@hupan56/wlkj 3.0.0 → 3.1.2

package/templates/.qoder/learning/eval-history.jsonl

@@ -1,28 +0,0 @@
-{"a1_reality": 0.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-good.md", "score_pct": 0.0, "time": "2026-06-23T15:06:47.512881"}
-{"a1_reality": 1.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-poor.md", "score_pct": 57.1, "time": "2026-06-23T15:06:48.151944"}
-{"a1_reality": 0.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-good.md", "score_pct": 0.0, "time": "2026-06-23T15:07:06.380363"}
-{"a1_reality": 1.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-poor.md", "score_pct": 57.1, "time": "2026-06-23T15:07:20.277445"}
-{"a1_reality": 0.5, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-good.md", "score_pct": 28.6, "time": "2026-06-23T15:08:26.677411"}
-{"a1_reality": 1.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-poor.md", "score_pct": 57.1, "time": "2026-06-23T15:08:27.209288"}
-{"a1_reality": 0.5, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-good.md", "score_pct": 28.6, "time": "2026-06-23T15:08:55.802944"}
-{"a1_reality": 1.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-poor.md", "score_pct": 57.1, "time": "2026-06-23T15:08:56.234057"}
-{"a1_reality": 0.5, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-good.md", "score_pct": 28.6, "time": "2026-06-23T15:11:32.306461"}
-{"a1_reality": 1.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-poor.md", "score_pct": 57.1, "time": "2026-06-23T15:11:32.761987"}
-{"a1_reality": 0.5, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-good.md", "score_pct": 28.6, "time": "2026-06-23T15:19:28.644771"}
-{"a1_reality": 1.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-poor.md", "score_pct": 57.1, "time": "2026-06-23T15:19:29.240469"}
-{"a1_reality": 0.5, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-good.md", "score_pct": 28.6, "time": "2026-06-23T15:26:51.416116"}
-{"a1_reality": 1.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-poor.md", "score_pct": 57.1, "time": "2026-06-23T15:26:52.129295"}
-{"a1_reality": 0.5, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-good.md", "score_pct": 28.6, "time": "2026-06-23T15:32:16.110663"}
-{"a1_reality": 1.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-poor.md", "score_pct": 57.1, "time": "2026-06-23T15:32:16.669572"}
-{"a1_reality": 0.5, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-good.md", "score_pct": 28.6, "time": "2026-06-23T15:35:31.618489"}
-{"a1_reality": 1.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-poor.md", "score_pct": 57.1, "time": "2026-06-23T15:35:32.085857"}
-{"a1_reality": 0.5, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-good.md", "score_pct": 28.6, "time": "2026-06-23T15:44:50.176279"}
-{"a1_reality": 1.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-poor.md", "score_pct": 57.1, "time": "2026-06-23T15:44:50.849511"}
-{"a1_reality": 0.5, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-good.md", "score_pct": 28.6, "time": "2026-06-23T15:49:54.375114"}
-{"a1_reality": 1.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-poor.md", "score_pct": 57.1, "time": "2026-06-23T15:49:54.852542"}
-{"a1_reality": 0.5, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-good.md", "score_pct": 28.6, "time": "2026-06-23T16:09:00.572162"}
-{"a1_reality": 1.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-poor.md", "score_pct": 57.1, "time": "2026-06-23T16:09:01.184157"}
