@thinking-models/mcp-server
Version:
A Model Context Protocol (MCP) server for thinking models
55 lines (54 loc) • 1.94 kB
JavaScript
/**
* 计算文本相似度的工具函数
*/
import { log } from './utils.js';
/**
* 计算两段文本之间的语义相似度
* 移除了OpenRouter依赖，仅使用本地算法计算
*/
export async function calculateSemanticSimilarity(text1, text2) {
try {
// 基本验证
if (!text1 || !text2 || text1.trim() === "" || text2.trim() === "") {
return 0;
}
// 规范化文本
const normalizeText = (text) => {
return text.replace(/\s+/g, ' ').trim().toLowerCase();
};
const normalizedText1 = normalizeText(text1);
const normalizedText2 = normalizeText(text2);
// 完全相同文本直接返回1.0
if (normalizedText1 === normalizedText2) {
return 1.0;
}
// 使用Jaccard相似度
return calculateJaccardSimilarity(normalizedText1, normalizedText2);
}
catch (error) {
log(`相似度计算过程中发生错误: ${error instanceof Error ? error.message : String(error)}`);
return 0;
}
}
/**
* 计算两个文本的Jaccard相似度
*/
function calculateJaccardSimilarity(text1, text2) {
try {
// 使用字词分割
const words1 = new Set(text1.split(/\s+/).filter(w => w.length > 1));
const words2 = new Set(text2.split(/\s+/).filter(w => w.length > 1));
if (words1.size === 0 || words2.size === 0)
return 0;
const intersection = new Set([...words1].filter(x => words2.has(x)));
const union = new Set([...words1, ...words2]);
if (union.size === 0)
return 0;
const jaccardScore = intersection.size / union.size;
return Math.min(0.8, jaccardScore * 1.5); // 稍微提高分数但限制最大值，模拟语义相似度
}
catch (e) {
log(`计算Jaccard相似度时出错: ${e instanceof Error ? e.message : String(e)}`);
return 0;
}
}