Implement image generation support

cli.php

@@ -90,7 +90,7 @@ function logError(string $message, int $exit_code = 1): void
 
 // Prompt input. Allow complex input as a JSON string.
 $promptInput = $positional_args[0];
-if (strpos($promptInput, '{') === 0 || strpos($promptInput, '[') === 0) {
+if (str_starts_with($promptInput, '{') || str_starts_with($promptInput, '[')) {
     $decodedInput = json_decode($promptInput, true);
     if ($decodedInput) {
         $promptInput = $decodedInput;
@@ -156,7 +156,11 @@ function logError(string $message, int $exit_code = 1): void
 }
 
 try {
-    $result = $promptBuilder->generateTextResult();
+    if ($outputFormat === 'image-json' || $outputFormat === 'image-base64') {
+        $result = $promptBuilder->generateImageResult();
+    } else {
+        $result = $promptBuilder->generateTextResult();
+    }
 } catch (InvalidArgumentException $e) {
     logError('Invalid arguments while trying to generate text result: ' . $e->getMessage());
 } catch (ResponseException $e) {
@@ -173,6 +177,12 @@ function logError(string $message, int $exit_code = 1): void
     case 'candidates-json':
         $output = json_encode($result->getCandidates(), JSON_PRETTY_PRINT);
         break;
+    case 'image-json':
+        $output = json_encode($result->toFile(), JSON_PRETTY_PRINT);
+        break;
+    case 'image-base64':
+        $output = $result->toFile()->getBase64Data();
+        break;
     case 'message-text':
     default:
         $output = $result->toText();

src/Common/AbstractEnum.php

@@ -328,7 +328,7 @@ protected static function determineClassEnumerations(string $className): array
     final public function __call(string $name, array $arguments): bool
     {
         // Handle is* methods
-        if (strpos($name, 'is') === 0) {
+        if (str_starts_with($name, 'is')) {
             $constantName = self::camelCaseToConstant(substr($name, 2));
             $constants = static::getConstants();
 

src/Files/ValueObjects/MimeType.php

@@ -204,7 +204,7 @@ public static function isValid(string $mimeType): bool
      */
     public function isType(string $mimeType): bool
     {
-        return strpos($this->value, strtolower($mimeType) . '/') === 0;
+        return str_starts_with($this->value, strtolower($mimeType) . '/');
     }
 
     /**

src/ProviderImplementations/Google/GoogleImageGenerationModel.php

@@ -0,0 +1,30 @@
+<?php
+
+declare(strict_types=1);
+
+namespace WordPress\AiClient\ProviderImplementations\Google;
+
+use WordPress\AiClient\Providers\Http\DTO\Request;
+use WordPress\AiClient\Providers\Http\Enums\HttpMethodEnum;
+use WordPress\AiClient\Providers\OpenAiCompatibleImplementation\AbstractOpenAiCompatibleImageGenerationModel;
+
+/**
+ * Class for a Google image generation model.
+ *
+ * @since n.e.x.t
+ */
+class GoogleImageGenerationModel extends AbstractOpenAiCompatibleImageGenerationModel
+{
+    /**
+     * @inheritDoc
+     */
+    protected function createRequest(HttpMethodEnum $method, string $path, array $headers = [], $data = null): Request
+    {
+        return new Request(
+            $method,
+            GoogleProvider::BASE_URI . '/openai/' . ltrim($path, '/'),
+            $headers,
+            $data
+        );
+    }
+}

src/ProviderImplementations/Google/GoogleModelMetadataDirectory.php

@@ -158,10 +158,13 @@ protected function parseResponseToModelMetadataList(Response $response): array
             new SupportedOption(OptionEnum::outputFileType(), [FileTypeEnum::inline()]),
             new SupportedOption(OptionEnum::outputMediaOrientation(), [
                 MediaOrientationEnum::square(),
-                MediaOrientationEnum::landscape(),
-                MediaOrientationEnum::portrait(),
+                // The following orientations are normally supported, but not when using the OpenAI compatible endpoint.
+                // MediaOrientationEnum::landscape(),
+                // MediaOrientationEnum::portrait(),
             ]),
-            new SupportedOption(OptionEnum::outputMediaAspectRatio(), ['1:1', '16:9', '4:3', '9:16', '3:4']),
+            // Aspect ratio is normally supported, but not when using the OpenAI compatible endpoint.
+            // new SupportedOption(OptionEnum::outputMediaAspectRatio(), ['1:1', '16:9', '4:3', '9:16', '3:4']),
+            new SupportedOption(OptionEnum::customOptions()),
         ];
 
