@fanchao8609/agent_brain_sync 0.1.0

package/src/store.js

@@ -0,0 +1,499 @@
+// src/store.js — CLI 命令实现：图谱读写层。
+// 命令: init / board / status / load / task / query / lint
+import { promises as fs } from 'node:fs';
+import { join, resolve } from 'node:path';
+import { requireBrain, findBrainRoot, brainPath } from './index.js';
+import { addTask, upsertTask, boardText, readTodo, ensureTodo, todoTemplate, today, localStamp, setBreakpoint, moveBlocked, insertDoneGrouped } from './todo.js';
+import { editFile, SKIP } from './lock.js';
+import { appendWrapup, strandedFor, wrapupLogPath } from './wrapup.js';
+
+// ---------- init: 建 .brain/ 骨架 ----------
+const BRAIN_DIRS = ['entities', 'concepts', 'sources', 'syntheses', 'sessions'];
+const BRAIN_FILES = ['index.md', 'log.md', 'todo.md'];
+
+export async function cmdInit({ dir }) {
+  const root = resolveProjectDir(dir);
+  const brain = join(root, '.brain');
+  try {
+    await fs.access(brain);
+  } catch (e) {
+    if (e && e.code !== 'ENOENT') throw e;
+    return createSkeleton(root, brain);
+  }
+  // 已存在 → 校验结构完整性；损坏则明确报错并指引 repair（不静默半初始化）
+  const report = await reportBrain({ dir: root });
+  if (report.ok) throw new Error(`已存在: ${brain} (abort)`);
+  throw new Error(
+    `已存在但结构不完整:\n  ${report.problems.join('\n  ')}\n` +
+    `  → 轻则补齐: abs init --repair    重则重建: 删掉 ${brain} 后重跑 abs init`
+  );
+}
+
+async function createSkeleton(root, brain) {
+  for (const d of BRAIN_DIRS) await fs.mkdir(join(brain, d), { recursive: true });
+  await fs.writeFile(join(brain, 'index.md'), indexTemplate(), 'utf8');
+  await fs.writeFile(join(brain, 'log.md'), logTemplate(), 'utf8');
+  await fs.writeFile(join(brain, 'todo.md'), todoTemplate(), 'utf8');
+  return `✓ 已建图谱 ${brain}\n  (模板见 SKILL.md「文件标准」)`;
+}
+
+/** 结构体检：6 目录 + 3 文件逐一核对。ok=false 时 problems 列出缺失项。 */
+export async function reportBrain({ dir }) {
+  const root = resolveProjectDir(dir);
+  const brain = join(root, '.brain');
+  const problems = [];
+  for (const d of BRAIN_DIRS) {
+    try {
+      const st = await fs.stat(join(brain, d));
+      if (!st.isDirectory()) problems.push(`缺目录 ${d}/ (被文件占位)`);
+    } catch {
+      problems.push(`缺目录 ${d}/`);
+    }
+  }
+  for (const f of BRAIN_FILES) {
+    try {
+      await fs.access(join(brain, f));
+    } catch {
+      problems.push(`缺文件 ${f}`);
+    }
+  }
+  return { ok: problems.length === 0, problems, brain };
+}
+
+/** init --repair: 只补缺失骨架，绝不覆盖已有文件。 */
+export async function cmdRepair({ dir }) {
+  const root = resolveProjectDir(dir);
+  const brain = join(root, '.brain');
+  try {
+    await fs.access(brain);
+  } catch {
+    return createSkeleton(root, brain); // 整个图谱都没有 = 全新建
+  }
+  const fixed = [];
+  for (const d of BRAIN_DIRS) {
+    try {
+      const st = await fs.stat(join(brain, d));
+      if (!st.isDirectory()) {
+        await fs.rm(join(brain, d), { force: true });
+        await fs.mkdir(join(brain, d), { recursive: true });
+        fixed.push(`${d}/`);
+      }
+    } catch {
+      await fs.mkdir(join(brain, d), { recursive: true });
+      fixed.push(`${d}/`);
+    }
+  }
+  for (const [f, tpl] of [['index.md', indexTemplate], ['log.md', logTemplate], ['todo.md', todoTemplate]]) {
+    try {
+      await fs.access(join(brain, f));
+    } catch {
+      await fs.writeFile(join(brain, f), tpl(), 'utf8');
+      fixed.push(f);
+    }
+  }
+  if (!fixed.length) return `✓ 结构完整，无需修复: ${brain}`;
+  return `✓ 已补齐 ${fixed.join(', ')} → ${brain}\n  (已有文件一律不覆盖)`;
+}
+
