// ═══════════════════════════════════════════════════════════════════════════
// Story Summary - Recall Engine (v5 - 统一命名)
//
// 命名规范：
// - 存储层用 L0/L1/L2/L3（StateAtom/Chunk/Event/Fact）
// - 召回层用语义名称：anchor/evidence/event/constraint
// ═══════════════════════════════════════════════════════════════════════════

import { getAllEventVectors, getChunksByFloors, getMeta, getChunkVectorsByIds } from '../storage/chunk-store.js';
import { getAllStateVectors, getStateAtoms } from '../storage/state-store.js';
import { getEngineFingerprint, embed } from '../utils/embedder.js';
import { xbLog } from '../../../../core/debug-core.js';
import { getContext } from '../../../../../../../extensions.js';
import { filterText } from '../utils/text-filter.js';
import { expandQueryCached, buildSearchText } from '../llm/query-expansion.js';
import { rerankChunks } from '../llm/reranker.js';
import { createMetrics, calcSimilarityStats } from './metrics.js';

const MODULE_ID = 'recall';

// ═══════════════════════════════════════════════════════════════════════════
// 配置
// ═══════════════════════════════════════════════════════════════════════════

const CONFIG = {
    // Query Expansion
    QUERY_EXPANSION_TIMEOUT: 6000,

    // Anchor (L0 StateAtoms) 配置
    ANCHOR_MIN_SIMILARITY: 0.58,

    // Evidence (L1 Chunks) 粗筛配置
    EVIDENCE_COARSE_MAX: 100,

    // Event (L2 Events) 配置
    EVENT_CANDIDATE_MAX: 100,
    EVENT_SELECT_MAX: 50,
    EVENT_MIN_SIMILARITY: 0.55,
    EVENT_MMR_LAMBDA: 0.72,

    // Rerank 配置
    RERANK_THRESHOLD: 80,
    RERANK_TOP_N: 50,
    RERANK_MIN_SCORE: 0.15,

    // 因果链
    CAUSAL_CHAIN_MAX_DEPTH: 10,
    CAUSAL_INJECT_MAX: 30,
};

// ═══════════════════════════════════════════════════════════════════════════
// 工具函数
// ═══════════════════════════════════════════════════════════════════════════

/**
 * 计算余弦相似度
 * @param {number[]} a - 向量A
 * @param {number[]} b - 向量B
 * @returns {number} 相似度 [0, 1]
 */
function cosineSimilarity(a, b) {
    if (!a?.length || !b?.length || a.length !== b.length) return 0;
    let dot = 0, nA = 0, nB = 0;
    for (let i = 0; i < a.length; i++) {
        dot += a[i] * b[i];
        nA += a[i] * a[i];
        nB += b[i] * b[i];
    }
    return nA && nB ? dot / (Math.sqrt(nA) * Math.sqrt(nB)) : 0;
}

/**
 * 标准化字符串（用于实体匹配）
 * @param {string} s - 输入字符串
 * @returns {string} 标准化后的字符串
 */
function normalize(s) {
    return String(s || '')
        .normalize('NFKC')
        .replace(/[\u200B-\u200D\uFEFF]/g, '')
        .trim()
        .toLowerCase();
}

/**
 * 清理文本用于召回
 * @param {string} text - 原始文本
 * @returns {string} 清理后的文本
 */
function cleanForRecall(text) {
    return filterText(text).replace(/\[tts:[^\]]*\]/gi, '').trim();
}

/**
 * 从 focus entities 中移除用户名
 * @param {string[]} focusEntities - 焦点实体列表
 * @param {string} userName - 用户名
 * @returns {string[]} 过滤后的实体列表
 */
function removeUserNameFromFocus(focusEntities, userName) {
    const u = normalize(userName);
    if (!u) return Array.isArray(focusEntities) ? focusEntities : [];

    return (focusEntities || [])
        .map(e => String(e || '').trim())
        .filter(Boolean)
        .filter(e => normalize(e) !== u);
}

