export interface LoadModelOptions {
    modelPath: string;
    mmprojPath?: string;
    gpuLayers?: number;
    contextSize?: number;
    threads?: number;
    debug?: boolean;
    /**
     * Print the final chat-template-rendered prompt sent to llama.cpp to stderr.
     *
     * This may include private user data. Intended for local debugging only.
     */
    logPrompts?: boolean;
    /**
     * Chat template to use for formatting messages.
     * - "auto" (default): Use the template embedded in the GGUF model file
     * - Template name: Use a specific built-in template (e.g., "llama3", "chatml", "gemma")
     */
    chatTemplate?: string;
    /**
     * Whether to load the model for embedding generation.
     * When true, creates an embedding context with mean pooling enabled.
     * Default: false
     */
    embedding?: boolean;
}
export interface ChatMessage {
    role: string;
    content: string;
    images?: ImageInput[];
    toolCalls?: NativeToolCall[];
    toolName?: string;
    toolCallId?: string;
}
export interface ImageInput {
    data: Uint8Array;
    mediaType?: string;
}
export interface NativeToolCall {
    id?: string;
    name: string;
    arguments: string;
}
export interface ToolDefinition {
    name: string;
    description?: string;
    parametersJson: string;
}
export interface GenerateOptions {
    /** Internal ID used to cancel one in-flight generation without affecting others. */
    requestId?: string;
    messages: ChatMessage[];
    tools?: ToolDefinition[];
    toolChoice?: "auto" | "required" | "none";
    parallelToolCalls?: boolean;
    maxTokens?: number;
    temperature?: number;
    topP?: number;
    topK?: number;
    seed?: number;
    stopSequences?: string[];
    /** GBNF grammar string for structured output */
    grammar?: string;
    /** Maximum generated tokens allowed inside the reasoning block. */
    reasoningBudgetTokens?: number;
    /** Marker that starts model reasoning content. */
    reasoningBudgetStart?: string;
    /** Marker that ends model reasoning content. */
    reasoningBudgetEnd?: string;
    /** Whether chat templates that support thinking should enable thinking. */
    enableThinking?: boolean;
    /** Reuse matching prompt prefixes from the previous request on this model. */
    promptCache?: boolean;
}
export interface GenerateResult {
    text: string;
    promptTokens: number;
    completionTokens: number;
    cacheReadTokens?: number;
    cacheWriteTokens?: number;
    finishReason: "stop" | "length" | "error";
    errorMessage?: string;
    toolCalls?: NativeToolCall[];
}
export interface EmbedOptions {
    texts: string[];
}
export interface EmbedResult {
    embeddings: Float32Array[];
    totalTokens: number;
}
export declare function loadModel(options: LoadModelOptions): Promise<number>;
export declare function unloadModel(handle: number): boolean;
export declare function generate(handle: number, options: GenerateOptions): Promise<GenerateResult>;
export declare function cancelGeneration(handle: number, requestId?: string): boolean;
export declare function generateStream(handle: number, options: GenerateOptions, onToken: (token: string) => void): Promise<GenerateResult>;
export declare function isModelLoaded(handle: number): boolean;
export declare function embed(handle: number, options: EmbedOptions): Promise<EmbedResult>;
//# sourceMappingURL=native-binding.d.ts.map