+function resolveProjectDir(dir) {
+  if (!dir) return process.cwd();
+  return resolve(dir);
+}
+
+export function indexTemplate() {
+  return [
+    '# 🗂 图谱索引',
+    '',
+    '本文件唯一入口。每新建/大改一页，同步在此分类下加一行 [[页面名]] — 一句话。',
+    '',
+    '## 当前路线 (Roadmap)',
+    '## Concepts',
+    '## Entities',
+    '## Sources',
+    '## Syntheses',
+    '## Sessions',
+    '',
+  ].join('\n');
+}
+
+export function logTemplate() {
+  return ['# 🗒 操作日志', '', '## [YYYY-MM-DD] ingest | 沉淀 <slug>', ''].join('\n');
+}
+
+// ---------- board: 看板 ----------
+export async function cmdBoard({ dir }) {
+  const root = await requireBrain(dir || process.cwd());
+  return boardText(root);
+}
+
+// ---------- status: 定位报告 + 图谱概要 ----------
+export async function cmdStatus({ dir }) {
+  const root = await requireBrain(dir || process.cwd());
+  const entries = await listBrainFiles(root);
+  return [
+    `📂 abs → 项目: ${root}`,
+    `图谱: ${brainPath(root)}`,
+    '',
+    entries.length ? entries.join('\n') : '(图谱为空)',
+  ].join('\n');
+}
+
+async function listBrainFiles(root) {
+  const out = [];
+  const dirs = ['concepts', 'entities', 'sources', 'syntheses', 'sessions'];
+  for (const d of dirs) {
+    const p = brainPath(root, d);
+    try {
+      const files = (await fs.readdir(p)).filter((f) => f.endsWith('.md'));
+      out.push(`${d}/: ${files.length} 页`);
+    } catch { out.push(`${d}/: 0 页`); }
+  }
+  return out;
+}
+
+// ---------- load: 开机读状态 ----------
+export async function cmdLoad({ dir }) {
+  const root = await requireBrain(dir || process.cwd());
+  const todo = await readTodo(root);
+  const index = await readIfExists(brainPath(root, 'index.md'));
+  const log = await readIfExists(brainPath(root, 'log.md'));
+  const stranded = await strandedFor(root);
+  const sections = [
+    `📂 abs → 项目: ${root}`,
+    '--- 当前路线 (index.md) ---',
+    index || '(index.md 为空)',
+    '',
+    '--- Todo 看板 (todo.md) ---',
+    todo.trim() || '(todo.md 为空)',
+    '',
+    '--- 最近动作 (log.md, 末尾 5 条) ---',
+    tailLines(log, 5) || '(log.md 为空)',
+  ];
+  if (stranded.length) {
+    const rows = stranded.map((t) => {
+      const bp = t.bp.length ? `\n    ${t.bp.map((b) => `↳ 断点: ${b}`).join('\n    ')}` : '';
+      return `  - ${t.body}${bp}`;
+    });
+    sections.splice(
+      0, 1,
+      `📂 abs → 项目: ${root}`,
+      '⏳ 上会话滞留（未 done，先对账）',
+      rows.join('\n'),
+      '→ 完成: abs task done <id>；未完: abs task note <id> --note 断点',
+      ''
+    );
+  }
+  return sections.join('\n');
+}
+
+// ---------- wrapup: 滞留快照（B）/ load 内展示由 cmdLoad 完成（A） ----------
+export async function cmdWrapup({ dir }) {
+  const root = await requireBrain(dir || process.cwd());
+  return appendWrapup(root);
+}
+
+async function readIfExists(p) {
+  try { return (await fs.readFile(p, 'utf8')).trim(); } catch { return ''; }
+}
+
+function tailLines(text, n) {
+  const lines = text.split('\n').filter((l) => l.trim());
+  return lines.slice(-n).join('\n');
+}
+
+// ---------- log: 追加工作成果沉淀摘要（用户/AI 主动 abs log "..." 记, 不收工具动作流水） ----------
+export async function cmdLog({ dir, title, kind = 'dev' }) {
+  const root = await requireBrain(dir || process.cwd());
+  const p = brainPath(root, 'log.md');
+  const stamp = localStamp();
+  const clean = String(title || '').replace(/\n/g, ' ').slice(0, 100); // 整行含前缀 ≤120
+  const line = `## [${stamp}] ${kind} | ${clean}`;
+  await editFile(p, (cur) => {
+    const text = cur ?? '# 🗒 操作日志\n';
+    // 倒序：新行插在标题后（若已是模板占位行则替换它）
+    const lines = text.split('\n');