/**
 * 构建 rerank 查询文本
 * @param {object} expansion - query expansion 结果
 * @param {object[]} lastMessages - 最近消息
 * @param {string} pendingUserMessage - 待发送的用户消息
 * @returns {string} 查询文本
 */
function buildRerankQuery(expansion, lastMessages, pendingUserMessage) {
    const parts = [];

    if (expansion?.focus?.length) {
        parts.push(expansion.focus.join(' '));
    }

    if (expansion?.queries?.length) {
        parts.push(...expansion.queries.slice(0, 3));
    }

    const recentTexts = (lastMessages || [])
        .slice(-2)
        .map(m => cleanForRecall(m.mes || '').slice(0, 150))
        .filter(Boolean);

    if (recentTexts.length) {
        parts.push(...recentTexts);
    }

    if (pendingUserMessage) {
        parts.push(cleanForRecall(pendingUserMessage).slice(0, 200));
    }

    return parts.filter(Boolean).join('\n').slice(0, 1500);
}

// ═══════════════════════════════════════════════════════════════════════════
// MMR 选择算法
// ═══════════════════════════════════════════════════════════════════════════

/**
 * Maximal Marginal Relevance 选择
 * @param {object[]} candidates - 候选项
 * @param {number} k - 选择数量
 * @param {number} lambda - 相关性/多样性权衡参数
 * @param {Function} getVector - 获取向量的函数
 * @param {Function} getScore - 获取分数的函数
 * @returns {object[]} 选中的候选项
 */
function mmrSelect(candidates, k, lambda, getVector, getScore) {
    const selected = [];
    const ids = new Set();

    while (selected.length < k && candidates.length) {
        let best = null;
        let bestScore = -Infinity;

        for (const c of candidates) {
            if (ids.has(c._id)) continue;

            const rel = getScore(c);
            let div = 0;

            if (selected.length) {
                const vC = getVector(c);
                if (vC?.length) {
                    for (const s of selected) {
                        const sim = cosineSimilarity(vC, getVector(s));
                        if (sim > div) div = sim;
                    }
                }
            }

            const score = lambda * rel - (1 - lambda) * div;
            if (score > bestScore) {
                bestScore = score;
                best = c;
            }
        }

        if (!best) break;
        selected.push(best);
        ids.add(best._id);
    }

    return selected;
}

// ═══════════════════════════════════════════════════════════════════════════
// [Anchors] L0 StateAtoms 检索
// ═══════════════════════════════════════════════════════════════════════════

/**
 * 检索语义锚点（L0 StateAtoms）
 * @param {number[]} queryVector - 查询向量
 * @param {object} vectorConfig - 向量配置
 * @param {object} metrics - 指标对象
 * @returns {Promise<{hits: object[], floors: Set<number>}>}
 */
async function recallAnchors(queryVector, vectorConfig, metrics) {
    const { chatId } = getContext();
    if (!chatId || !queryVector?.length) {
        return { hits: [], floors: new Set() };
    }

    const meta = await getMeta(chatId);
    const fp = getEngineFingerprint(vectorConfig);
    if (meta.fingerprint && meta.fingerprint !== fp) {
        xbLog.warn(MODULE_ID, 'Anchor fingerprint 不匹配');
        return { hits: [], floors: new Set() };
    }

    const stateVectors = await getAllStateVectors(chatId);
    if (!stateVectors.length) {
        return { hits: [], floors: new Set() };
    }

    const atomsList = getStateAtoms();
    const atomMap = new Map(atomsList.map(a => [a.atomId, a]));