-{"a1_reality": 0.5, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-good.md", "score_pct": 28.6, "time": "2026-06-23T16:28:54.661703"}
-{"a1_reality": 1.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-poor.md", "score_pct": 57.1, "time": "2026-06-23T16:28:55.227433"}
-{"a1_reality": 0.5, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-good.md", "score_pct": 28.6, "time": "2026-06-23T16:46:57.600191"}
-{"a1_reality": 1.0, "a2_style": null, "a3_template": 0.0, "passed": false, "prd": "req-poor.md", "score_pct": 57.1, "time": "2026-06-23T16:46:58.135094"}

package/templates/data/index/wiki-index.json

@@ -1,8 +0,0 @@
-{
-  "index": {},
-  "stats": {
-    "projects": 0,
-    "pages": 0,
-    "keywords": 0
-  }
-}

package/templates/qoder/nul

@@ -1,4 +0,0 @@
-dir: cannot access '/s': No such file or directory
-dir: cannot access '/b': No such file or directory
-dir: cannot access '*.py': No such file or directory
-dir: cannot access '*.md': No such file or directory

package/templates/qoder/scripts/kg/extract.py

@@ -1,398 +0,0 @@
-#!/usr/bin/env python3
-# -*- coding: utf-8 -*-
-"""extract.py - 三种提取策略 (Strategy 模式, v3.0 合并3文件)
-
-v3.0 合并自: extract_routes(regex扫路由) + extract_routes_tree(API拉路由树) + extract_api_params(提取API参数)。
-三种"提取"是同一抽象的不同策略, 合一用子命令分发。
-
-用法:
-    python extract.py routes         regex扫前端路由→frontend-routes.json
-    python extract.py routes-tree    API拉菜单树→menu-routes-tree.json (需登录态)
-    python extract.py api-params     提取API参数→api-params.json
-"""
-# v3.0 路径自举: 引导到 common/bootstrap, 统一 sys.path 逻辑
-import os as _o, sys as _s
-_cp = _o.path.join(_o.path.dirname(_o.path.dirname(_o.path.abspath(__file__))), 'common')
-if _cp not in _s.path: _s.path.insert(0, _cp)
-from bootstrap import setup; setup()
-
-import sys, os, re, glob, json
-from common.paths import PROJECT_ROOT
-
-# ==================== routes-tree 的辅助函数 ====================
-def read_developer():
-    f = os.path.join(PROJECT, '.qoder', '.developer')
-    for ln in open(f, encoding='utf-8'):
-        if ln.strip().startswith('name='):
-            return ln.strip().split('=', 1)[1].strip()
-    raise RuntimeError('无 .developer, 先 /wl-init')
-
-
-
-# ==================== Strategy 1: routes (regex扫) ====================
-def main_routes():
-    """提取前端路由 (regex扫.ts文件). 原extract_routes.py"""
-    router_dir = str(PROJECT_ROOT / 'data' / 'code' / 'fywl-ui' / 'apps' / 'ics' / 'src' / 'router')
-    routes = []
-
-    for rf in glob.glob(os.path.join(router_dir, '**', '*.ts'), recursive=True):
-        try:
-            txt = open(rf, encoding='utf-8', errors='ignore').read()
-        except Exception:
-            continue
-        # 匹配 path: 'xxx' 和 component: () => import('xxx')
-        # 或 path: 'xxx', name: 'yyy'
-        for m in re.finditer(
-            r"path:\s*['\"`]([^'\"`]+)['\"`]"
-            r"[^}]*?(?:name:\s*['\"]([^'\"]+)['\"])?"
-            r"[^}]*?(?:component:\s*\([^)]*\)\s*=>\s*import\(['\"]([^'\"]+)['\"]\))?",
-            txt, re.S):
-            path, name, component = m.group(1), m.group(2) or '', m.group(3) or ''
-            if path and not path.startswith(':') and len(path) > 1:
-                routes.append({'path': path, 'name': name, 'component': component})
-
-    # 也扫后台菜单路由(通常在数据库 sys_menu 表, 但前端可能有硬编码)
-    # 搜 views 目录下的 Vue 文件路径作为页面标识
-    views_dir = str(PROJECT_ROOT / 'data' / 'code' / 'fywl-ui' / 'apps' / 'ics' / 'src' / 'views')
-    vue_views = glob.glob(os.path.join(views_dir, '**', '*.vue'), recursive=True)
