerOptions
+     * @return array<string, mixed>
+     */
+    protected function buildSpeechConfig(array $providerOptions): array
+    {
+        if (isset($providerOptions['multiSpeaker']) && is_array($providerOptions['multiSpeaker'])) {
+            $multiSpeakerConfig = $this->buildMultiSpeakerConfig($providerOptions['multiSpeaker']);
+
+            if ($multiSpeakerConfig !== []) {
+                return $multiSpeakerConfig;
+            }
+        }
+
+        if ($this->request->voice()) {
+            return $this->buildSingleVoiceConfig($this->request->voice());
+        }
+
+        return [];
+    }
+
+    /**
+     * @return array<string, array<string, array<string, string>>>
+     */
+    protected function buildSingleVoiceConfig(string $voiceName): array
+    {
+        return [
+            'voiceConfig' => [
+                'prebuiltVoiceConfig' => [
+                    'voiceName' => $voiceName,
+                ],
+            ],
+        ];
+    }
+
+    /**
+     * @param array<int, array{speaker?: string, voiceName?: string}> $speakers
+     * @return array<string, mixed>
+     */
+    protected function buildMultiSpeakerConfig(array $speakers): array
+    {
+        $speakerVoiceConfigs = [];
+
+        foreach ($speakers as $speaker) {
+            if (!isset($speaker['speaker']) || !isset($speaker['voiceName'])) {
+                continue;
+            }
+
+            $speakerVoiceConfigs[] = [
+                'speaker' => $speaker['speaker'],
+                'voiceConfig' => [
+                    'prebuiltVoiceConfig' => [
+                        'voiceName' => $speaker['voiceName'],
+                    ],
+                ],
+            ];
+        }
+
+        return $speakerVoiceConfigs !== [] ? [
+            'multiSpeakerVoiceConfig' => [
+                'speakerVoiceConfigs' => $speakerVoiceConfigs,
+            ],
+        ] : [];
+    }
+
+    protected function provider(): string|Provider
+    {
+        return Provider::Gemini;
+    }
+}

tests/Fixtures/gemini/tts-flash-1-1.json

@@ -0,0 +1,16 @@
+{
+  "candidates": [
+    {
+      "content": {
+        "parts": [
+          {
+            "inlineData": {
+              "mimeType": "audio/wav",
+              "data": "UklGRiQAAABXQVZFZm10IBAAAAABAAEAQB8AAEAfAAABAAgAZGF0YQAAAAA="
+            }
+          }
+        ]
+      }
+    }
+  ]
+}

tests/Fixtures/gemini/tts-flash-multi-speaker-1.json

@@ -0,0 +1,16 @@
+{
+  "candidates": [
+    {
+      "content": {
+        "parts": [
+          {
+            "inlineData": {
+              "mimeType": "audio/wav",
+              "data": "UklGRiQAAABXQVZFZm10IBAAAAABAAEAQB8AAEAfAAABAAgAZGF0YQAAAAA="
+            }
+          }
+        ]
+      }
+    }
+  ]
+}

tests/Fixtures/gemini/tts-flash-voice-option-1.json

@@ -0,0 +1,16 @@
+{
+  "candidates": [
+    {
+      "content": {
+        "parts": [
+          {
+            "inlineData": {
+              "mimeType": "audio/wav",
+              "data": "UklGRiQAAABXQVZFZm10IBAAAAABAAEAQB8AAEAfAAABAAgAZGF0YQAAAAA="
+            }
+          }
+        ]
+      }
+    }
+  ]
+}

tests/Fixtures/gemini/tts-pro-1-1.json

@@ -0,0 +1,16 @@
+{
+  "candidates": [
+    {
+      "content": {
+        "parts": [
+          {
+            "inlineData": {
+              "mimeType": "audio/wav",
+              "data": "UklGRiQAAABXQVZFZm10IBAAAAABAAEAQB8AAEAfAAABAAgAZGF0YQAAAAA="
+            }
+          }
+        ]
+      }
+    }
+  ]
+}

tests/Fixtures/gemini/tts-pro-multi-speaker-1.json

@@ -0,0 +1,16 @@
+{
+  "candidates": [
+    {
+      "content": {
+        "parts": [
+          {
+            "inlineData": {
+              "mimeType": "audio/wav",
+              "data": "UklGRiQAAABXQVZFZm10IBAAAAABAAEAQB8AAEAfAAABAAgAZGF0YQAAAAA="
+            }
+          }
+        ]
+      }
+    }
+  ]
+}

tests/Fixtures/gemini/tts-pro-voice-option-1.json

@@ -0,0 +1,16 @@
+{
+  "candidates": [
+    {
+      "content": {
+        "parts": [
+          {
+            "inlineData": {
+              "mimeType": "audio/wav",
+              "data": "UklGRiQAAABXQVZFZm10IBAAAAABAAEAQB8AAEAfAAABAAgAZGF0YQAAAAA="
+            }
+          }
+        ]
+      }
+    }
+  ]
+}