@aeye/openai — API Reference

OpenAIProvider

class OpenAIProvider<TConfig extends OpenAIConfig = OpenAIConfig>
  implements Provider<TConfig>

Requires Node >= 22 and the openai SDK v7.

Constructor

new OpenAIProvider(config: OpenAIConfig)

OpenAIConfig

interface OpenAIConfig {
  apiKey: string;
  baseURL?: string;
  organization?: string;
  project?: string;
  store?: boolean;
  api?: 'auto' | 'chat' | 'responses'; // default 'auto'
  retry?: RetryConfig;
  retryEvents?: RetryEvents;
  defaultModels?: {
    chat?: ModelInput;
    imageGenerate?: ModelInput;   // default 'gpt-image-2'
    imageEdit?: ModelInput;       // default 'gpt-image-2'
    imageAnalyze?: ModelInput;    // default 'gpt-5.4-mini'
    transcription?: ModelInput;   // default 'gpt-transcribe'
    speech?: ModelInput;          // default 'gpt-4o-mini-tts'
    embedding?: ModelInput;       // default 'text-embedding-3-small'
  };
  hooks?: OpenAIHooks;
}

With api: 'auto', chat uses Chat Completions unless the model is Responses-only (metadata.responsesOnly in its registry entry, or by id: codex, pre-5.6 pro, deep-research, computer-use), the request has hostedTools, or reason sets mode, summary, or context; then it uses the Responses API.

Subclasses can override the protected defaults property to supply their own default model ids.

RetryConfig

interface RetryConfig {
  maxRetries?: number;
  initialDelay?: number;
  maxDelay?: number;
  backoffMultiplier?: number;
  jitter?: boolean;
  retryableStatuses?: number[];
  timeout?: number;
}

OpenAIHooks

type OpenAIHooks = {
  [K in keyof OpenAIOperations]?: {
    beforeRequest?: (request, params, ctx, metadata) => void;
    afterRequest?: (request, params, response, ctx, metadata) => void;
    onError?: (request, params, error, ctx, metadata) => void;
  };
};
// Operations: chat, responses, imageGenerate, imageEdit, transcribe, speech, embed

Image analysis runs as chat, so it reports through the chat (or responses) hooks.

Error Types

Class	Description
ProviderError	Base provider error
ProviderAuthError	Authentication failure
RateLimitError	Rate limit exceeded
ProviderRateLimitError	Provider-specific rate limit
ProviderQuotaError	Account quota exceeded
ContextWindowError	Input exceeds context window