@hupan56/wlkj 3.1.32 → 3.3.0

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

@@ -9,7 +9,7 @@ for _ in range(10):
         break
 if _cp not in _s.path: _s.path.insert(0, _cp)
 from bootstrap import setup; setup()
-from foundation.core.paths import get_repo_root
+from foundation.core.paths import get_repo_root
 
 """
 kg.py — QODER 知识图谱统一 CLI 入口（全系列通用）
@@ -46,8 +46,8 @@ try:
 except Exception:
     pass
 
-SCRIPTS_DIR = get_repo_root()
-BASE_DIR = get_repo_root()
+SCRIPTS_DIR = get_repo_root()
+BASE_DIR = get_repo_root()
 sys.path.insert(0, SCRIPTS_DIR)
 _KG_DIR = os.path.dirname(os.path.abspath(__file__))
 if _KG_DIR not in sys.path: sys.path.insert(0, _KG_DIR)  # 同包import
@@ -69,8 +69,10 @@ def _kg_db_path():
 
 
 def _has_duckdb():
-    """知识图谱 DuckDB 是否已构建。"""
-    return os.path.isfile(_kg_db_path())
+    """知识图谱本地 DuckDB 是否已构建。
+    ★ 全迁 PG 后本地不再有 DuckDB——这个函数返回 False，让所有旧路径走 MCP 远程。
+    """
+    return False  # 本地零 DuckDB，全部走 MCP → 工作台 PG
 
 
 def _hint_rebuild(action_cn='知识图谱'):
@@ -821,6 +823,36 @@ def _try_daemon(cmd, rest):
     return True
 
 
+def _undo_msys_pathconv(arg):
+    """Undo Git Bash (MSYS2) path conversion on a single argument.
+
+    On Windows Git Bash, a leading-slash token like '/insurance' is
+    auto-rewritten to '<git-install>/insurance' (e.g. D:/programming/tools/
+    Git/Git/insurance) before Python ever sees argv. kg.py endpoints start
+    with '/', so every '/xxx' query is corrupted in that shell.
+
+    Detection: arg looks like an absolute Windows path (X:/...) AND contains
+    '/Git/' — then the real intent was '/<tail>'. Safe no-op for genuine paths
+    (file commands pass real paths that rarely contain /Git/, and the worst
+    case is a real path stays unchanged).
+    """
+    if not arg or len(arg) < 2 or arg[1] != ':':
+        return arg
+    # e.g. D:/programming/tools/Git/Git/insurance -> /insurance
+    # Use rfind: the install path itself may contain "Git" (e.g. .../Git/Git/),
+    # so the LAST /Git/ is the one MSYS inserts.
+    marker = '/Git/'
+    idx = arg.rfind(marker)
+    if idx < 0:
+        return arg
+    tail = arg[idx + len(marker):]
+    # Only restore when the tail looks like an endpoint fragment (no drive,
+    # no '.ext' that would indicate a real file path).
+    if ':' in tail or ('.' in tail and tail.rfind('.') > tail.rfind('/')):
+        return arg
+    return '/' + tail
+
+
 def main(argv=None):
     argv = argv if argv is not None else sys.argv[1:]
 
@@ -830,6 +862,11 @@ def main(argv=None):
         use_daemon = False
         argv = [a for a in argv if a != '--no-daemon']
 
+    # Undo MSYS path conversion on arguments (Git Bash rewrites '/xxx' →
+    # '<git-install>/xxx', which corrupts endpoint queries like /insurance).
+    if len(argv) > 1:
+        argv[1:] = [_undo_msys_pathconv(a) for a in argv[1:]]
+
     if not argv or argv[0] in ('-h', '--help', 'help'):
         print(_usage())
         return 0

