Interface ImageEditStreamCapability<TInput, TOutput>

Streaming image editing capability.

Represents the ability to perform image edits while streaming partial or incremental results.

Streaming image editing is designed for:

multi-turn or iterative edits
progressive refinement
long-running or expensive edit operations

This interface mirrors the streaming semantics of chat and image generation capabilities.

Template: TRequest

Client-specific image edit request shape

interface ImageEditStreamCapability<
    TInput = unknown,
    TOutput = NormalizedImage[],
> {
    editImageStream(
        request: AIRequest<TInput>,
        ctx: MultiModalExecutionContext,
        signal?: AbortSignal,
    ): AsyncGenerator<AIResponseChunk<TOutput>>;
}

Type Parameters

TInput = unknown
TOutput = NormalizedImage[]

Hierarchy (View Summary)

ProviderCapability
- ImageEditStreamCapability

Implemented by

Index

Methods

editImageStream

Methods

editImageStream

editImageStream(
    request: AIRequest<TInput>,
    ctx: MultiModalExecutionContext,
    signal?: AbortSignal,
): AsyncGenerator<AIResponseChunk<TOutput>>
Performs a streaming image edit operation.
Parameters
- request: AIRequest<TInput>
  Unified AI request containing:
  
  edit prompt
  
  reference images (base image, mask, style, etc.)
  
  optional provider/model parameters
- ctx: MultiModalExecutionContext
  Execution context
- Optionalsignal: AbortSignal
Returns AsyncGenerator<AIResponseChunk<TOutput>>
AsyncGenerator yielding AIResponseChunk objects containing partial or complete normalized images
- Defined in src/core/provider/capabilities/ImageCapability.ts:193