/**
 * Cursor LLM 客户端 - 零成本架构的核心实现
 * 通过 MCP 协议向 Cursor IDE 请求 LLM 补全，实现智能对话
 */
import type { LLMRequest, LLMResponse, ConversationTurn, AccumulatedContext } from './types.js';
export declare class CursorLLMClient {
    private mcpServer;
    constructor(mcpServer: any);
    /**
     * 核心方法：向 Cursor 请求 LLM 补全
     * 这是零成本架构的关键 - 使用用户在 Cursor 中已配置的模型
     */
    requestCompletion(request: LLMRequest): Promise<LLMResponse>;
    /**
     * 意图分类 - 使用专门的分类提示词
     */
    classifyIntent(userInput: string): Promise<{
        project_type: string;
        confidence: number;
        reasoning: string;
    }>;
    /**
     * 构建对话响应 - 使用专家角色卡和上下文
     */
    generateConversationResponse(userInput: string, expertPrompt: string, conversationHistory: ConversationTurn[], accumulatedContext: AccumulatedContext): Promise<{
        response: string;
        type: 'question' | 'suggestion' | 'completion';
        confidence: number;
    }>;
    /**
     * 分析对话准备度 - 判断是否可以生成 PRD
     */
    analyzeReadiness(conversationHistory: ConversationTurn[], accumulatedContext: AccumulatedContext): Promise<{
        ready: boolean;
        confidence: number;
        missing_info: string[];
        next_questions: string[];
    }>;
    private buildIntentClassificationPrompt;
    private buildConversationPrompt;
    private buildReadinessAnalysisPrompt;
    private parseIntentResponse;
    private parseConversationResponse;
    private parseReadinessResponse;
    private estimateConfidence;
    /**
     * 使用 OpenAI 官方 REST 接口的兜底实现
     */
    private requestViaOpenAI;
}
//# sourceMappingURL=llm-client.d.ts.map