stellavault 0.2.0 → 0.3.0

package/packages/core/src/api/graph-data.ts

@@ -1,286 +0,0 @@
-// Design Ref: §4.2 — Graph Data 생성 알고리즘
-// Design Ref: §6.2 — K-Means 클러스터링
-
-import type { VectorStore } from '../store/types.js';
-import type { GraphNode, GraphEdge, Cluster, GraphData } from '../types/graph.js';
-import { createHash } from 'node:crypto';
-
-const CLUSTER_COLORS = [
-  '#6366f1', '#ec4899', '#f59e0b', '#10b981', '#3b82f6',
-  '#8b5cf6', '#ef4444', '#14b8a6', '#f97316', '#06b6d4',
-  '#84cc16', '#e879f9', '#22d3ee', '#a3e635', '#fb923c',
-];
-
-export type GraphMode = 'semantic' | 'folder';
-
-export interface BuildGraphOptions {
-  mode?: GraphMode;            // default: 'semantic'
-  edgeThreshold?: number;      // default: 0.15
-  maxEdgesPerNode?: number;    // default: 5
-  clusterCount?: number;       // 0 = auto (Elbow)
-}
-
-export async function buildGraphData(
-  store: VectorStore,
-  options: BuildGraphOptions = {},
-): Promise<GraphData> {
-  const {
-    edgeThreshold = 0.15,
-    maxEdgesPerNode = 5,
-  } = options;
-
-  // 1. 모든 문서 + 임베딩 로드
-  const [docs, embeddings] = await Promise.all([
-    store.getAllDocuments(),
-    store.getDocumentEmbeddings(),
-  ]);
-
-  // 2. k-NN으로 엣지 생성
-  const edges: GraphEdge[] = [];
-  const edgeCounts = new Map<string, number>();
-
-  for (const doc of docs) {
-    const vec = embeddings.get(doc.id);
-    if (!vec) continue;
-
-    // 모든 다른 문서와 코사인 유사도 계산
-    const similarities: Array<{ id: string; sim: number }> = [];
-    for (const other of docs) {
-      if (other.id === doc.id) continue;
-      const otherVec = embeddings.get(other.id);
-      if (!otherVec) continue;
-      const sim = cosineSimilarity(vec, otherVec);
-      if (sim >= edgeThreshold) {
-        similarities.push({ id: other.id, sim });
-      }
-    }
-
-    // 상위 k개만
-    similarities.sort((a, b) => b.sim - a.sim);
-    const topK = similarities.slice(0, maxEdgesPerNode);
-
-    for (const { id: targetId, sim } of topK) {
-      // 중복 엣지 방지 (A→B, B→A)
-      const edgeKey = [doc.id, targetId].sort().join(':');
-      if (!edgeCounts.has(edgeKey)) {
-        edges.push({ source: doc.id, target: targetId, weight: sim });
-        edgeCounts.set(edgeKey, 1);
-      }
-    }
-  }
-
-  // 3. 클러스터링 — 모드에 따라 분기
-  const mode = options.mode ?? 'semantic';
-  let assignmentMap: Map<string, number>;
-  let clusters: Cluster[];
-
-  if (mode === 'folder') {
-    // 폴더 기반: 최상위 폴더를 클러스터로 사용
-    const folderMap = new Map<string, number>();
-    const folderNames: string[] = [];
-
-    for (const doc of docs) {
-      const topFolder = doc.filePath.split('/')[0] ?? 'root';
-      if (!folderMap.has(topFolder)) {
-        folderMap.set(topFolder, folderNames.length);
-        folderNames.push(topFolder);
-      }
-    }
-
-    assignmentMap = new Map<string, number>();
-    for (const doc of docs) {
-      const topFolder = doc.filePath.split('/')[0] ?? 'root';
-      assignmentMap.set(doc.id, folderMap.get(topFolder)!);
-    }
-
-    // 폴더별 클러스터
-    const folderCounts = new Map<number, number>();
-    for (const [, cId] of assignmentMap) {
