import type { EmbeddingModelV4, EmbeddingModelV4CallOptions, EmbeddingModelV4Result } from "@ai-sdk/provider";
import type { LlamaCppProviderConfig } from "./llama-cpp-provider-config.js";
export declare class LlamaCppEmbeddingModel implements EmbeddingModelV4 {
    readonly specificationVersion: "v4";
    readonly provider = "llama.cpp";
    readonly modelId: string;
    /**
     * Maximum number of embeddings that can be generated in a single call.
     * Local models can handle large batches, but we limit to prevent memory issues.
     */
    readonly maxEmbeddingsPerCall: number;
    /**
     * Whether the model supports parallel calls.
     * We use a single model instance, so parallel calls are not supported.
     */
    readonly supportsParallelCalls: boolean;
    private modelHandle;
    private readonly config;
    private initPromise;
    constructor(config: LlamaCppProviderConfig);
    private ensureModelLoaded;
    /**
     * Dispose of the model and free resources.
     */
    dispose(): Promise<void>;
    doEmbed(options: EmbeddingModelV4CallOptions): Promise<EmbeddingModelV4Result>;
}
//# sourceMappingURL=llama-cpp-embedding-model.d.ts.map