@hupan56/wlkj 2.7.12 → 3.0.0

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

@@ -0,0 +1,127 @@
+# -*- coding: utf-8 -*-
+"""build_relations.py — 从 trace-chain + code-api 推导关系边, 生成 relations.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()
+
+产出 data/index/_build_cache/relations.json, 供 kg_duckdb 导入 edges 表,
+让 kg.py feature 的端点/按钮数据非零, kg.py hop 多跳遍历有边可走。
+
+结构 (与 kg_duckdb.bulk_import_from_json 的 edges_data 契约对齐):
+  {
+    "edges": {
+      "click_to_api": [
+        {"vue_file":..., "button":..., "handler":..., "endpoint":..., "confidence":...}
+      ],
+      "api_to_controller": {"<endpoint>": "<controller_file>"},
+      "controller_to_tests": {"<controller>": [{"test":..., "dep":..., "assertions":[...]}]}
+    }
+  }
+
+数据源:
+  - trace-chain.json: traces (vue_file→button→handler→api_fn→endpoint→verb→confidence)
+                      api_functions (fnName→{url,verb,file})
+  - code-api.json: {endpoint: controller_file} (api_to_controller 直接用)
+  - test-assertions.json: tests (controller_to_tests 推导)
+"""
+import json
+import os
+import sys
+from datetime import datetime
+
+THIS_DIR = os.path.dirname(os.path.abspath(__file__))
+sys.path.insert(0, os.path.join(THIS_DIR, '..'))
+from common.paths import DATA_INDEX_DIR, BUILD_CACHE_DIR
+
+
+def load(name):
+    for d in (str(BUILD_CACHE_DIR), str(DATA_INDEX_DIR)):
+        p = os.path.join(d, name)
+        if os.path.isfile(p):
+            try:
+                return json.load(open(p, encoding='utf-8'))
+            except (json.JSONDecodeError, OSError):
+                pass
+    return {}
+
+
+def build():
+    tc = load('trace-chain.json')
+    api = load('code-api.json')
+
+    # 1. click_to_api: 直接从 traces 转 (字段名对齐 kg_duckdb 期望)
+    click_to_api = []
+    for t in tc.get('traces', []):
+        if not t.get('endpoint'):
+            continue
+        click_to_api.append({
+            'vue_file': t.get('vue_file', ''),
+            'button': t.get('button', ''),
+            'handler': t.get('handler', ''),
+            'api_fn': t.get('api_fn', ''),
+            'endpoint': t.get('endpoint', ''),
+            'verb': t.get('verb', ''),
+            'confidence': t.get('confidence', ''),
+        })
+
+    # 2. api_to_controller: code-api.json 已经是 {endpoint: controller_file}
+    # 过滤掉空值
+    api_to_controller = {ep: f for ep, f in api.items() if ep and f}
+
+    # 3. controller_to_tests: 从 test-assertions.json 推导
+    # test 有 package/class 信息, controller 名从 controller_file 提取, 反向匹配 test
+    ta = load('test-assertions.json')
+    controller_to_tests = {}
+    # 建 controller 名 → test 的映射 (test 的 package 含 controller 关键字则关联)
+    ctrl_names = set()
+    for ctrl_file in api_to_controller.values():
+        cname = ctrl_file.split('/')[-1].replace('.java', '')
+        if cname:
+            ctrl_names.add(cname)
+    for t in ta.get('tests', []):
+        test_loc = (t.get('package', '') + '.' + t.get('class', '')).lower()
+        for cname in ctrl_names:
+            if cname.lower() in test_loc:
+                controller_to_tests.setdefault(cname, []).append({
+                    'test': t.get('display_name') or (t.get('class', '') + '.' + t.get('method', '')),
+                    'dep': '',
+                    'assertions': t.get('assertions', []),
+                })
+                break  # 一个 test 只关联一个 controller
+
+    relations = {
+        'version': '1.0',
+        'updated': datetime.now().isoformat(),
+        'edges': {
+            'click_to_api': click_to_api,
+            'api_to_controller': api_to_controller,
+            'controller_to_tests': controller_to_tests,
+        },
+        'stats': {
+            'click_edges': len(click_to_api),
+            'api_to_ctrl': len(api_to_controller),
+            'ctrl_to_test': len(controller_to_tests),
+        },
+    }
+    return relations
+
+
+def main():
+    print('Building relations...')
+    rel = build()
+    out = os.path.join(str(BUILD_CACHE_DIR), 'relations.json')
+    os.makedirs(os.path.dirname(out), exist_ok=True)
+    with open(out, 'w', encoding='utf-8') as f:
+        json.dump(rel, f, ensure_ascii=False, indent=2, sort_keys=True)
+    s = rel['stats']
+    print('  click_to_api edges: %d' % s['click_edges'])
+    print('  api_to_controller: %d' % s['api_to_ctrl'])
+    print('  controller_to_tests: %d' % s['ctrl_to_test'])
+    print('  Saved to: %s' % out)
+
+
+if __name__ == '__main__':
+    main()