package/templates/qoder/scripts/domain/kg/search/_remote.py

@@ -0,0 +1,187 @@
+"""本地引擎 → 工作台 MCP 远程查询层。
+
+替代直连本地 kg.duckdb（已删），所有知识查询走 MCP HTTP → 工作台 PostgreSQL。
+下游命令（kg.py / wl-search / context_pack）零改动：函数签名不变，print 格式不变。
+
+用法（search_index.py 内）：
+    from domain.kg.search._remote import call_remote, render_search_code
+    data = call_remote("search_code", {"keyword": query})
+    sys.stdout.write(render_search_code(data))
+"""
+import json
+import sys
+import os
+
+# McpCap 路径（相对于 .qoder/scripts/）
+_scripts = os.path.dirname(os.path.dirname(os.path.dirname(os.path.dirname(__file__))))
+if _scripts not in sys.path:
+    sys.path.insert(0, _scripts)
+
+_mcp = None
+
+
+def _get_mcp():
+    """惰性初始化 McpCap（首次调用连工作台 MCP server）。"""
+    global _mcp
+    if _mcp is None:
+        from capability.adapters.mcp import McpCap
+        _mcp = McpCap()
+    return _mcp
+
+
+def call_remote(tool, args):
+    """调工作台 MCP 工具，返回 dict。失败抛 RuntimeError（无本地降级，kg.duckdb 已删）。"""
+    cap = _get_mcp()
+    result = cap.call(tool, args)
+    if result.is_error:
+        raise RuntimeError("MCP工具 %s 调用失败: %s" % (tool, getattr(result, 'text', '')[:200]))
+    try:
+        data = json.loads(result.text)
+    except Exception:
+        raise RuntimeError("MCP返回非JSON: %s" % result.text[:200])
+    if isinstance(data, dict) and data.get("degraded") and data.get("error"):
+        raise RuntimeError("工作台降级: %s" % data["error"])
+    return data
+
+
+# ============================================================
+# 渲染器：把 MCP JSON → 原本地 print 文本格式（下游命令零改动）
+# ============================================================
+
+def render_search_code(data):
+    """渲染 search_code 结果（仿 search_index.py _search_keywords_impl 格式）。"""
+    items = data.get("items", [])
+    if not items:
+        return "No matches.\n"
+    lines = ["[%d matches]" % len(items)]
+    for it in items[:20]:
+        t = it.get("type", "")
+        name = it.get("name", "")
+        label = it.get("label", "")
+        f = it.get("file", "")
+        plat = it.get("platform", "")
+        short = f.replace("\\", "/").split("/")[-1] if f else ""
+        suffix = "  ->  data/code/%s" % f.replace("\\", "/") if f else ""
+        line = "  [%s] %s" % (t, name)
+        if label and label != name:
+            line += "（%s）" % label
+        if f:
+            line += "  %s%s" % (short, suffix)
+        lines.append(line)
+    return "\n".join(lines) + "\n"
+
+
+def render_search_api(data):
+    """渲染 search_api 结果。"""
+    items = data.get("items", [])
+    if not items:
+        return "No API matches.\n"
+    lines = ["[%d API matches]" % len(items)]
+    for it in items[:20]:
+        verb = (it.get("verb") or "").upper()
+        url = it.get("url", "")
+        fn = it.get("fn", "")
+        lines.append("  %s %s -> %s" % (verb or "GET", url, fn))
+    return "\n".join(lines) + "\n"
+
+
+def render_search_field(data):
+    """渲染 search_field 结果。"""
+    items = data.get("items", [])
+    if not items:
+        return "No field matches.\n"
+    lines = ["[%d field matches]" % len(items)]
+    for it in items[:20]:
+        field = it.get("field", "")
+        title = it.get("title", "")
+        count = it.get("count", 0)
+        lines.append("  %s = %s (用%d次)" % (field, title, count))
+    return "\n".join(lines) + "\n"
+
+
+def render_search_prds(data):
+    """渲染 PRD 搜索结果。"""
+    items = data.get("items", [])
+    if not items:
+        return "No PRD matches.\n"
+    lines = ["[%d PRD matches]" % len(items)]
+    for it in items[:10]:
+        title = it.get("title", "")
+        status = it.get("status", "")
+        lines.append("  [%s] %s" % (status, title))
+    return "\n".join(lines) + "\n"
+
+
+def render_search_style(data):
+    """渲染风格搜索结果。"""
+    items = data.get("items", [])
+    if not items:
+        return "No style matches.\n"
+    lines = ["[%d style matches]" % len(items)]
+    for it in items[:10]:
+        lines.append("  %s: %s" % (it.get("title", ""), it.get("description", "")[:60]))
+    return "\n".join(lines) + "\n"
+
+
+def render_impact(data):
+    """渲染影响分析结果。"""
+    entity = data.get("entity")
+    if not entity:
+        return "Entity not found.\n"
+    totals = data.get("totals", {})
+    by_type = data.get("byType", {})
+    lines = [
+        "影响分析: %s (%s)" % (entity.get("name", ""), entity.get("type", "")),
+        "  影响范围: %s 节点 / %s 关系" % (totals.get("nodes", 0), totals.get("links", 0)),
+    ]
+    if by_type:
+        lines.append("  按类型: " + ", ".join("%s·%d" % (k, v) for k, v in by_type.items()))
+    for section in [("upstream", "⬆ 谁依赖它"), ("downstream", "⬇ 它依赖谁")]:
+        key, label = section
+        layers = data.get(key, [])
+        if layers:
+            lines.append(label)
+            for layer in layers:
+                depth = layer.get("depth", "?")
+                items = layer.get("items", [])
+                names = [it.get("name", "")[:30] for it in items[:5]]
+                lines.append("  L%s: %s" % (depth, ", ".join(names)))
+                if len(items) > 5:
+                    lines.append("       +%d more" % (len(items) - 5))
+    return "\n".join(lines) + "\n"
+
+
+def render_context360(data):
+    """渲染 context_360 结果。"""
+    parts = []
+    for key, label in [("code", "CODE"), ("api", "API"), ("field", "FIELD")]:
+        section = data.get(key, {})
+        # context_360 返回 {code: {items:[...], total:N}, ...} 不是直接 list
+        items = section.get("items", []) if isinstance(section, dict) else (section if isinstance(section, list) else [])
+        if items:
+            parts.append("[%s] %d matches" % (label, len(items)))
+            for it in items[:5]:
+                if isinstance(it, dict):
+                    name = it.get("name", it.get("url", it.get("field", "")))
+                    t = it.get("type", "")
+                    extra = it.get("title", it.get("label", ""))
+                    line = "  %s %s" % (t, name) if t else "  %s" % name
+                    if extra and extra != name:
+                        line += "（%s）" % extra
+                    parts.append(line)
+    if not parts:
+        return "No context found.\n"
+    return "\n".join(parts) + "\n"
+
+
+def render_rag_search(data):
+    """渲染 RAG 语义检索结果。"""
+    items = data.get("items", [])
+    if not items:
+        return "No semantic matches.\n"
+    lines = ["[%d semantic matches] vec=%s" % (len(items), data.get("vector_ok"))]
+    for it in items[:10]:
+        name = it.get("name", "")
+        score = it.get("vec_score", 0)
+        lines.append("  [%.2f] %s (%s)" % (score, name, it.get("type", "")))
+    return "\n".join(lines) + "\n"

