Struct OpenAiAdapter

pub struct OpenAiAdapter { /* private fields */ }

Thin wrapper that wires the HTTP client [OpenAiClient] into a value that implements artificial_core::backend::Backend.

Think of it as the service locator for the OpenAI back-end:

• stores the API key (and optionally a custom base URL in the future),
• owns a shareable, connection-pooled reqwest::Client,
• provides a fluent OpenAiAdapterBuilder so callers don’t have to juggle Option<String> manually.

The type itself purposefully exposes no additional methods—all user- facing functionality sits on the generic artificial_core::ArtificialClient once the adapter is plugged in.

Trait Implementations

impl Backend for OpenAiAdapter

Implementation of Backend for the OpenAiAdapter.

The type is only a thin glue layer—almost all heavy lifting is done by the inner OpenAiClient (HTTP) and the adapter’s config (API key, base URL, …).

Responsibilities:

1. Convert the generic prompt into OpenAI‐compatible chat messages.
2. Enrich the request with a JSON Schema derived from Prompt::Output.
3. Call the /v1/chat/completions endpoint and bubble up transport errors.
4. Validate & deserialize the returned JSON into Prompt::Output.

The implementation purposefully rejects any streaming or multi-choice responses for now; this keeps the surface minimal and makes error handling easier to reason about.

type Message = ChatCompletionMessage

Provider-specific chat message type.

fn chat_complete<P>( &self, prompt: P, ) -> Pin<Box<dyn Future<Output = Result<P::Output>> + Send>>

where P: PromptTemplate + Send + 'static, <P as IntoPrompt>::Message: Into<Self::Message>,

Perform a non-streaming chat completion and deserialize the result.

The method is object-safe by returning a boxed Future rather than using async/await syntax directly.