package/templates/qoder/scripts/{build_style_index.py → kg/build_style_index.py}

@@ -1,5 +1,11 @@
-#!/usr/bin/env python3
+#!/usr/bin/env python3
 # -*- 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()
+
 """
 Build style index from Vue/HTML files.
 Extracts UI patterns: tables, forms, layouts, colors, components.
@@ -39,7 +45,14 @@ try:
 except Exception:
     from common.extract import build_api_fn_table, trace_clicks
     _EXTRACT_ENGINE = 'regex (tree-sitter not installed)'
-from common.test_extract import extract_test_assertions  # noqa: E402
+# test_extract 是可选增强层 (test-assertions 索引), 模块可能未提供。
+# 用 lazy import + 降级: 没有就跳过 test-assertions, 不阻断 ui-style 主线构建。
+try:
+    from common.test_extract import extract_test_assertions  # noqa: E402
+    _HAS_TEST_EXTRACT = True
+except ImportError:
+    extract_test_assertions = None
+    _HAS_TEST_EXTRACT = False
 
 
 def extract_vue_patterns(filepath, content):
@@ -377,7 +390,12 @@ def scan_project(proj_name, proj_path):
 
         for f in files:
             filepath = os.path.join(root, f)
-            rel_path = os.path.relpath(filepath, proj_path).replace('\\', '/')
+            # relpath 在 Windows 上对特殊路径 (\\.\nul 等设备/跨盘符) 会抛 ValueError。
+            # 这些根本不是源码, 跳过即可, 不让一个异常文件炸掉整个扫描。
+            try:
+                rel_path = os.path.relpath(filepath, proj_path).replace('\\', '/')
+            except ValueError:
+                continue
 
             if f.endswith('.vue'):
                 vue_count += 1
@@ -415,7 +433,7 @@ def scan_project(proj_name, proj_path):
                         comp_m = re.findall(r"component:\s*['\"]([^'\"]+)['\"]", content)
                         for comp in comp_m:
                             form_components[comp] = form_components.get(comp, 0) + 1
-                except:
+                except Exception:
                     pass
 
     return {
@@ -662,13 +680,20 @@ def main():
 
     # test-assertions.json — E1: 测试代码提取 (Given-When-Then + 断言 + 注入依赖)
     # 第二层知识: 生产代码提取不到的业务预期和断言标准, 补上'怎么算对'天花板
-    print('Building test assertions index...')
+    # 依赖 common.test_extract (可选模块); 缺失则跳过, 不阻断 ui-style 等主线。
     all_tests = []
-    for proj in sorted(os.listdir(DATA_CODE)):
-        proj_path = os.path.join(DATA_CODE, proj)
-        if os.path.isdir(proj_path):
-            tests = extract_test_assertions(proj_path)
-            all_tests.extend(tests)
+    if not _HAS_TEST_EXTRACT:
+        print('Skipping test assertions index (common.test_extract not available - optional)')
+    else:
+        print('Building test assertions index...')
+        for proj in sorted(os.listdir(DATA_CODE)):
+            proj_path = os.path.join(DATA_CODE, proj)
+            if os.path.isdir(proj_path):
+                try:
+                    tests = extract_test_assertions(proj_path)
+                    all_tests.extend(tests)
+                except Exception as e:
+                    print('  WARN: test_extract failed for {}: {}'.format(proj, str(e)[:80]))
     test_index = {
         'version': '1.0', 'updated': datetime.now().isoformat(),
         'tests': all_tests,
@@ -679,6 +704,8 @@ def main():
         json.dump(test_index, f, ensure_ascii=False, indent=2, sort_keys=True)
 
     # ── 知识图谱层: relations + features + entity-registry + DuckDB ──
+    # build_entity_registry 已实现 (FEATURE/ENDPOINT entity 基础层)。
+    # build_relations/features/workflows 是增强层, 未实现时 entities 表仍有基础数据。
     print('Building knowledge graph layers...')
     import subprocess
     for script in ['build_relations.py', 'build_features.py',
@@ -690,6 +717,8 @@ def main():
                 print('  %s OK' % script)
             except Exception as e:
                 print('  %s SKIP: %s' % (script, str(e)[:60]))
+        elif script != 'kg_duckdb.py':
+            print('  %s N/A (语义层脚本未实现, entities/workflows 表将空)' % script)
 
     print()
     print('=== Done ===')

