/**
 * Tokenizer-Aware Encoding - Optimize for LLM tokenizers
 *
 * Adjusts encoding based on how LLM tokenizers (GPT, Claude, etc.) process text.
 * Minimizes token count by choosing optimal delimiters, spacing, and formatting.
 *
 * Features:
 * - Tokenizer-friendly delimiter selection
 * - Optimal whitespace usage
 * - Case-sensitive optimization
 * - Number formatting optimization
 *
 * Example optimizations:
 * - Use spaces instead of tabs (1 token vs 1-2 tokens)
 * - Prefer common delimiters (comma over pipe)
 * - Avoid unnecessary quotes
 * - Use compact number formats
 *
 * Token Savings: 5-15% through tokenizer-aware choices
 */
import type { TokenizerAwareOptions } from './types.js';
/**
 * Tokenizer analysis result
 */
export interface TokenizerAnalysis {
    estimatedTokens: number;
    recommendedDelimiter: string;
    recommendedQuoting: 'minimal' | 'conservative';
    optimizations: string[];
}
/**
 * Tokenizer-aware encoding manager
 */
export declare class TokenizerAware {
    private options;
    constructor(options?: Partial<TokenizerAwareOptions>);
    /**
     * Analyze text for tokenizer optimization
     *
     * @param text - Text to analyze
     * @returns Analysis result
     */
    analyzeText(text: string): TokenizerAnalysis;
    /**
     * Estimate token count for text
     *
     * Simple estimation: ~4 characters per token on average
     *
     * @param text - Text to estimate
     * @returns Estimated token count
     */
    estimateTokens(text: string): number;
    /**
     * Recommend best delimiter for data
     *
     * @param sample - Sample data
     * @returns Recommended delimiter
     */
    recommendDelimiter(sample: string): string;
    /**
     * Optimize text for tokenizer
     *
     * @param text - Original text
     * @returns Optimized text
     */
    optimize(text: string): string;
    /**
     * Compact number formatting
     *
     * @param text - Text with numbers
     * @returns Text with compacted numbers
     */
    private compactNumbers;
    /**
     * Generate tokenizer hint directive
     *
     * Format: @tokenizer target
     *
     * @param target - Target tokenizer
     * @returns TONL directive
     */
    generateDirective(target: string): string;
    /**
     * Parse tokenizer directive
     *
     * @param directive - Directive string
     * @returns Target tokenizer
     */
    parseDirective(directive: string): string;
    /**
     * Estimate savings from optimization
     *
     * @param original - Original text
     * @returns Estimated token savings
     */
    estimateSavings(original: string): number;
    /**
     * Check if optimization would be beneficial
     *
     * @param text - Text to check
     * @returns True if recommended
     */
    shouldOptimize(text: string): boolean;
}
//# sourceMappingURL=tokenizer-aware.d.ts.map