Class OpenAIImageGenerationCapabilityImpl

Adapts OpenAI image generation responses into ProviderPlaneAI's normalized image artifact surface.

Supports both non-streaming and streaming generation through the Responses API and normalizes generated image payloads into NormalizedImage[].

Implements

Index

Constructors

Methods

generateImage generateImageStream

Constructors

constructor

new OpenAIImageGenerationCapabilityImpl(
provider: BaseProvider,
client: OpenAI,
): OpenAIImageGenerationCapabilityImpl
Creates a new OpenAI image generation capability adapter.
Parameters
- provider: BaseProvider
  Owning provider instance used for initialization checks and merged config access.
- client: OpenAI
  Initialized OpenAI SDK client.
Returns OpenAIImageGenerationCapabilityImpl
- Defined in src/providers/openai/capabilities/OpenAIImageGenerationCapabilityImpl.ts:44

Methods

generateImage

generateImage(
    request: AIRequest<ClientImageGenerationRequest>,
    _executionContext?: MultiModalExecutionContext,
    signal?: AbortSignal,
): Promise<AIResponse<NormalizedImage[]>>
Generates images using OpenAI Responses API (non-streaming).

Steps:
- Validates prompt
- Merges provider/model/request options
- Builds prompt and reference image content
- Sends concurrent requests if count > 1
- Normalizes and returns image output
Parameters
- request: AIRequest<ClientImageGenerationRequest>
  Unified image generation request envelope.
- Optional_executionContext: MultiModalExecutionContext
  Optional multimodal execution context. Unused directly in this adapter.
- Optionalsignal: AbortSignal
  Optional cancellation signal.
Returns Promise<AIResponse<NormalizedImage[]>>
Provider-normalized generated image artifacts.

Throws
If the prompt is missing or generation returns no image artifacts.
Implementation of ImageGenerationCapability.generateImage
- Defined in src/providers/openai/capabilities/OpenAIImageGenerationCapabilityImpl.ts:65

generateImageStream

generateImageStream(
    request: AIRequest<ClientImageGenerationRequest>,
    _executionContext?: MultiModalExecutionContext,
    signal?: AbortSignal,
): AsyncGenerator<AIResponseChunk<NormalizedImage[]>>
Streaming image generation using OpenAI Responses API. Emits partial image chunks as they become available.

Note: OpenAI image generation streams lifecycle events only. Image payloads are delivered atomically once generation completes. Consumers should expect exactly one image-bearing chunk.

Steps:
- Validates prompt
- Merges provider/model/request options
- Builds prompt and reference image content
- Starts streaming generation for each image
- Yields image chunks as they are received
- Handles errors by yielding an error chunk
Parameters
- request: AIRequest<ClientImageGenerationRequest>
  Unified image generation request envelope.
- Optional_executionContext: MultiModalExecutionContext
  Optional multimodal execution context. Unused directly in this adapter.
- Optionalsignal: AbortSignal
  Optional cancellation signal.
Returns AsyncGenerator<AIResponseChunk<NormalizedImage[]>>
Async generator of incremental image chunks and a terminal completion chunk.

Throws
If the prompt is missing.
Implementation of ImageGenerationStreamCapability.generateImageStream
- Defined in src/providers/openai/capabilities/OpenAIImageGenerationCapabilityImpl.ts:147