    // 按阈值过滤，不设硬上限
    const scored = stateVectors
        .map(sv => {
            const atom = atomMap.get(sv.atomId);
            if (!atom) return null;

            return {
                atomId: sv.atomId,
                floor: sv.floor,
                similarity: cosineSimilarity(queryVector, sv.vector),
                atom,
            };
        })
        .filter(Boolean)
        .filter(s => s.similarity >= CONFIG.ANCHOR_MIN_SIMILARITY)
        .sort((a, b) => b.similarity - a.similarity);

    const floors = new Set(scored.map(s => s.floor));

    if (metrics) {
        metrics.anchor.matched = scored.length;
        metrics.anchor.floorsHit = floors.size;
        metrics.anchor.topHits = scored.slice(0, 5).map(s => ({
            floor: s.floor,
            semantic: s.atom?.semantic?.slice(0, 50),
            similarity: Math.round(s.similarity * 1000) / 1000,
        }));
    }

    return { hits: scored, floors };
}

// ═══════════════════════════════════════════════════════════════════════════
// [Evidence] L1 Chunks 拉取 + 粗筛 + Rerank
// ═══════════════════════════════════════════════════════════════════════════

/**
 * 统计 evidence 类型构成
 * @param {object[]} chunks - chunk 列表
 * @returns {{anchorVirtual: number, chunkReal: number}}
 */
function countEvidenceByType(chunks) {
    let anchorVirtual = 0;
    let chunkReal = 0;

    for (const c of chunks || []) {
        if (c.isAnchorVirtual) {
            anchorVirtual++;
        } else {
            chunkReal++;
        }
    }

    return { anchorVirtual, chunkReal };
}

/**
 * 根据锚点命中楼层拉取证据（L1 Chunks）
 * @param {Set<number>} anchorFloors - 锚点命中的楼层
 * @param {object[]} anchorHits - 锚点命中结果
 * @param {number[]} queryVector - 查询向量
 * @param {string} queryText - rerank 查询文本
 * @param {object} metrics - 指标对象
 * @returns {Promise<object[]>} 证据 chunks
 */
async function pullEvidenceByFloors(anchorFloors, anchorHits, queryVector, queryText, metrics) {
    const { chatId } = getContext();
    if (!chatId || !anchorFloors.size) {
        return [];
    }

    const floorArray = Array.from(anchorFloors);

    // 1. 构建锚点虚拟 chunks（来自 L0 StateAtoms）
    const anchorVirtualChunks = (anchorHits || []).map(a => ({
        chunkId: `anchor-${a.atomId}`,
        floor: a.floor,
        chunkIdx: -1,
        speaker: '📌',
        isUser: false,
        text: a.atom?.semantic || '',
        similarity: a.similarity,
        isAnchorVirtual: true,
        _atom: a.atom,
    }));

    // 2. 拉取真实 chunks（来自 L1）
    let dbChunks = [];
    try {
        dbChunks = await getChunksByFloors(chatId, floorArray);
    } catch (e) {
        xbLog.warn(MODULE_ID, '从 DB 拉取 chunks 失败', e);
    }

    // 3. L1 向量粗筛
    let coarseFiltered = [];
    if (dbChunks.length > 0 && queryVector?.length) {
        const chunkIds = dbChunks.map(c => c.chunkId);
        let chunkVectors = [];
        try {
            chunkVectors = await getChunkVectorsByIds(chatId, chunkIds);
        } catch (e) {
            xbLog.warn(MODULE_ID, 'L1 向量获取失败', e);
        }

        const vectorMap = new Map(chunkVectors.map(v => [v.chunkId, v.vector]));

        coarseFiltered = dbChunks
            .map(c => {
                const vec = vectorMap.get(c.chunkId);
                if (!vec?.length) return null;

                return {
                    ...c,
                    isAnchorVirtual: false,
                    similarity: cosineSimilarity(queryVector, vec),
                };
            })
            .filter(Boolean)
            .sort((a, b) => b.similarity - a.similarity)
            .slice(0, CONFIG.EVIDENCE_COARSE_MAX);
    }

    // 4. 合并
    const allEvidence = [...anchorVirtualChunks, ...coarseFiltered];

    // 更新 metrics
    if (metrics) {
        metrics.evidence.floorsFromAnchors = floorArray.length;
        metrics.evidence.chunkTotal = dbChunks.length;
        metrics.evidence.chunkAfterCoarse = coarseFiltered.length;
        metrics.evidence.merged = allEvidence.length;
        metrics.evidence.mergedByType = countEvidenceByType(allEvidence);
    }

