@thinking-models/mcp-server
Version:
A Model Context Protocol (MCP) server for thinking models
63 lines (62 loc) • 2.66 kB
JavaScript
/**
* 统一的推荐系统 - 整合不同的推荐策略和算法
*/
import { log } from './utils.js';
import { calculateModelSimilarity, DEFAULT_BASIC_WEIGHTS, DEFAULT_ENHANCED_WEIGHTS } from './similarity_engine.js';
/** 推荐系统模式 */
export var RecommendationMode;
(function (RecommendationMode) {
RecommendationMode["ENHANCED"] = "enhanced";
RecommendationMode["BASIC"] = "basic"; // 基础模式（轻量快速的相似度评估）
})(RecommendationMode || (RecommendationMode = {}));
/**
* 获取模型推荐
* @param sourceModel 源模型
* @param candidateModels 候选模型数组
* @param lang 语言
* @param mode 推荐模式
* @param limit 返回的最大结果数量
* @returns 推荐结果数组
*/
export async function getModelRecommendations(sourceModel, candidateModels, lang, mode = RecommendationMode.ENHANCED, limit = 5) {
try {
log(`开始为「${sourceModel.name}」生成${mode === RecommendationMode.ENHANCED ? '增强' : '基础'}模式的推荐`);
// 根据模式选择合适的权重配置
const weights = mode === RecommendationMode.ENHANCED
? DEFAULT_ENHANCED_WEIGHTS
: DEFAULT_BASIC_WEIGHTS;
// 根据模式决定是否使用语义相似度
const useSemanticSimilarity = mode === RecommendationMode.ENHANCED;
// 计算每个候选模型的相似度
const scoredModelsPromises = candidateModels.map(async (candidate) => {
const { score, reasons } = await calculateModelSimilarity({
sourceModel,
targetModel: candidate,
weights,
useSemanticSimilarity
});
if (score <= 0)
return null;
return {
id: candidate.id,
name: candidate.name,
definition: candidate.definition || "",
relevance_score: score,
match_reasons: reasons
};
});
// 等待所有相似度计算完成
const scoredModelsWithNulls = await Promise.all(scoredModelsPromises);
// 过滤掉null结果并按得分排序
const recommendations = scoredModelsWithNulls
.filter((result) => result !== null)
.sort((a, b) => b.relevance_score - a.relevance_score)
.slice(0, limit);
log(`${mode}模式推荐系统生成了 ${recommendations.length} 个相关模型推荐`);
return recommendations;
}
catch (error) {
log(`生成模型推荐时出错: ${error instanceof Error ? error.message : String(error)}`);
return [];
}
}