Class OpenAIImageProvider

constructor

• new OpenAIImageProvider(): OpenAIImageProvider

• Returns OpenAIImageProvider

initialize

• initialize(config): Promise<void>

• Parameters

  • config: Record<string, unknown>

  Returns Promise<void>

generateImage

• generateImage(request): Promise<ImageGenerationResult>

• Parameters

  • request: ImageGenerationRequest

  Returns Promise<ImageGenerationResult>

editImage

• editImage(request): Promise<ImageGenerationResult>

• Edits an image using the OpenAI /v1/images/edits endpoint.

  Supports both img2img (prompt-guided transformation) and inpainting (mask-guided regional editing). The endpoint expects multipart form data with the source image and an optional mask.

  Parameters

  • request: ImageEditRequest

    Edit request with source image buffer and prompt.

  Returns Promise<ImageGenerationResult>

  Generation result containing the edited image(s).

  Throws

  When the provider is not initialised.

  Throws

  When the API returns an HTTP error status.

  See

  https://platform.openai.com/docs/api-reference/images/createEdit

styleTransferModel

• styleTransferModel(modelId): string

• The model a style transfer runs on for modelId, or a refusal. An empty id takes the configured default when it is one of OPENAI_STYLE_TRANSFER_MODELS, else Sunburst. A named model must be one of them, unless the provider's base URL is not OpenAI's: a custom endpoint may serve any name.

  Parameters

  • modelId: string

    The model the call names, or an empty string.

  Returns string

  Throws

  For a model the style transfer does not run.

transferStyle

• transferStyle(request): Promise<ImageGenerationResult>

• Restyles an image after a style reference through /v1/images/edits, which takes "up to 16 images" for the GPT Image models (openai-node, src/resources/images.ts): the source and the reference go as two image[] parts, the source first, with an instruction that names them.

  The endpoint has no strength, seed or negative prompt: strength picks the instruction (see styleTransfer.ts), and the others are not sent. Before the request, the images are checked to be PNG, WebP or JPEG files under 50 MB each, and the prompt as sent to be at most 32,000 characters.

  Parameters

  • request: ImageStyleTransferRequest

    The source, the reference and the prompt.

  Returns Promise<ImageGenerationResult>

  Generation result with the restyled image(s).

  Throws

  When the provider is not initialised, the model, an image or the prompt is refused, the API returns an HTTP error status, or the answer holds no image.

variateImage

• variateImage(request): Promise<ImageGenerationResult>

• Makes variations of an image through the /v1/images/edits endpoint, with a GPT image model and a prompt that asks for a variation. OpenAI retired /v1/images/variations and the DALL·E models it served: "Use the image edits endpoint with a GPT Image model and a prompt to create a variation of an image" (openai-node, src/resources/images.ts).

  The edits endpoint has no strength parameter, so variance picks the prompt: below 0.5 asks for a close variation, otherwise a looser one.

  Parameters

  • request: ImageVariateRequest

    Variation request with the source image buffer.

  Returns Promise<ImageGenerationResult>

  Generation result containing the variation image(s).

  Throws

  When the provider is not initialised.

  Throws

  When the API returns an HTTP error status.

  See

  https://github.com/openai/openai-node/blob/master/src/resources/images.ts

listAvailableModels

• listAvailableModels(): Promise<ImageModelInfo[]>

• Returns Promise<ImageModelInfo[]>

Readonly providerId

isInitialized

Optional defaultModelId