    // 5. 是否需要 Rerank
    if (allEvidence.length <= CONFIG.RERANK_THRESHOLD) {
        if (metrics) {
            metrics.evidence.rerankApplied = false;
            metrics.evidence.selected = allEvidence.length;
            metrics.evidence.selectedByType = countEvidenceByType(allEvidence);
        }
        return allEvidence;
    }

    // 6. Rerank 精排
    const T_Rerank_Start = performance.now();

    const reranked = await rerankChunks(queryText, allEvidence, {
        topN: CONFIG.RERANK_TOP_N,
        minScore: CONFIG.RERANK_MIN_SCORE,
    });

    const rerankTime = Math.round(performance.now() - T_Rerank_Start);

    if (metrics) {
        metrics.evidence.rerankApplied = true;
        metrics.evidence.beforeRerank = allEvidence.length;
        metrics.evidence.afterRerank = reranked.length;
        metrics.evidence.selected = reranked.length;
        metrics.evidence.selectedByType = countEvidenceByType(reranked);
        metrics.evidence.rerankTime = rerankTime;
        metrics.timing.evidenceRerank = rerankTime;

        const scores = reranked.map(c => c._rerankScore || 0).filter(s => s > 0);
        if (scores.length > 0) {
            scores.sort((a, b) => a - b);
            metrics.evidence.rerankScores = {
                min: Number(scores[0].toFixed(3)),
                max: Number(scores[scores.length - 1].toFixed(3)),
                mean: Number((scores.reduce((a, b) => a + b, 0) / scores.length).toFixed(3)),
            };
        }
    }

    xbLog.info(MODULE_ID, `Evidence: ${dbChunks.length} L1 → ${coarseFiltered.length} coarse → ${reranked.length} rerank (${rerankTime}ms)`);

    return reranked;
}

// ═══════════════════════════════════════════════════════════════════════════
// [Events] L2 Events 检索
// ═══════════════════════════════════════════════════════════════════════════

/**
 * 检索事件（L2 Events）
 * @param {number[]} queryVector - 查询向量
 * @param {object[]} allEvents - 所有事件
 * @param {object} vectorConfig - 向量配置
 * @param {string[]} focusEntities - 焦点实体
 * @param {object} metrics - 指标对象
 * @returns {Promise<object[]>} 事件命中结果
 */
async function recallEvents(queryVector, allEvents, vectorConfig, focusEntities, metrics) {
    const { chatId } = getContext();
    if (!chatId || !queryVector?.length || !allEvents?.length) {
        return [];
    }

    const meta = await getMeta(chatId);
    const fp = getEngineFingerprint(vectorConfig);
    if (meta.fingerprint && meta.fingerprint !== fp) {
        xbLog.warn(MODULE_ID, 'Event fingerprint 不匹配');
        return [];
    }

    const eventVectors = await getAllEventVectors(chatId);
    const vectorMap = new Map(eventVectors.map(v => [v.eventId, v.vector]));

    if (!vectorMap.size) {
        return [];
    }

    const focusSet = new Set((focusEntities || []).map(normalize));

    const scored = allEvents.map(event => {
        const v = vectorMap.get(event.id);
        const baseSim = v ? cosineSimilarity(queryVector, v) : 0;

        const participants = (event.participants || []).map(p => normalize(p));
        const hasEntityMatch = participants.some(p => focusSet.has(p));

        const bonus = hasEntityMatch ? 0.05 : 0;

        return {
            _id: event.id,
            event,
            similarity: baseSim + bonus,
            _baseSim: baseSim,
            _hasEntityMatch: hasEntityMatch,
            vector: v,
        };
    });

    if (metrics) {
        metrics.event.inStore = allEvents.length;
    }

    let candidates = scored
        .filter(s => s.similarity >= CONFIG.EVENT_MIN_SIMILARITY)
        .sort((a, b) => b.similarity - a.similarity)
        .slice(0, CONFIG.EVENT_CANDIDATE_MAX);

    if (metrics) {
        metrics.event.considered = candidates.length;
    }

    // 实体过滤
    if (focusSet.size > 0) {
        const beforeFilter = candidates.length;

        candidates = candidates.filter(c => {
            if (c.similarity >= 0.85) return true;
            return c._hasEntityMatch;
        });

        if (metrics) {
            metrics.event.entityFilter = {
                focusEntities: focusEntities || [],
                before: beforeFilter,
                after: candidates.length,
                filtered: beforeFilter - candidates.length,
            };
        }
    }

