@aeye/openrouter — API Reference

OpenRouterProvider

Extends OpenAIProvider with OpenRouter-specific features.

class OpenRouterProvider extends OpenAIProvider<OpenRouterConfig>

OpenRouterConfig

Extends OpenAIConfig (minus api) with:

interface OpenRouterConfig extends Omit<OpenAIConfig, 'api'> {
  defaultParams?: {
    siteUrl?: string;
    appName?: string;
    providers?: {
      order?: string[];
      allowFallbacks?: boolean;
      requireParameters?: boolean;      // auto-on for strict json_schema when undefined
      dataCollection?: 'deny' | 'allow';
      zdr?: boolean;
      enforceDistillableText?: boolean; // only providers that allow training on outputs
      only?: string[];
      ignore?: string[];
      quantizations?: OpenRouterQuantization[]; // 'int4' | 'int8' | 'fp4' | 'mxfp4' | 'nvfp4' | 'fp6' | 'fp8' | 'mxfp8' | 'fp16' | 'bf16' | 'fp32' | 'unknown'
      sort?: OpenRouterSort | { by: OpenRouterSort; partition?: 'model' | 'none' }; // 'price' | 'throughput' | 'latency' | 'exacto'
      preferredMinThroughput?: number;  // tokens/second, median
      preferredMaxLatency?: number;     // seconds, median
      maxPrice?: { prompt?: number; completion?: number; image?: number; audio?: number; request?: number };
    };
    fallbackModels?: string[];          // sent as `models`
    plugins?: OpenRouterPlugin[];       // file-parser | response-healing | context-compression
    /** @deprecated `middle-out` is sent as the `context-compression` plugin */
    transforms?: string[];
  };
}

OpenRouterConfig omits OpenAIConfig's api option: requests always go through Chat Completions, where routing, plugins, server tools, and reasoning options work.

OpenRouterModelSource

Dynamic model source for fetching available models:

class OpenRouterModelSource implements ModelSource {
  constructor(config?: { apiKey?: string; includeZDR?: boolean });
  fetchModels(): Promise<ModelInfo[]>;
}

Exported Functions

fetchModels(apiKey?)

Fetch available models from OpenRouter API, including speech, transcription, embedding, and image models.

fetchZDRModels(apiKey?)

Fetch models that support Zero Data Retention.

fetchEndpointMetrics(modelId, apiKey?)

Best median latency (ms), median throughput (tokens/s), and uptime across the providers serving a model, from /models/{id}/endpoints.