+    const headerIdx = lines.findIndex((l) => l.startsWith('#'));
+    lines.splice(headerIdx + 1, 0, line);
+    return { text: lines.join('\n') };
+  });
+  return `✓ log → ${p}\n  ${line}`;
+}
+
+// ---------- task: 登记/推进（幂等键 = 行首 id；hook 也调这个） ----------
+// 纪律: task 过程动作(start/done/note/blocked)只改 todo.md, 不写 log.md。
+// log.md 是「工作成果沉淀摘要」(用户/AI 主动 abs log "..." 记), 不收工具动作流水。
+export async function cmdTask({ dir, action, id, section, note }) {
+  const root = await requireBrain(dir || process.cwd());
+  if (action === 'start') {
+    const text = `${id}${note ? ' — ' + note : ''}`;
+    const r = await upsertTask(root, {
+      section: section || 'Today / In Progress',
+      text: `${text} (认领 ${today()})`,
+    });
+    return `✓ 任务${r.updated ? '更新(幂等)' : '登记'} → ${brainPath(root, 'todo.md')}\n  ${id}${note ? ' — ' + note : ''}`;
+  }
+  if (action === 'blocked') {
+    const r = await moveBlocked(root, { id, reason: note });
+    return r.msg;
+  }
+  if (action === 'note') {
+    if (!note) return '用法: abs task note <id> --note "断点/进度"（实时落 ↳ 断点 行）';
+    const r = await setBreakpoint(root, { id, text: note });
+    return r.msg;
+  }
+  if (action === 'done') {
+    // 找到匹配 id 的行，勾选并归位 Done（简化：若行在某 section 则标记完成）
+    const res = await markDone(brainPath(root, 'todo.md'), id);
+    return res;
+  }
+  throw new Error(`unknown task action: ${action}`);
+}
+
+async function markDone(file, id) {
+  const res = await editFile(file, (text) => {
+    const lines = text.split('\n');
+    let changed = false;
+    let moved = null;
+    const kept = [];
+    for (let i = 0; i < lines.length; i++) {
+      const l = lines[i];
+      if (!moved && l.startsWith('- [ ]') && l.includes(id)) {
+        changed = true;
+        const head = l.replace('- [ ]', '- [x]').replace(/\(认领[^)]*\)/, '') + ` (完成 ${today()})`;
+        const bp = [];
+        while (i + 1 < lines.length && lines[i + 1].trimStart().startsWith('↳')) bp.push(lines[++i]);
+        moved = [head, ...bp];
+        continue;
+      }
+      kept.push(l);
+    }
+    if (!changed || !moved) return SKIP;
+    // 归位 + 按日期分组：insertDoneGrouped 一次重建 Done 区（新日期在前，未标日期归尾）
+    return { text: insertDoneGrouped(kept.join('\n'), moved) };
+  });
+  return res === SKIP
+    ? `(未找到含 "${id}" 的未完成任务行)`
+    : `✓ 已完成并归位 Done: ${id}`;
+}
+
+// ---------- show: 查看 index/todo/log（只读面） ----------
+export async function cmdShow({ dir, view }) {
+  const v = String(view || '').toLowerCase();
+  if (!['todo', 'index', 'log'].includes(v)) {
+    return '用法: abs <todo|index|log>  — todo=看板(原 board), index=图谱索引, log=操作流水';
+  }
+  const root = await requireBrain(dir || process.cwd());
+  const p = brainPath(root, `${v === 'todo' ? 'todo' : v}.md`);
+  // todo 走 readTodo（含老格式惰性迁移写回）；index/log 直接读
+  let text;
+  try {
+    text = v === 'todo'
+      ? await readTodo(root)          // 迁移老格式 In Progress/Todo → B4
+      : (await fs.readFile(p, 'utf8')).trim();
+  } catch {
+    return `${v}.md 不存在于 ${brainPath(root)}。初始化/补齐: abs init --repair`;
+  }
+  if (!text) return `(${v}.md 为空)`;
+  return v === 'todo' ? boardText(root, text) : text;
+}
+// ---------- query: 检索知识图谱（多词 OR，扫全 .md 页） ----------
+const KNOWN_SLUG_HINT = /模板残留|\[\[slug\]\]/;
+
+export async function cmdQuery({ dir, terms }) {
+  const words = (terms || []).map((w) => String(w).trim()).filter(Boolean);
+  if (!words.length) {
+    return '用法: abs query <词1> [词2 …]  — 多词 OR 检索 .brain/ 全部知识页';
+  }
+  let root;
+  try {