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

@@ -0,0 +1,144 @@
+# -*- coding: utf-8 -*-
+"""build_workflows.py — 从 traces + endpoints 推导业务流程链, 生成 workflow-index.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()
+
+产出 data/index/_build_cache/workflow-index.json, 供 kg_duckdb 导入 workflows 表,
+让 kg.py workflow <模块> 能返回该模块的状态流转链 (query→create→audit→...)。
+
+workflows 表结构: module, state, seq, button, endpoint
+
+推导逻辑 (基于 verb + 路径关键词的启发式状态推断):
+  - GET/read/list/page/export → "query" (查询)
+  - POST/add/create/save/submit/apply → "create" (新增/提交)
+  - PUT/update/edit/audit/approve/reject → "audit" (审核/修改)
+  - DELETE/remove/cancel/delete/del → "cancel" (取消/删除)
+
+数据源: trace-chain.json (vue_file→button→endpoint→verb) + entity-registry.json
+(module→endpoint 邻接, 用于把 endpoint 归到 module)。
+"""
+import json
+import os
+import sys
+from datetime import datetime
+
+THIS_DIR = os.path.dirname(os.path.abspath(__file__))
+sys.path.insert(0, os.path.join(THIS_DIR, '..'))
+from common.paths import DATA_INDEX_DIR, BUILD_CACHE_DIR
+
+
+def load(name):
+    for d in (str(BUILD_CACHE_DIR), str(DATA_INDEX_DIR)):
+        p = os.path.join(d, name)
+        if os.path.isfile(p):
+            try:
+                return json.load(open(p, encoding='utf-8'))
+            except (json.JSONDecodeError, OSError):
+                pass
+    return {}
+
+
+# 状态推断规则 (按优先级, 先匹配先用)
+STATE_RULES = [
+    ('query', ['get', 'read', 'list', 'page', 'export', 'detail', 'info', 'count']),
+    ('create', ['post', 'add', 'create', 'save', 'submit', 'apply', 'import', 'write']),
+    ('audit', ['put', 'update', 'edit', 'audit', 'approve', 'reject', 'review', 'confirm', 'send']),
+    ('cancel', ['delete', 'remove', 'cancel', 'del', 'revoke', 'close']),
+]
+# 状态排序 (业务流程顺序)
+STATE_ORDER = {'query': 1, 'create': 2, 'audit': 3, 'cancel': 4}
+
+
+def infer_state(verb, endpoint):
+    """从 HTTP verb + endpoint 路径推断业务状态。"""
+    key = ((verb or '') + ' ' + (endpoint or '')).lower()
+    for state, patterns in STATE_RULES:
+        for p in patterns:
+            if p in key:
+                return state
+    return 'other'
+
+
+def build():
+    er = load('entity-registry.json')
+    tc = load('trace-chain.json')
+
+    # module → [(state, button, endpoint)] 从 feature_adjacency + traces 推导
+    # feature_adjacency 已有 endpoints + buttons, 配合 trace 的 verb 推断 state
+    fa = er.get('feature_adjacency', {})
+    entities = er.get('entities', {})
+
+    # 建 endpoint → (verb, button) 映射 (从 traces)
+    ep_info = {}  # endpoint -> {verb, buttons:set}
+    for t in tc.get('traces', []):
+        ep = t.get('endpoint', '')
+        if not ep:
+            continue
+        info = ep_info.setdefault(ep, {'verb': '', 'buttons': set()})
+        if t.get('verb') and not info['verb']:
+            info['verb'] = t.get('verb', '')
+        if t.get('button'):
+            info['buttons'].add(t.get('button'))
+
+    # 按 kg_duckdb.import_workflow_layer 期望的结构组装:
+    # {module: {chain: [{state, operations: [{button, endpoint}]}]}}
+    workflows = {}
+    module_states = {}
+    total_steps = 0
+    for feat_eid, adj in fa.items():
+        if not isinstance(adj, dict):
+            continue
+        ent = entities.get(feat_eid, {})
+        module = ent.get('canonical', feat_eid.replace('F:', ''))
+        eps = adj.get('endpoints', [])
+        if not eps:
+            continue
+        # 按 state 分组端点
+        state_ops = {}  # state -> [{button, endpoint}]
+        for ep_eid in eps:
+            ep = ep_eid.replace('E:', '', 1)
+            info = ep_info.get(ep, {'verb': '', 'buttons': set()})
+            state = infer_state(info['verb'], ep)
+            btn = next(iter(info['buttons']), '') if info['buttons'] else ''
+            state_ops.setdefault(state, []).append({'button': btn, 'endpoint': ep})
+        # 按 STATE_ORDER 排序成 chain
+        chain = []
+        for state in sorted(state_ops.keys(), key=lambda s: STATE_ORDER.get(s, 5)):
+            chain.append({'state': state, 'operations': state_ops[state]})
+            total_steps += len(state_ops[state])
+        if chain:
+            workflows[module] = {'chain': chain}
+            module_states[module] = len(state_ops)
+
+    result = {
+        'version': '1.0',
+        'updated': datetime.now().isoformat(),
+        'workflows': workflows,
+        'stats': {
+            'modules': len(module_states),
+            'flow_steps': total_steps,
+            'modules_with_multi_state': sum(1 for n in module_states.values() if n > 1),
+        },
+    }
+    return result
+
+
+def main():
+    print('Building workflows...')
+    wf = build()
+    out = os.path.join(str(BUILD_CACHE_DIR), 'workflow-index.json')
+    os.makedirs(os.path.dirname(out), exist_ok=True)
+    with open(out, 'w', encoding='utf-8') as f:
+        json.dump(wf, f, ensure_ascii=False, indent=2, sort_keys=True)
+    s = wf['stats']
+    print('  modules: %d (%d 有多状态流转)' % (s['modules'], s['modules_with_multi_state']))
+    print('  flow steps: %d' % s['flow_steps'])
+    print('  Saved to: %s' % out)
+
+
+if __name__ == '__main__':
+    main()

