autosnippet
Version:
Extract code patterns into a knowledge base for AI coding assistants
85 lines (84 loc) • 2.5 kB
TypeScript
/**
* QualityScorer v2 — Recipe 质量评分器
*
* 面向知识管理场景重新设计，采用渐进式评分（非二元判断），
* 充分利用 KnowledgeEntry 所有可用字段。
*
* 5 维度加权:
* - completeness (0.25): 结构完整性 — 核心字段齐全度
* - contentDepth (0.30): 内容深度 — markdown 丰富度、推理、溯源
* - deliveryReady (0.20): 交付就绪 — trigger/language/tags/category
* - actionability (0.15): 可操作性 — coreCode、do/dont/when 质量
* - provenance (0.10): 溯源可信 — confidence、sources、authority
*
* 设计参考:
* - RAG Triad (TruLens): Relevance + Groundedness + Answer Relevance
* - RAGAS: Context Precision + Faithfulness + Factual Correctness
* - SonarQube: 多维度渐进评级，非二元判断
*/
export interface RecipeInput {
title?: string;
trigger?: string;
description?: string;
language?: string;
category?: string;
doClause?: string;
dontClause?: string;
whenClause?: string;
coreCode?: string;
usageGuide?: string;
contentMarkdown?: string;
contentRationale?: string;
reasoningWhyStandard?: string;
reasoningSources?: string[];
reasoningConfidence?: number;
source?: string;
headers?: string[];
tags?: string[];
views?: number;
clicks?: number;
rating?: number;
[key: string]: unknown;
}
export declare class QualityScorer {
#private;
constructor(options?: {
weights?: Record<string, number>;
});
/**
* 计算综合质量分
* @returns { score: 0-1, dimensions: Record<string,number>, grade: A-F }
*/
score(recipe: RecipeInput): {
score: number;
dimensions: {
completeness: number;
contentDepth: number;
deliveryReady: number;
actionability: number;
provenance: number;
};
grade: string;
};
/** 批量评分 */
scoreBatch(recipes: RecipeInput[]): {
score: number;
dimensions: {
completeness: number;
contentDepth: number;
deliveryReady: number;
actionability: number;
provenance: number;
};
grade: string;
recipe: RecipeInput;
}[];
/** 获取维度权重 */
getWeights(): {
completeness: 0.25;
contentDepth: 0.3;
deliveryReady: 0.2;
actionability: 0.15;
provenance: 0.1;
};
}