-      folderCounts.set(cId, (folderCounts.get(cId) ?? 0) + 1);
-    }
-
-    clusters = folderNames.map((name, i) => ({
-      id: i,
-      label: name.replace(/^\d+_/, ''),  // "04_Projects" → "Projects"
-      color: CLUSTER_COLORS[i % CLUSTER_COLORS.length],
-      nodeCount: folderCounts.get(i) ?? 0,
-    }));
-  } else {
-    // 시맨틱 기반: K-means
-    const docIds = docs.filter(d => embeddings.has(d.id)).map(d => d.id);
-    const vectors = docIds.map(id => embeddings.get(id)!);
-    const k = Math.min(Math.max(5, Math.round(Math.sqrt(docIds.length / 5))), 10);
-    const assignments = kMeans(vectors, k);
-
-    // 클러스터별 문서 수집 (id + title)
-    const clusterDocInfos = new Map<number, Array<{ id: string; title: string }>>();
-    for (let i = 0; i < docIds.length; i++) {
-      const cId = assignments[i];
-      if (!clusterDocInfos.has(cId)) clusterDocInfos.set(cId, []);
-      const doc = docs.find(d => d.id === docIds[i]);
-      if (doc) clusterDocInfos.get(cId)!.push({ id: doc.id, title: doc.title });
-    }
-
-    // 엣지 기반 연결 수 계산 (대표 노드 선정용)
-    const connCounts = new Map<string, number>();
-    for (const e of edges) {
-      connCounts.set(e.source, (connCounts.get(e.source) ?? 0) + 1);
-      connCounts.set(e.target, (connCounts.get(e.target) ?? 0) + 1);
-    }
-
-    clusters = [];
-    for (const [cId, docInfos] of clusterDocInfos) {
-      // 가장 연결이 많은 노드 = 대표 노드
-      const sorted = [...docInfos].sort((a, b) =>
-        (connCounts.get(b.id) ?? 0) - (connCounts.get(a.id) ?? 0)
-      );
-      const representative = sorted[0]?.title ?? 'Cluster';
-      clusters.push({
-        id: cId,
-        label: `${representative} (${docInfos.length})`,
-        color: CLUSTER_COLORS[cId % CLUSTER_COLORS.length],
-        nodeCount: docInfos.length,
-      });
-    }
-
-    assignmentMap = new Map<string, number>();
-    for (let i = 0; i < docIds.length; i++) {
-      assignmentMap.set(docIds[i], assignments[i]);
-    }
-  }
-
-  // 4. 노드 생성
-
-  const connectionCounts = new Map<string, number>();
-  for (const edge of edges) {
-    connectionCounts.set(edge.source, (connectionCounts.get(edge.source) ?? 0) + 1);
-    connectionCounts.set(edge.target, (connectionCounts.get(edge.target) ?? 0) + 1);
-  }
-  const maxConnections = Math.max(1, ...connectionCounts.values());
-
-  const nodes: GraphNode[] = docs.map(doc => {
-    const conns = connectionCounts.get(doc.id) ?? 0;
-    const ratio = conns / maxConnections; // 0~1
-    // 지수 스케일: 상위 노드만 극적으로 커짐 (ratio^0.5 → 중심부 강조)
-    const size = 1 + 6 * Math.pow(ratio, 0.5);
-    return {
-      id: doc.id,
-      label: doc.title.slice(0, 40),
-      filePath: doc.filePath,
-      tags: doc.tags,
-      clusterId: assignmentMap.get(doc.id) ?? 0,
-      size,
-      source: doc.source ?? 'local',
-      type: doc.type ?? 'note',
-      lastModified: doc.lastModified,
-    };
-  });
-
-  // 5. cache key
-  const cacheKey = createHash('md5')
-    .update(JSON.stringify({ nodeCount: nodes.length, edgeCount: edges.length }))
-    .digest('hex')
-    .slice(0, 8);
-
-  return {
-    nodes,
-    edges,
-    clusters,
-    stats: {
-      nodeCount: nodes.length,