-    view_paths = [v.replace('\\', '/').split('/views/')[-1].replace('.vue', '') for v in vue_views]
-
-    print('路由定义: %d 条' % len(routes))
-    print('Vue 页面: %d 个' % len(view_paths))
-
-    # 保存路由索引
-    output = {
-        'version': '1.0',
-        'total_routes': len(routes),
-        'total_views': len(view_paths),
-        'routes': routes,
-        'views': sorted(view_paths),
-    }
-    out_path = str(PROJECT_ROOT / 'data' / 'index' / 'frontend-routes.json')
-    with open(out_path, 'w', encoding='utf-8') as f:
-        json.dump(output, f, ensure_ascii=False, indent=2)
-    print('已保存: %s' % out_path)
-
-    # 样本
-    print('\n样本路由:')
-    for r in routes[:8]:
-        print('  %s -> %s' % (r['path'], r['component'][:50] if r['component'] else r['name']))
-
-    # 保险相关
-    print('\n保险相关路由:')
-    for r in routes:
-        if 'insur' in r['path'].lower() or '保险' in r.get('name', ''):
-            print('  %s' % r['path'])
-
-
-# ==================== Strategy 2: routes-tree (API拉) ====================
-def main_routes_tree():
-    """从getRouters API拉菜单树(准确path). 原extract_routes_tree.py"""
-    dev = read_developer()
-    auth_file = os.path.join(PROJECT, 'workspace', 'members', dev, 'auth-state.json')
-    if not os.path.exists(auth_file):
-        print('无 auth-state, 先登录: python .qoder/scripts/kg_auto_login.py'); sys.exit(1)
-    auth = json.load(open(auth_file, encoding='utf-8'))
-    token = auth['access_token']
-    base = auth['base_url']; prefix = auth.get('api_prefix', '/test-api')
-
-    print('拉取 getRouters...')
-    r = requests.get(base + prefix + '/system/menu/getRouters',
-                     headers={'Authorization': 'Bearer ' + token, 'clientid': 'e5cd7e4891bf95d1d19206ce24a7b32e'},
-                     timeout=20)
-    d = r.json()
-    if d.get('code') != 200:
-        print('getRouters 失败:', d.get('msg')); sys.exit(1)
-    tree = d.get('data') or []
-
-
-    def walk(nodes, parent_path=''):
-        """递归拼接准确 path, 返回 [{full_path, name, component}].
-        name 优先用 meta.title(中文), 回退 name(i18n key)."""
-        out = []
-        for n in nodes:
-            seg = n.get('path', '').lstrip('/')
-            meta = n.get('meta') or {}
-            name = meta.get('title', '') or n.get('name', '')
-            comp = n.get('component', '')
-            full = (parent_path.rstrip('/') + '/' + seg) if parent_path else '/' + seg
-            while '//' in full:
-                full = full.replace('//', '/')
-            kids = n.get('children') or []
-            if kids:
-                out.extend(walk(kids, full))
-            elif comp and comp != '#' and ':' not in full:
-                out.append({'path': full, 'name': name, 'component': comp, 'i18n_key': n.get('name', '')})
-        return out
-
-
-    pages = walk(tree)
-    # 去重
-    seen = set(); uniq = []
-    for p in pages:
-        if p['path'] not in seen:
-            seen.add(p['path']); uniq.append(p)
-
-    out_file = os.path.join(PROJECT, 'data', 'index', 'menu-routes-tree.json')
-    json.dump({
-        'source': 'getRouters API (父子树递归拼接, 准确 path)',
-        'total': len(uniq),
-        'routes': uniq,
-    }, open(out_file, 'w', encoding='utf-8'), ensure_ascii=False, indent=2)
-
-    print('已存 %d 条准确路由 -> %s' % (len(uniq), out_file))
-    print()
-    # 验证关键页面
-    for kw in ['abnormalRecords', 'insurance', 'ticket']:
-        hits = [p for p in uniq if kw.lower() in (p['path'] + p['name']).lower()]
-        for h in hits[:2]:
-            print('  %-12s -> %s' % (h['name'][:10], h['path']))
-
-
-# ==================== Strategy 3: api-params (来自extract_api_params) ====================
-BASE = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-KG_PATH = os.path.join(BASE, 'data', 'index', 'kg.duckdb')
-PARAMS_FILE = os.path.join(BASE, 'data', 'index', 'api-params.json')
-
-
-# ── TS 解析 ──────────────────────────────────────────────────
-
-def parse_ts_param_types():
-    """扫所有 TS api 文件, 建 {fn_name: param_type} 映射。"""
-    fn_map = {}
-    ts_root = os.path.join(BASE, 'data', 'code', 'fywl-ui', 'apps', 'ics', 'src', 'api')
-    for tf in glob.glob(os.path.join(ts_root, '**', '*.ts'), recursive=True):
-        try:
-            if os.path.getsize(tf) > 50000:
-                continue
-            txt = open(tf, encoding='utf-8', errors='ignore').read()
-        except Exception:
-            continue
-        for m in re.finditer(r'export\s+(?:async\s+)?function\s+(\w+)\s*\(\s*(?:\w+)?\s*(?::\s*(\w+|Partial<[^>]+>))?', txt):
-            fn_map[m.group(1)] = m.group(2) or 'none'
-        for m in re.finditer(r'export\s+const\s+(\w+)\s*=\s*\(\s*(?:\w+)?\s*(?::\s*(\w+|Partial<[^>]+>|Record<[^>]+>))?', txt):
-            fn_map[m.group(1)] = m.group(2) or 'none'
-    return fn_map
-
-
-# ── Java Bo 解析 ─────────────────────────────────────────────
-
-def build_bo_index():
-    """预扫所有 Bo/Dto/Query 类, 建 {类名: 路径} 索引。"""
-    index = {}
-    root = os.path.join(BASE, 'data', 'code', 'fywl-ics')
-    for pattern in ['**/*Bo.java', '**/*Dto.java', '**/*QueryBo.java', '**/*Query.java']:
-        for f in glob.glob(os.path.join(root, pattern), recursive=True):
-            if 'target' in f or '/classes/' in f:
-                continue
-            base_name = os.path.basename(f).replace('.java', '')
-            for suffix in ['QueryBo', 'Bo', 'Dto', 'Query']:
-                if base_name.endswith(suffix):
-                    index[base_name] = f
-                    short = base_name[:-len(suffix)]
-                    if short not in index:
-                        index[short] = f
-                    break
-            else:
-                index[base_name] = f
-    return index
-
-
-def gen_default(java_type, field_name):
-    """按 Java 类型 + 字段名生成智能默认值。"""
-    FIELD_DEFAULTS = {
-        'pageNum': 1, 'pageSize': 10, 'page': 1, 'size': 10, 'current': 1,
-        'id': 1, 'parentId': 0, 'pid': 0, 'deptId': 100,
-        'userId': 1, 'tenantId': '000000',
-        'status': 0, 'type': '', 'code': '', 'name': 'test',
-        'phone': '13800000000', 'mobile': '13800000000', 'email': 'test@test.com',
-        'isAsc': 'asc', 'orderByColumn': 'id',
-    }
-    if field_name in FIELD_DEFAULTS:
-        return FIELD_DEFAULTS[field_name]
-    fn = (field_name or '').lower()
-    if 'phone' in fn or 'mobile' in fn:
-        return '13800000000'
-    if 'email' in fn:
-        return 'test@test.com'
-    if 'date' in fn or 'time' in fn:
-        return '2026-01-01'
-    if 'name' in fn:
-        return 'test'
-    if 'code' in fn:
-        return 'TEST001'
-    if fn == 'id' or fn.endswith('id'):
-        return 1
-    jt = (java_type or '').strip()
-    if jt in ('Long', 'long', 'Integer', 'int', 'BigInteger', 'BigDecimal', 'Number'):
-        return 1
-    if jt in ('String', 'CharSequence', 'char'):
-        return 'test'
-    if jt in ('Boolean', 'boolean'):
-        return False
-    if jt in ('Date', 'LocalDate', 'LocalDateTime', 'Timestamp'):
