Package bg.codexio.ai.openai.api.http

Interface OpenAIHttpExecutor<I extends Streamable,O extends Mergeable<O>>

Type Parameters:

I - The request (Input) model

O - The response (Output) model

All Known Implementing Classes:

ChatHttpExecutor, CreateImageHttpExecutor, DefaultOpenAIHttpExecutor, EditImageHttpExecutor, ImageVariationHttpExecutor, SpeechHttpExecutor, TranscriptionHttpExecutor, TranslationHttpExecutor, VisionHttpExecutor

public interface OpenAIHttpExecutor<I extends Streamable,O extends Mergeable<O>>

Abstraction over HTTP execution to OpenAPI. It takes a potentially Streamable request, and sends HTTP request, which execution can be immediately executed in a blocking manner, or scheduled for execution in asynchronous (reactive included) fashion.

Nested Class Summary

Nested Classes

Modifier and Type	Interface	Description
static class	OpenAIHttpExecutor.ReactiveExecution<O>	Data holder for the response of executeReactive(Streamable)

Method Summary

Modifier and Type	Method	Description
boolean	canStream(I input)	Denotes whether the response can be streamed, mostly taking into account the value of Streamable.stream() which usually is implemented like inputModel -> inputModel.stream() && specificApiLogic.
O	execute(I request)	Executes HTTP request synchronously
void	executeAsync(I request, Consumer<String> callBack, Consumer<O> finalizer)	Executes HTTP request asynchronously.
OpenAIHttpExecutor.ReactiveExecution<O>	executeReactive(I request)	Executes HTTP request in reactive fashion.

Method Detail

execute

O execute(I request)

Executes HTTP request synchronously

Parameters:

request - The request (Input) model

Returns:

deserialized HTTP Response in the Output model

executeAsync

void executeAsync(I request,
                  Consumer<String> callBack,
                  Consumer<O> finalizer)

Executes HTTP request asynchronously. Since response can be streamed, it can be potentially beneficial for the developer, to subscribe to each line, hence the callback parameter. It makes a little bit more sense to subscribe, to the whole response, using the finalizer parameter.

Parameters:

request - The request (Input) model

callBack - A callback of type stringLine -> consume(stringLine)

finalizer - A callback of type outputModel -> consume(outputModel)

executeReactive

OpenAIHttpExecutor.ReactiveExecution<O> executeReactive(I request)

Executes HTTP request in reactive fashion. We strongly recommend to use this only if a real reactive runtime is present, such as Reactor netty.

Parameters:

request - The request (Input) model

Returns:

OpenAIHttpExecutor.ReactiveExecution object holding a single observable (Mono) to the whole response, and a multi-emit observable (Flux) to each response line.

canStream

boolean canStream(I input)

Denotes whether the response can be streamed, mostly taking into account the value of Streamable.stream() which usually is implemented like inputModel -> inputModel.stream() && specificApiLogic.

Parameters:

input - The request (Input) model

Returns:

boolean saying if you can stream the response or not. Usually beneficial for internal calls like execute(Streamable)