+    root = await requireBrain(dir || process.cwd());
+  } catch {
+    return `未找到 .brain/ 图谱（无记忆可查）。先在项目根运行: abs init`;
+  }
+  const hits = [];
+  const dirs = ['concepts', 'entities', 'sources', 'syntheses', 'sessions'];
+  for (const d of dirs) {
+    const p = brainPath(root, d);
+    let files;
+    try {
+      files = await fs.readdir(p);
+    } catch {
+      continue;
+    }
+    for (const f of files) {
+      if (!f.endsWith('.md') || f.startsWith('_')) continue;
+      const full = join(p, f);
+      const body = await fs.readFile(full, 'utf8').catch(() => '');
+      const matched = words.filter((w) => body.toLowerCase().includes(w.toLowerCase()));
+      if (matched.length) {
+        hits.push({ full, slug: f.replace(/\.md$/, ''), matched, snippet: firstHitLine(body, words) });
+      }
+    }
+  }
+  if (!hits.length) return `query [${words.join(', ')}]: 无命中。用 abs lint 看图谱健康；首次使用先 abs init。`;
+  const lines = hits.map((h) => `📄 ${h.slug}  (命中: ${h.matched.join(', ')})\n    ${h.snippet}`);
+  return [`query [${words.join(', ')}] → ${hits.length} 页:`, '', ...lines].join('\n');
+}
+
+function firstHitLine(body, words) {
+  const lower = body.toLowerCase();
+  for (const line of body.split('\n')) {
+    const l = line.toLowerCase();
+    if (words.some((w) => l.includes(w.toLowerCase())) && line.trim() && !KNOWN_SLUG_HINT.test(line)) {
+      return line.trim().slice(0, 100);
+    }
+  }
+  return '';
+}
+
+// ---------- note: 经验实时暂存（source 页，一念一落，防流失） ----------
+const NOTE_DEDUP_MS = 60 * 1000;
+
+export async function cmdNote({ dir, text, tags }) {
+  const clean = String(text || '').trim();
+  if (!clean) return '用法: abs note "经验/坑/技巧一句话"（落 sources/ 暂存页，实时不流失）';
+  let root;
+  try {
+    root = await requireBrain(dir || process.cwd());
+  } catch {
+    return `未找到 .brain/ 图谱。先在项目根运行: abs init`;
+  }
+  const srcDir = brainPath(root, 'sources');
+  await fs.mkdir(srcDir, { recursive: true });
+  // 幂等: 同文本 60s 内只落一份
+  const existing = (await fs.readdir(srcDir).catch(() => [])).filter((f) => f.endsWith('.md'));
+  for (const f of existing) {
+    const body = await fs.readFile(join(srcDir, f), 'utf8').catch(() => '');
+    if (body.includes(clean)) {
+      return `• 60s 内已落同文本 → ${f} (跳过重复)`;
+    }
+  }
+  const tagList = String(tags || '').split(',').map((t) => t.trim()).filter(Boolean);
+  const fmTags = ['source', ...tagList].join(', ');
+  const slugSrc = clean.slice(0, 24).replace(/[^\w一-鿿]+/g, '-').replace(/^-+|-+$/g, '').toLowerCase();
+  const file = `${today()}-${slugSrc || 'note'}.md`;
+  const body = [
+    '---',
+    `tags: [${fmTags}]`,
+    `updated: ${today()}`,
+    'status: draft',
+    '---',
+    '',
+    `# 来源：${clean.slice(0, 40)}`,
+    '',
+    `## 记录（实时暂存，Teardown 时提炼进 concepts/ 后本页可删）`,
+    `- ${clean}`,
+    '',
+    '## 关联连接',
+    '（提炼成 concepts 规律页后，在此挂双链到该页）',
+    '',
+  ].join('\n');
+  // 源文件是新写唯一文件：tmp+rename 原子落盘（避免并发读读到半写文件）
+  const srcFile = join(srcDir, file);
+  const tmp = join(srcDir, `.${file}.tmp-${Date.now()}-${Math.random().toString(36).slice(2, 6)}`);
+  await fs.writeFile(tmp, body, 'utf8');
+  await fs.rename(tmp, srcFile);
+  // index Sources 区登记（锁内幂等：别页已登记则跳过，防并发重复） + log 一行
+  const iP = brainPath(root, 'index.md');
+  const slug = file.replace(/\.md$/, '');
+  const line = `- [[${slug}]] — ${clean.slice(0, 40)}`;
+  await editFile(iP, (index) => {
+    if (!index || index.includes(`[[${slug}]]`)) return SKIP;
+    const sIdx = index.indexOf('## Sources');
+    if (sIdx === -1) return SKIP;