         $modelsData = (array) $responseData['models'];

src/ProviderImplementations/Google/GoogleProvider.php

@@ -38,10 +38,7 @@ protected static function createModel(
                 return new GoogleTextGenerationModel($modelMetadata, $providerMetadata);
             }
             if ($capability->isImageGeneration()) {
-                // TODO: Implement GoogleImageGenerationModel.
-                throw new RuntimeException(
-                    'Google image generation model class is not yet implemented.'
-                );
+                return new GoogleImageGenerationModel($modelMetadata, $providerMetadata);
             }
         }
 

src/ProviderImplementations/OpenAi/OpenAiImageGenerationModel.php

@@ -0,0 +1,50 @@
+<?php
+
+declare(strict_types=1);
+
+namespace WordPress\AiClient\ProviderImplementations\OpenAi;
+
+use WordPress\AiClient\Providers\Http\DTO\Request;
+use WordPress\AiClient\Providers\Http\Enums\HttpMethodEnum;
+use WordPress\AiClient\Providers\OpenAiCompatibleImplementation\AbstractOpenAiCompatibleImageGenerationModel;
+
+/**
+ * Class for an OpenAI image generation model.
+ *
+ * @since n.e.x.t
+ */
+class OpenAiImageGenerationModel extends AbstractOpenAiCompatibleImageGenerationModel
+{
+    /**
+     * @inheritDoc
+     */
+    protected function createRequest(HttpMethodEnum $method, string $path, array $headers = [], $data = null): Request
+    {
+        return new Request(
+            $method,
+            OpenAiProvider::BASE_URI . '/' . ltrim($path, '/'),
+            $headers,
+            $data
+        );
+    }
+
+    /**
+     * @inheritDoc
+     */
+    protected function prepareGenerateImageParams(array $prompt): array
+    {
+        $params = parent::prepareGenerateImageParams($prompt);
+
+        /*
+         * Only the newer 'gpt-image-' models support passing a MIME type ('output_format').
+         * Conversely, they do not support 'response_format', but always return a base64 encoded image.
+         */
+        if (isset($params['model']) && is_string($params['model']) && str_starts_with($params['model'], 'gpt-image-')) {
+            unset($params['response_format']);
+        } else {
+            unset($params['output_format']);
+        }
+
+        return $params;
+    }
+}

src/ProviderImplementations/OpenAi/OpenAiModelMetadataDirectory.php

@@ -126,6 +126,7 @@ protected function parseResponseToModelMetadataList(Response $response): array
                 MediaOrientationEnum::portrait(),
             ]),
             new SupportedOption(OptionEnum::outputMediaAspectRatio(), ['1:1', '7:4', '4:7']),
+            new SupportedOption(OptionEnum::customOptions()),
         ];
         $gptImageOptions = [
             new SupportedOption(OptionEnum::inputModalities(), [[ModalityEnum::text()]]),
@@ -139,6 +140,7 @@ protected function parseResponseToModelMetadataList(Response $response): array
                 MediaOrientationEnum::portrait(),
             ]),
             new SupportedOption(OptionEnum::outputMediaAspectRatio(), ['1:1', '3:2', '2:3']),
+            new SupportedOption(OptionEnum::customOptions()),
         ];
         $ttsCapabilities = [
             CapabilityEnum::textToSpeechConversion(),
@@ -154,6 +156,7 @@ protected function parseResponseToModelMetadataList(Response $response): array
                 'audio/aac',
             ]),
             new SupportedOption(OptionEnum::outputSpeechVoice()),
+            new SupportedOption(OptionEnum::customOptions()),
         ];
 
         $modelsData = (array) $responseData['data'];

src/ProviderImplementations/OpenAi/OpenAiProvider.php

@@ -38,10 +38,7 @@ protected static function createModel(
                 return new OpenAiTextGenerationModel($modelMetadata, $providerMetadata);
             }
             if ($capability->isImageGeneration()) {
-                // TODO: Implement OpenAiImageGenerationModel.
-                throw new RuntimeException(
-                    'OpenAI image generation model class is not yet implemented.'
-                );
+                return new OpenAiImageGenerationModel($modelMetadata, $providerMetadata);
             }
             if ($capability->isTextToSpeechConversion()) {
                 // TODO: Implement OpenAiTextToSpeechConversionModel.

src/Providers/Models/Enums/OptionEnum.php

@@ -97,7 +97,7 @@ protected static function determineClassEnumerations(string $className): array
 
         // Add ModelConfig constants that start with KEY_
         foreach ($modelConfigConstants as $constantName => $constantValue) {
-            if (strpos($constantName, 'KEY_') === 0) {
+            if (str_starts_with($constantName, 'KEY_')) {
                 // Remove KEY_ prefix to get the enum constant name
                 $enumConstantName = substr($constantName, 4);