route-claudecode/dist/preprocessing/max-token-preprocessor.d.ts

Advanced routing and transformation system for Claude Code outputs to multiple AI providers

/**
 * Max Token Preprocessing Module
 * 最大Token预处理模块
 * Owner: Jason Zhang
 *
 * 提供多种max token处理策略：
 * 1. 动态截断历史记录
 * 2. 路由重定向到long context模型
 * 3. Long context模型压缩
 */
export interface MaxTokenStrategy {
    name: string;
    enabled: boolean;
    priority: number;
}
export interface DynamicTruncationConfig extends MaxTokenStrategy {
    name: 'dynamic_truncation';
    truncatePosition: 'head' | 'tail' | 'middle';
    tokenRatio: number;
    preserveSystemPrompt: boolean;
    preserveLatestMessages: number;
    enableSimplifiedTools: boolean;
}
export interface RouteRedirectionConfig extends MaxTokenStrategy {
    name: 'route_redirection';
    longContextCategory: string;
    tokenThreshold: number;
}
export interface LongContextCompressionConfig extends MaxTokenStrategy {
    name: 'long_context_compression';
    compressionRatio: number;
    compressionModel: string;
}
export interface MaxTokenPreprocessorConfig {
    enabled: boolean;
    strategies: {
        dynamicTruncation: DynamicTruncationConfig;
        routeRedirection: RouteRedirectionConfig;
        longContextCompression: LongContextCompressionConfig;
    };
}
export interface Message {
    role: string;
    content: string;
    tool_calls?: any[];
    tool_call_id?: string;
}
export interface PreprocessedRequest {
    messages: Message[];
    model: string;
    max_tokens?: number;
    temperature?: number;
    tools?: any[];
    redirectedCategory?: string;
    appliedStrategies: string[];
    originalTokenCount: number;
    processedTokenCount: number;
}
export interface TokenAnalysis {
    totalTokens: number;
    systemTokens: number;
    userTokens: number;
    assistantTokens: number;
    toolTokens: number;
    exceedsLimit: boolean;
    exceedsBy: number;
}
export declare class MaxTokenPreprocessor {
    private config;
    constructor(config: MaxTokenPreprocessorConfig);
    /**
     * 主要预处理入口
     */
    preprocessRequest(request: any, maxTokenLimit: number, routingCategory?: string): Promise<PreprocessedRequest>;
    /**
     * 分析请求的token使用情况
     */
    private analyzeTokens;
    /**
     * 简化的token估算 (实际应该使用准确的tokenizer)
     */
    private estimateTokens;
    /**
     * 按优先级获取启用的策略
     */
    private getEnabledStrategiesByPriority;
    /**
     * 应用路由重定向策略
     */
    private applyRouteRedirection;
    /**
     * 应用动态截断策略
     */
    private applyDynamicTruncation;
    /**
     * 计算消息列表的总token数
     */
    private calculateMessagesTokens;
    /**
     * 简化工具定义
     */
    private simplifyToolDefinitions;
    /**
     * 应用长上下文压缩策略
     */
    private applyLongContextCompression;
    /**
     * 压缩消息 (占位符实现)
     */
    private compressMessages;
    /**
     * 获取默认配置
     */
    static getDefaultConfig(): MaxTokenPreprocessorConfig;
}
//# sourceMappingURL=max-token-preprocessor.d.ts.map