    // MMR 选择
    const selected = mmrSelect(
        candidates,
        CONFIG.EVENT_SELECT_MAX,
        CONFIG.EVENT_MMR_LAMBDA,
        c => c.vector,
        c => c.similarity
    );

    let directCount = 0;
    let relatedCount = 0;

    const results = selected.map(s => {
        const recallType = s._hasEntityMatch ? 'DIRECT' : 'RELATED';
        if (recallType === 'DIRECT') directCount++;
        else relatedCount++;

        return {
            event: s.event,
            similarity: s.similarity,
            _recallType: recallType,
            _baseSim: s._baseSim,
        };
    });

    if (metrics) {
        metrics.event.selected = results.length;
        metrics.event.byRecallType = { direct: directCount, related: relatedCount, causal: 0 };
        metrics.event.similarityDistribution = calcSimilarityStats(results.map(r => r.similarity));
    }

    return results;
}

// ═══════════════════════════════════════════════════════════════════════════
// [Causation] 因果链追溯
// ═══════════════════════════════════════════════════════════════════════════

/**
 * 构建事件索引
 * @param {object[]} allEvents - 所有事件
 * @returns {Map<string, object>} 事件索引
 */
function buildEventIndex(allEvents) {
    const map = new Map();
    for (const e of allEvents || []) {
        if (e?.id) map.set(e.id, e);
    }
    return map;
}

/**
 * 追溯因果链
 * @param {object[]} eventHits - 事件命中结果
 * @param {Map<string, object>} eventIndex - 事件索引
 * @param {number} maxDepth - 最大深度
 * @returns {{results: object[], maxDepth: number}}
 */
function traceCausation(eventHits, eventIndex, maxDepth = CONFIG.CAUSAL_CHAIN_MAX_DEPTH) {
    const out = new Map();
    const idRe = /^evt-\d+$/;
    let maxActualDepth = 0;

    function visit(parentId, depth, chainFrom) {
        if (depth > maxDepth) return;
        if (!idRe.test(parentId)) return;

        const ev = eventIndex.get(parentId);
        if (!ev) return;

        if (depth > maxActualDepth) maxActualDepth = depth;

        const existed = out.get(parentId);
        if (!existed) {
            out.set(parentId, { event: ev, depth, chainFrom: [chainFrom] });
        } else {
            if (depth < existed.depth) existed.depth = depth;
            if (!existed.chainFrom.includes(chainFrom)) existed.chainFrom.push(chainFrom);
        }

        for (const next of (ev.causedBy || [])) {
            visit(String(next || '').trim(), depth + 1, chainFrom);
        }
    }

    for (const r of eventHits || []) {
        const rid = r?.event?.id;
        if (!rid) continue;
        for (const cid of (r.event?.causedBy || [])) {
            visit(String(cid || '').trim(), 1, rid);
        }
    }

    const results = Array.from(out.values())
        .sort((a, b) => {
            const refDiff = b.chainFrom.length - a.chainFrom.length;
            if (refDiff !== 0) return refDiff;
            return a.depth - b.depth;
        })
        .slice(0, CONFIG.CAUSAL_INJECT_MAX);

    return { results, maxDepth: maxActualDepth };
}

// ═══════════════════════════════════════════════════════════════════════════
// 辅助函数
// ═══════════════════════════════════════════════════════════════════════════