-        return '2026-01-01'
-    if jt in ('Double', 'double', 'Float', 'float'):
-        return 1.0
-    if jt.startswith('List') or jt.startswith('Set') or jt.startswith('Collection'):
-        return []
-    if jt.startswith('Map'):
-        return {}
-    return None
-
-
-def parse_bo_fields(java_path):
-    """解析 Java Bo 类的字段+类型+校验注解。"""
-    try:
-        if not os.path.isfile(java_path) or os.path.getsize(java_path) > 80000:
-            return []
-        txt = open(java_path, encoding='utf-8', errors='ignore').read()
-    except Exception:
-        return []
-    fields = []
-    try:
-        for m in re.finditer(
-            r'((?:@\w+(?:\([^)]*\))?\s*)*)'
-            r'(?:private\s+)?(\w+(?:<[^>]+>)?)\s+(\w+)\s*;', txt):
-            anns, jtype, fname = m.group(1), m.group(2), m.group(3)
-            if fname in ('serialVersionUID', 'logger', 'LOG') or fname.isupper():
-                continue
-            required = bool(re.search(r'@(?:NotNull|NotBlank|NotEmpty|Valid)\b', anns))
-            fields.append({'name': fname, 'type': jtype, 'required': required})
-    except Exception:
-        pass
-    return fields
-
-
-_BO_CACHE = {}
-
-
-def bo_to_params(bo_class, bo_index):
-    """从 Bo 类名解析出参数 JSON (带缓存)。"""
-    if not bo_class or bo_class in _BO_CACHE:
-        return _BO_CACHE.get(bo_class)
-    jf = bo_index.get(bo_class) or bo_index.get(bo_class + 'Bo') or bo_index.get(bo_class + 'Dto')
-    if not jf:
-        _BO_CACHE[bo_class] = None
-        return None
-    fields = parse_bo_fields(jf)
-    if not fields:
-        _BO_CACHE[bo_class] = None
-        return None
-    params = {}
-    for f in fields:
-        val = gen_default(f['type'], f['name'])
-        if f['required'] and val is None:
-            val = 1 if 'id' in f['name'].lower() else 'test'
-        params[f['name']] = val
-    _BO_CACHE[bo_class] = params
-    return params
-
-
-# ── 参数生成 ─────────────────────────────────────────────────
-
-def gen_params(ptype, bo_index=None):
-    """按 TS 参数类型生成默认参数 JSON。Bo 类型解析 Java 字段。"""
-    if ptype == 'PageQuery':
-        return {'pageNum': 1, 'pageSize': 10, 'isAsc': 'asc', 'orderByColumn': 'id'}, 'PageQuery'
-    if ptype in ('ID', 'Id'):
-        return {'id': 1}, 'ID'
-    if ptype == 'IDS':
-        return {'ids': [1]}, 'IDS'
-    if ptype == 'string':
-        return {'id': '1'}, 'string'
-    if ptype == 'number':
-        return {'id': 1}, 'number'
-    if ptype and ptype not in ('any', 'none', 'Recordable', 'Partial', 'Record', 'boolean', 'Blob'):
-        if bo_index:
-            bp = bo_to_params(ptype, bo_index)
-            if bp:
-                return bp, 'Bo:' + ptype
-        return {}, 'Bo:' + ptype
-    if ptype == 'any':
-        return {}, 'any'
-    return {}, 'none'
-
-
-# ── 主流程 ───────────────────────────────────────────────────
-
-def main_api_params():
-    t0 = time.time()
-
-    # 1. 读主库(只读)拿接口列表
-    kg = duckdb.connect(KG_PATH, read_only=True)
-    rows = kg.execute("SELECT fn_name, url, verb FROM api_functions WHERE url IS NOT NULL").fetchall()
-    kg.close()
-    sys.stderr.write('[%.1fs] DuckDB: %d rows\n' % (time.time() - t0, len(rows)))
-
-    # 2. TS 参数类型
-    fn_map = parse_ts_param_types()
-    sys.stderr.write('[%.1fs] TS: %d fns\n' % (time.time() - t0, len(fn_map)))
-
-    # 2b. Bo 类索引
-    bo_index = build_bo_index()
-    sys.stderr.write('[%.1fs] Bo index: %d\n' % (time.time() - t0, len(bo_index)))
-
-    # 3. 生成参数
-    results = []
-    source_dist = {}