package/templates/qoder/scripts/{context_pack.py → kg/context_pack.py}

@@ -1,5 +1,11 @@
 #!/usr/bin/env python3
 # -*- 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()
+
 """
 QODER Context Pack - 一次调用返回写 PRD/原型所需的全部上下文
 
@@ -34,7 +40,7 @@ except (AttributeError, TypeError, OSError, IOError):
         pass
 
 sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
-from common.terms import CN_MAP, BUSINESS_PATH_MAP, CN_TO_EN, get_platform_map
+from common.terms import CN_MAP, BUSINESS_PATH_MAP, CN_TO_EN, get_platform_map, expand_chinese_query
 
 BASE = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
 INDEX_DIR = os.path.join(BASE, 'data', 'index')
@@ -62,11 +68,14 @@ def load(filename):
 
 
 def expand_query(query):
-    words = []
-    if query in CN_MAP:
-        words.extend(CN_MAP[query].split())
-    if query in BUSINESS_PATH_MAP:
-        words.extend(p.strip('/-').lower() for p in BUSINESS_PATH_MAP[query])
+    """词扩展: 委托给 expand_chinese_query (三级匹配: 精确/分词/兜底)。
+
+    旧版只查精确匹配, 复合词如 "保险单OCR" / "异常记录" 查不到就 0 命中。
+    现在用 expand_chinese_query, 能拆 "保险单OCR" → "保险" + "ocr",
+    "异常记录" → "异常" + "记录", 大幅减少单词查询的 0 命中。
+    """
+    words = expand_chinese_query(query)
+    # 兜底: 拆不出词就用原词小写 (让英文索引做子串匹配)
     return sorted(set(w.lower() for w in words)) or [query.lower()]
 
 
@@ -315,10 +324,9 @@ def _compute_and_print(query, platform, page_type):
 
     # ── 8. 数据库表结构 (从 kg_db.duckdb, 合并了原来要单独调 MCP 的查询) ──
     try:
-        import duckdb as _ddb2
-        db_path = os.path.join(INDEX_DIR, 'kg_db.duckdb')
-        if os.path.isfile(db_path):
-            con = _ddb2.connect(db_path, read_only=True)
+        from search_index import _get_kg_db
+        con = _get_kg_db()  # 惰性缓存的只读连接, 不在此 close (供后续调用复用)
+        if con is not None:
             # 关键词匹配表名 (英文词或snake_case, camelCase自动转)
             kw_lower = query.lower()
             # camelCase → snake_case (qualityCase → quality_case)
@@ -349,7 +357,6 @@ def _compute_and_print(query, platform, page_type):
                     if ents:
                         print('  {} → Entity: {}'.format(
                             tn, ', '.join(e[0] for e in ents)))
-            con.close()
     except Exception:
         pass  # kg_db不存在/被占用 → 跳过, 不影响其它段