-      edgeCount: edges.length,
-      clusterCount: clusters.length,
-    },
-  };
-}
-
-// --- 유틸리티 ---
-
-function cosineSimilarity(a: number[], b: number[]): number {
-  let dot = 0, normA = 0, normB = 0;
-  for (let i = 0; i < a.length; i++) {
-    dot += a[i] * b[i];
-    normA += a[i] * a[i];
-    normB += b[i] * b[i];
-  }
-  const denom = Math.sqrt(normA) * Math.sqrt(normB);
-  return denom === 0 ? 0 : dot / denom;
-}
-
-function kMeans(vectors: number[][], k: number, maxIter: number = 50): number[] {
-  if (vectors.length === 0) return [];
-  const dims = vectors[0].length;
-
-  // K-means++ 초기화
-  const centroids: number[][] = [vectors[Math.floor(Math.random() * vectors.length)].slice()];
-  for (let c = 1; c < k; c++) {
-    const dists = vectors.map(v => {
-      let minD = Infinity;
-      for (const cent of centroids) {
-        const d = euclideanDist(v, cent);
-        if (d < minD) minD = d;
-      }
-      return minD;
-    });
-    const totalDist = dists.reduce((a, b) => a + b, 0);
-    let r = Math.random() * totalDist;
-    for (let i = 0; i < dists.length; i++) {
-      r -= dists[i];
-      if (r <= 0) { centroids.push(vectors[i].slice()); break; }
-    }
-    if (centroids.length <= c) centroids.push(vectors[Math.floor(Math.random() * vectors.length)].slice());
-  }
-
-  const assignments = new Array<number>(vectors.length).fill(0);
-
-  for (let iter = 0; iter < maxIter; iter++) {
-    let changed = false;
-
-    // Assign
-    for (let i = 0; i < vectors.length; i++) {
-      let bestC = 0, bestD = Infinity;
-      for (let c = 0; c < k; c++) {
-        const d = euclideanDist(vectors[i], centroids[c]);
-        if (d < bestD) { bestD = d; bestC = c; }
-      }
-      if (assignments[i] !== bestC) { assignments[i] = bestC; changed = true; }
-    }
-
-    if (!changed) break;
-
-    // Update centroids
-    for (let c = 0; c < k; c++) {
-      const members = vectors.filter((_, i) => assignments[i] === c);
-      if (members.length === 0) continue;
-      for (let d = 0; d < dims; d++) {
-        centroids[c][d] = members.reduce((sum, v) => sum + v[d], 0) / members.length;
-      }
-    }
-  }
-
-  return assignments;
-}
-
-function euclideanDist(a: number[], b: number[]): number {
-  let sum = 0;
-  for (let i = 0; i < a.length; i++) {
-    const diff = a[i] - b[i];
-    sum += diff * diff;
-  }
-  return Math.sqrt(sum);
-}
-
-function extractClusterLabel(titles: string[]): string {
-  // 제목에서 빈출 단어 추출 (2글자 이상)
-  const wordCounts = new Map<string, number>();
-  const stopWords = new Set(['the', 'a', 'an', 'in', 'on', 'at', 'to', 'for', 'of', 'and', 'or', 'is', 'are', 'was', 'be']);
-  for (const title of titles) {
-    const words = title.replace(/[^\w가-힣\s]/g, '').split(/\s+/).filter(w => w.length >= 2 && !stopWords.has(w.toLowerCase()));
-    const seen = new Set<string>();
-    for (const w of words) {
-      if (!seen.has(w)) { wordCounts.set(w, (wordCounts.get(w) ?? 0) + 1); seen.add(w); }
-    }
-  }
-  const sorted = [...wordCounts.entries()].sort((a, b) => b[1] - a[1]);
-  return sorted.slice(0, 3).map(([w]) => w).join(', ') || `Cluster`;
-}