-    for fn, url, verb in rows:
-        if not url or url.startswith('$'):
-            continue
-        ptype = fn_map.get(fn, 'none')
-        params, source = gen_params(ptype, bo_index)
-        req = [k for k, v in params.items() if v is not None]
-        results.append({
-            'url': url, 'verb': verb or '', 'param_source': source,
-            'params': params, 'required': req,
-        })
-        skey = source.split(':')[0]
-        source_dist[skey] = source_dist.get(skey, 0) + 1
-
-    sys.stderr.write('[%.1fs] Generated: %d\n' % (time.time() - t0, len(results)))
-
-    # 4. 写 JSON
-    output = {
-        'version': '1.0',
-        'extracted_at': datetime.now().isoformat()[:19],
-        'total': len(results),
-        'apis': results,
-    }
-    with open(PARAMS_FILE, 'w', encoding='utf-8') as f:
-        json.dump(output, f, ensure_ascii=False, indent=2)
-    sys.stderr.write('[%.1fs] Written: %s\n' % (time.time() - t0, PARAMS_FILE))
-
-    # 5. 报告
-    sys.stderr.write('\n=== API 参数提取 ===\n')
-    sys.stderr.write('总接口: %d, 已提取: %d\n' % (len(rows), len(results)))
-    for s, c in sorted(source_dist.items(), key=lambda x: -x[1]):
-        sys.stderr.write('  %-15s %d\n' % (s, c))
-
-    # 样本
-    for r in results:
-        if r['param_source'] not in ('none', 'any') and r['params']:
-            sys.stderr.write('\n样本: %s [%s] %s\n' % (
-                r['url'][:50], r['param_source'],
-                json.dumps(r['params'], ensure_ascii=False)[:60]))
-            break
-
-
-
-
-# ============================================================
-# Facade: 统一入口分发
-# ============================================================
-def main():
-    cmd = sys.argv[1] if len(sys.argv) > 1 else 'routes'
-    sys.argv = [sys.argv[0]] + sys.argv[2:]
-    if cmd == 'routes':
-        main_routes()
-    elif cmd == 'routes-tree':
-        main_routes_tree()
-    elif cmd == 'api-params':
-        main_api_params()
-    else:
-        print('用法: extract.py [routes|routes-tree|api-params]')
-
-if __name__ == '__main__':
-    main()

package/templates/qoder/scripts/kg/kg_semantic.py

@@ -1,156 +0,0 @@
-# -*- coding: utf-8 -*-
-# v3.0 路径自举: 引导到 common/bootstrap, 统一 sys.path 逻辑
-import os as _o, sys as _s
-_cp = _o.path.join(_o.path.dirname(_o.path.dirname(_o.path.abspath(__file__))), 'common')
-if _cp not in _s.path: _s.path.insert(0, _cp)
-from bootstrap import setup; setup()
-
-"""
-LLM semantic labeling layer for the knowledge graph.
-
-After kg_build.py extracts structural data (AST-level), this module:
-  1. Finds entities that lack semantic labels (cn == dir, no business_domain)
-  2. Outputs a labeling prompt for the AI (QoderWork) to fill in
-  3. Provides a write-back function to store labels in DuckDB
-
-The AI (QoderWork's LLM) sees the prompt during session-start or when
-kg_build reports new modules, and labels them — zero human trigger.
-
-Usage (called by AI automatically):
-  python kg_semantic.py --check       # check what needs labeling
-  python kg_semantic.py --label '{"assets": {"cn": "资产管理", "domain": "固定资产"}}'
-"""
-import json
-import os
-import sys
-
-SCRIPTS = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))  # 子包→scripts/根
-sys.path.insert(0, SCRIPTS)
-from common.paths import DATA_INDEX_DIR
-from kg_duckdb import get_db
-
-DB_PATH = str(DATA_INDEX_DIR / 'kg.duckdb')
-
-
-def check_unlabeled():
-    """Find modules that need semantic labeling (cn == dir or cn is empty)."""