/**
 * 获取最近消息
 * @param {object[]} chat - 聊天记录
 * @param {number} count - 消息数量
 * @param {boolean} excludeLastAi - 是否排除最后的 AI 消息
 * @returns {object[]} 最近消息
 */
function getLastMessages(chat, count = 4, excludeLastAi = false) {
    if (!chat?.length) return [];

    let messages = [...chat];

    if (excludeLastAi && messages.length > 0 && !messages[messages.length - 1]?.is_user) {
        messages = messages.slice(0, -1);
    }

    return messages.slice(-count);
}

/**
 * 构建查询文本
 * @param {object[]} chat - 聊天记录
 * @param {number} count - 消息数量
 * @param {boolean} excludeLastAi - 是否排除最后的 AI 消息
 * @returns {string} 查询文本
 */
export function buildQueryText(chat, count = 2, excludeLastAi = false) {
    if (!chat?.length) return '';

    let messages = chat;
    if (excludeLastAi && messages.length > 0 && !messages[messages.length - 1]?.is_user) {
        messages = messages.slice(0, -1);
    }

    return messages.slice(-count).map(m => {
        const text = cleanForRecall(m.mes);
        const speaker = m.name || (m.is_user ? '用户' : '角色');
        return `${speaker}: ${text.slice(0, 500)}`;
    }).filter(Boolean).join('\n');
}

// ═══════════════════════════════════════════════════════════════════════════
// 主函数
// ═══════════════════════════════════════════════════════════════════════════

/**
 * 执行记忆召回
 * @param {string} queryText - 查询文本
 * @param {object[]} allEvents - 所有事件（L2）
 * @param {object} vectorConfig - 向量配置
 * @param {object} options - 选项
 * @returns {Promise<object>} 召回结果
 */
export async function recallMemory(queryText, allEvents, vectorConfig, options = {}) {
    const T0 = performance.now();
    const { chat, name1 } = getContext();
    const { pendingUserMessage = null, excludeLastAi = false } = options;

    const metrics = createMetrics();

    if (!allEvents?.length) {
        metrics.anchor.needRecall = false;
        return {
            events: [],
            evidenceChunks: [],
            causalChain: [],
            focusEntities: [],
            elapsed: 0,
            logText: 'No events.',
            metrics,
        };
    }

    // ═══════════════════════════════════════════════════════════════════════
    // Step 1: Query Expansion
    // ═══════════════════════════════════════════════════════════════════════

    const T_QE_Start = performance.now();

    const lastMessages = getLastMessages(chat, 4, excludeLastAi);

    let expansion = { focus: [], queries: [] };
    try {
        expansion = await expandQueryCached(lastMessages, {
            pendingUserMessage,
            timeout: CONFIG.QUERY_EXPANSION_TIMEOUT,
        });
        xbLog.info(MODULE_ID, `Query Expansion: focus=[${expansion.focus.join(',')}] queries=${expansion.queries.length}`);
    } catch (e) {
        xbLog.warn(MODULE_ID, 'Query Expansion 失败，降级使用原始文本', e);
    }

    const searchText = buildSearchText(expansion);
    const finalSearchText = searchText || queryText || lastMessages.map(m => cleanForRecall(m.mes || '').slice(0, 200)).join(' ');

    const focusEntities = removeUserNameFromFocus(expansion.focus, name1);

    metrics.anchor.needRecall = true;
    metrics.anchor.focusEntities = focusEntities;
    metrics.anchor.queries = expansion.queries || [];
    metrics.anchor.queryExpansionTime = Math.round(performance.now() - T_QE_Start);
    metrics.timing.queryExpansion = metrics.anchor.queryExpansionTime;

    // ═══════════════════════════════════════════════════════════════════════
    // Step 2: 向量化查询
    // ═══════════════════════════════════════════════════════════════════════

    let queryVector;
    try {
        const [vec] = await embed([finalSearchText], vectorConfig, { timeout: 10000 });
        queryVector = vec;
    } catch (e) {
        xbLog.error(MODULE_ID, '向量化失败', e);
        metrics.timing.total = Math.round(performance.now() - T0);
        return {
            events: [],
            evidenceChunks: [],
            causalChain: [],
            focusEntities,
            elapsed: metrics.timing.total,
            logText: 'Embedding failed.',
            metrics,
        };
    }

    if (!queryVector?.length) {
        metrics.timing.total = Math.round(performance.now() - T0);
        return {
            events: [],
            evidenceChunks: [],
            causalChain: [],
            focusEntities,
            elapsed: metrics.timing.total,
            logText: 'Empty query vector.',
            metrics,
        };
    }

