autosnippet
Version:
Extract code patterns into a knowledge base for AI coding assistants
82 lines (81 loc) • 2.6 kB
TypeScript
/**
* dimension-configs.js — v3.0 维度配置 + Tier Reflection
*
* **v2: 从统一维度注册表 (DimensionRegistry) 派生**
*
* 从 orchestrator.js 拆分，包含:
* - DIMENSION_CONFIGS_V3: 维度的 outputType + allowedKnowledgeTypes
* 现在从 DimensionRegistry 自动生成，无需手动维护
* - getFullDimensionConfig(): 合并 baseDimensions + V3 专属配置
* - buildTierReflection: Tier 级反思聚合 (规则化, 不需要 AI)
*
* @module pipeline/dimension-configs
*/
/**
* 从统一注册表生成 V3 配置映射
* 所有维度统一为 candidate-only outputType
*/
export declare const DIMENSION_CONFIGS_V3: Record<string, {
outputType: string;
allowedKnowledgeTypes: string[];
}>;
/**
* 获取完整维度配置（合并 baseDimensions + V3 专属配置 + SOP）
*
* @param dimId 维度 ID
* @returns 完整维度配置，或 null（未知维度）
*/
export declare function getFullDimensionConfig(dimId: string): {
id: string;
label: string;
guide: string;
outputType: string;
allowedKnowledgeTypes: string[];
skillWorthy: boolean;
dualOutput: boolean;
knowledgeTypes: string[];
sopSteps: import("#domain/dimension/DimensionSop.js").FullSopStep[] | null;
commonMistakes: string[];
timeEstimate: string | null;
focusKeywords: string[];
} | null;
/** A single finding from a dimension analysis */
interface DimensionFinding {
dimId?: string;
importance?: number;
evidence?: string | unknown;
finding?: string;
}
/** Minimal session store interface for tier reflection */
interface TierSessionStore {
getDimensionReport(dimId: string): {
findings?: DimensionFinding[];
digest?: {
gaps?: string[];
remainingTasks?: Array<{
signal?: string;
reason?: string;
}>;
};
} | undefined;
}
/**
* 构建 Tier 级 Reflection — 在每个 Tier 完成后调用
*
* 无需 AI 调用，通过规则化聚合维度发现:
* - 收集所有维度的关键发现并按重要性排序
* - 检测跨维度重复模式
* - 为下一 Tier 生成建议
*
* @param tierIndex Tier 索引 (0-based)
* @param tierResults 本 Tier 的维度结果
* @returns TierReflection
*/
export declare function buildTierReflection(tierIndex: number, tierResults: Map<string, unknown>, sessionStore: TierSessionStore): {
tierIndex: number;
completedDimensions: string[];
topFindings: DimensionFinding[];
crossDimensionPatterns: string[];
suggestionsForNextTier: string[];
};
export {};