-    con = get_db()
-    # Find FEATURE entities where cn == dir (unlabeled) or cn is empty
-    rows = con.execute("""
-        SELECT id, canonical, cn FROM entities
-        WHERE type = 'FEATURE'
-          AND (cn = canonical OR cn = '' OR cn IS NULL)
-        ORDER BY canonical
-    """).fetchall()
-
-    if not rows:
-        return None
-
-    # Build labeling prompt
-    modules = []
-    for eid, dirname, cn in rows:
-        # Gather context: what endpoints/pages does this module have?
-        ep_count = con.execute(
-            "SELECT COUNT(*) FROM edges WHERE from_id = ? AND edge_type = 'has_endpoint'",
-            [eid]
-        ).fetchone()[0]
-        btn_count = con.execute(
-            "SELECT COUNT(*) FROM edges WHERE from_id = ? AND edge_type = 'has_button'",
-            [eid]
-        ).fetchone()[0]
-        # Sample endpoints for context
-        sample_eps = con.execute("""
-            SELECT e.to_id FROM edges e
-            WHERE e.from_id = ? AND e.edge_type = 'has_endpoint' LIMIT 5
-        """, [eid]).fetchall()
-
-        modules.append({
-            'dir': dirname,
-            'endpoints': ep_count,
-            'buttons': btn_count,
-            'sample_apis': [r[0].replace('E:', '') for r in sample_eps],
-        })
-
-    con.close()
-
-    prompt = {
-        'action': 'semantic_labeling_needed',
-        'description': '以下模块需要语义标注（中文名+业务领域）。请根据目录名和API端点推断。',
-        'modules': modules,
-        'instructions': (
-            '对每个模块返回: {"目录名": {"cn": "中文名", "domain": "业务领域"}}\n'
-            '示例: {"assets": {"cn": "资产管理", "domain": "固定资产管理"}}\n'
-            '然后用 kg_semantic.py --label 写回'
-        ),
-    }
-    return prompt
-
-
-def apply_labels(labels_json):
-    """Write semantic labels back to DuckDB.
-
-    labels_json: {"assets": {"cn": "资产管理", "domain": "固定资产"}, ...}
-    """
-    labels = json.loads(labels_json) if isinstance(labels_json, str) else labels_json
-    con = get_db()
-    updated = 0
-
-    for dirname, info in labels.items():
-        cn = info.get('cn', dirname)
-        domain = info.get('domain', '')
-
-        # Find the FEATURE entity by canonical name (dir)
-        rows = con.execute(
-            "SELECT id FROM entities WHERE type='FEATURE' AND canonical = ?",
-            [dirname]
-        ).fetchall()
-
-        for row in rows:
-            eid = row[0]
-            # Update cn
-            con.execute(
-                "UPDATE entities SET cn = ? WHERE id = ?",
-                [cn, eid]
-            )
-            # Update props with domain
-            props_row = con.execute(
-                "SELECT props FROM entities WHERE id = ?", [eid]
-            ).fetchone()
-            if props_row:
-                try:
-                    props = json.loads(props_row[0]) if props_row[0] else {}
-                except Exception:
-                    props = {}
-                props['business_domain'] = domain
-                props['labeled'] = True
-                con.execute(
-                    "UPDATE entities SET props = ? WHERE id = ?",
-                    [json.dumps(props, ensure_ascii=False), eid]
-                )
-            # Update alias
-            con.execute(
-                "INSERT OR REPLACE INTO aliases VALUES (?, ?)",
-                [cn.lower(), eid]
-            )
-            updated += 1
-
-    con.close()
-    return updated
-
-
-if __name__ == '__main__':
-    if len(sys.argv) > 1 and sys.argv[1] == '--check':
-        result = check_unlabeled()
-        if result:
-            print(json.dumps(result, ensure_ascii=False, indent=2))
-        else:
-            print('{"status": "all_labeled", "message": "所有模块已有语义标注"}')
-
-    elif len(sys.argv) > 2 and sys.argv[1] == '--label':
-        count = apply_labels(sys.argv[2])
-        print('Updated %d entities with semantic labels' % count)
-
-    else:
-        print('Usage:')
-        print('  python kg_semantic.py --check    # 查看待标注模块')
-        print('  python kg_semantic.py --label \'{"assets":{"cn":"资产管理","domain":"固定资产"}}\'')