    // ═══════════════════════════════════════════════════════════════════════
    // Step 3: Anchor (L0) 检索
    // ═══════════════════════════════════════════════════════════════════════

    const T_Anchor_Start = performance.now();

    const { hits: anchorHits, floors: anchorFloors } = await recallAnchors(queryVector, vectorConfig, metrics);

    metrics.timing.anchorSearch = Math.round(performance.now() - T_Anchor_Start);

    // ═══════════════════════════════════════════════════════════════════════
    // Step 4: Evidence (L1) 拉取 + 粗筛 + Rerank
    // ═══════════════════════════════════════════════════════════════════════

    const T_Evidence_Start = performance.now();

    const rerankQuery = buildRerankQuery(expansion, lastMessages, pendingUserMessage);
    const evidenceChunks = await pullEvidenceByFloors(anchorFloors, anchorHits, queryVector, rerankQuery, metrics);

    metrics.timing.evidenceRetrieval = Math.round(performance.now() - T_Evidence_Start);

    // ═══════════════════════════════════════════════════════════════════════
    // Step 5: Event (L2) 独立检索
    // ═══════════════════════════════════════════════════════════════════════

    const T_Event_Start = performance.now();

    const eventHits = await recallEvents(queryVector, allEvents, vectorConfig, focusEntities, metrics);

    metrics.timing.eventRetrieval = Math.round(performance.now() - T_Event_Start);

    // ═══════════════════════════════════════════════════════════════════════
    // Step 6: 因果链追溯
    // ═══════════════════════════════════════════════════════════════════════

    const eventIndex = buildEventIndex(allEvents);
    const { results: causalMap, maxDepth: causalMaxDepth } = traceCausation(eventHits, eventIndex);

    const recalledIdSet = new Set(eventHits.map(x => x?.event?.id).filter(Boolean));
    const causalChain = causalMap
        .filter(x => x?.event?.id && !recalledIdSet.has(x.event.id))
        .map(x => ({
            event: x.event,
            similarity: 0,
            _recallType: 'CAUSAL',
            _causalDepth: x.depth,
            chainFrom: x.chainFrom,
        }));

    if (metrics.event.byRecallType) {
        metrics.event.byRecallType.causal = causalChain.length;
    }
    metrics.event.causalChainDepth = causalMaxDepth;
    metrics.event.causalCount = causalChain.length;

    // ═══════════════════════════════════════════════════════════════════════
    // 完成
    // ═══════════════════════════════════════════════════════════════════════

    metrics.timing.total = Math.round(performance.now() - T0);

    metrics.event.entityNames = focusEntities;
    metrics.event.entitiesUsed = focusEntities.length;

    console.group('%c[Recall v5]', 'color: #7c3aed; font-weight: bold');
    console.log(`Elapsed: ${metrics.timing.total}ms`);
    console.log(`Query Expansion: focus=[${expansion.focus.join(', ')}]`);
    console.log(`Anchors: ${anchorHits.length} hits → ${anchorFloors.size} floors`);
    console.log(`Evidence: ${metrics.evidence.chunkTotal || 0} L1 → ${metrics.evidence.chunkAfterCoarse || 0} coarse → ${evidenceChunks.length} final`);
    if (metrics.evidence.rerankApplied) {
        console.log(`Evidence Rerank: ${metrics.evidence.beforeRerank} → ${metrics.evidence.afterRerank} (${metrics.evidence.rerankTime}ms)`);
    }
    console.log(`Events: ${eventHits.length} hits, ${causalChain.length} causal`);
    console.groupEnd();

    return {
        events: eventHits,
        causalChain,
        evidenceChunks,
        expansion,
        focusEntities,
        elapsed: metrics.timing.total,
        metrics,
    };
}