+    const after = index.indexOf('\n## ', sIdx + 1);
+    const next = after === -1
+      ? `${index.replace(/\s*$/, '')}\n${line}\n`
+      : index.slice(0, after) + `\n${line}` + index.slice(after);
+    return { text: next };
+  });
+  await cmdLog({ dir: root, title: clean.slice(0, 60), kind: 'note' });
+  return `✓ 经验暂存 → sources/${file}\n  ${clean.slice(0, 60)}`;
+}
+// ---------- lint: 体检（与 scripts/lint.sh 同规则的 Node 版，供 CLI/MCP 直调） ----------
+export async function cmdLint({ dir }) {
+  let root;
+  try {
+    root = await requireBrain(dir || process.cwd());
+  } catch {
+    return `未找到 .brain/ 图谱。先在项目根运行: abs init`;
+  }
+  const vault = brainPath(root);
+  const pages = await listPages(vault);
+  const names = new Set(pages.map((p) => p.slug));
+  const linkedNames = new Set(pages.flatMap((p) => p.links));
+  const issues = [];
+
+  for (const pg of pages) {
+    if (!pg.hasFrontmatter) issues.push(`NO-FRONTMATTER: ${pg.rel}`);
+    for (const ln of pg.links) {
+      if (/slug|Name|name|Date|页面名$/.test(ln)) issues.push(`TEMPLATE-LINK: ${pg.rel} -> [[${ln}]]`);
+      if (!names.has(ln)) issues.push(`DEAD-LINK: ${pg.rel} -> [[${ln}]]`);
+    }
+    // ORPHAN: sources/ 暂存页豁免（暂存线索天然孤立，提炼成 concept 前不强制挂链）
+    if (pg.dir !== 'sources' && !pg.links.length && !linkedNames.has(pg.slug)) {
+      issues.push(`ORPHAN-PAGE: ${pg.rel} (no links out, no links in)`);
+    }
+    if (/知识冲突/.test(pg.body) && /status: draft/.test(pg.frontmatter)) {
+      issues.push(`UNRESOLVED-CONFLICT: ${pg.rel}`);
+    }
+    if (['concepts', 'entities', 'syntheses'].includes(pg.dir)) {
+      if (pg.lines > 150 || pg.bytes > 5 * 1024) {
+        issues.push(`OVER-SIZE: ${pg.rel} (${pg.lines}L/${pg.bytes}B > 150L/5120B; 拆或外链)`);
+      }
+    }
+    if (pg.dir !== 'sources' && !pg.indexed) {
+      issues.push(`INDEX-MISSING: ${pg.rel} not listed as [[${pg.slug}]] in index.md`);
+    }
+  }
+
+  const nsrc = pages.filter((p) => p.dir === 'sources').length;
+  if (nsrc > 10) issues.push(`SOURCES-PILED-UP: sources/ has ${nsrc} files > 10; 提炼归档旧 source`);
+
+  const n = issues.length;
+  return [
+    ...(issues.length ? issues : []),
+    '',
+    `lint: ${n} problem(s).`,
+    n === 0 ? '✓ 图谱健康' : '',
+  ].filter(Boolean).join('\n');
+}
+
+const PAGE_DIRS = ['entities', 'concepts', 'sources', 'syntheses', 'sessions'];
+
+async function listPages(vault) {
+  let indexText = '';
+  try {
+    indexText = await fs.readFile(join(vault, 'index.md'), 'utf8');
+  } catch { /* no index yet */ }
+  const pages = [];
+  for (const d of PAGE_DIRS) {
+    const dp = join(vault, d);
+    let files;
+    try {
+      files = await fs.readdir(dp);
+    } catch {
+      continue;
+    }
+    for (const f of files) {
+      if (!f.endsWith('.md') || f.startsWith('_')) continue;
+      const full = join(dp, f);
+      const body = await fs.readFile(full, 'utf8').catch(() => '');
+      const fm = body.match(/^---\n([\s\S]*?)\n---/);
+      const links = [...new Set([...body.matchAll(/\[\[([^\]]+)\]\]/g)].map((m) => m[1].split('|')[0]))];
+      pages.push({
+        dir: d,
+        rel: `${d}/${f}`,
+        slug: f.replace(/\.md$/, ''),
+        body,
+        frontmatter: fm ? fm[1] : '',
+        hasFrontmatter: body.startsWith('---\n'),
+        links,
+        lines: body.split('\n').length,
+        bytes: Buffer.byteLength(body, 'utf8'),
+        indexed: indexText.includes(`[[${f.replace(/\.md$/, '')}]]`),
+      });
+    }
+  }
+  return pages;
+}