package/templates/qoder/scripts/domain/kg/search/context_pack.py

@@ -198,8 +198,38 @@ def main():
 
 
 def _compute_and_print(query, platform, page_type):
-    """实际的 context_pack 计算 (原 main 逻辑)。"""
-
+    """实际的 context_pack 计算 → 走工作台 MCP context_pack（本地 JSON 索引已废弃）。"""
+    try:
+        from domain.kg.search._remote import call_remote
+        data = call_remote("context_pack", {"query": query, "platform": platform})
+        # MCP context_pack 返回聚合结果，直接渲染
+        print('# Context Pack: {} (platform: {})'.format(query, platform or 'all'))
+        for section_key, section_label in [("code", "相关代码"), ("api", "API"),
+                                            ("field", "字段"), ("wiki", "Wiki"),
+                                            ("prd", "PRD")]:
+            items = data.get(section_key, [])
+            if not items:
+                continue
+            print('\n## %s (%d)' % (section_label, len(items)))
+            for it in items[:8]:
+                if isinstance(it, dict):
+                    name = it.get("name", it.get("url", it.get("field", "")))
+                    label = it.get("label", it.get("title", ""))
+                    line = "- %s" % name
+                    if label and label != name:
+                        line += "（%s）" % label
+                    print(line)
+        print("\n（数据来自工作台 PostgreSQL，MCP context_pack 工具）")
+        return
+    except Exception as e:
+        print("# Context Pack: MCP 调用失败（工作台未启动？）")
+        print("# 错误: %s" % str(e)[:200])
+        print("# 提示: 确保工作台后端在跑（start_pg.bat），mcp_config.json 配置正确")
+        return
+
+
+def _compute_and_print_local(query, platform, page_type):
+    """[已废弃] 原本地 JSON 索引逻辑（kg.duckdb 已删，load_index 会失败）。"""
     targets = []
     if platform and platform.lower() not in ('both', '两端'):
         targets = [PLATFORM_MAP.get(platform.lower(), platform)]

package/templates/qoder/scripts/domain/kg/search/search_index.py

@@ -413,6 +413,22 @@ def _check_index_freshness():
 
 
 def search_keywords(query, platform=None):
+    """搜代码关键词 → 走工作台 MCP search_code（本地 kg.duckdb 已删，不再直连）。"""
+    try:
+        from domain.kg.search._remote import call_remote, render_search_code
+        data = call_remote("search_code", {"keyword": query, "platform": platform})
+        sys.stdout.write(render_search_code(data))
+        return
+    except Exception as e:
+        print("MCP search_code 失败（工作台未启动？）: %s" % str(e)[:200])
+        print("提示: 确保工作台后端在跑（start_pg.bat），mcp_config.json 配置正确")
+        return
+
+
+def search_keywords_local(query, platform=None):
+    """[已废弃] 原本地直连 kg.duckdb 的搜索逻辑。保留仅供回退参考。
+    本地 kg.duckdb 已删除，此函数会报 code-keyword.json not found。
+    """
     # 先查缓存 (命中即数据未变, 直接还)。缓存 key 已含 kg.duckdb mtime, 命中就说明
     # 数据源没变, 跳过 _check_index_freshness 的 os.walk —— 50 人团队每天几十次
     # 搜索, 每次都 walk 目录是纯浪费。未命中才做新鲜度检查 + 计算。
@@ -581,6 +597,17 @@ def _semantic_fallback(query, platform=None):
 
 
 def search_prds(query):
+    """搜 PRD → 走工作台 MCP search_prds。"""
+    try:
+        from domain.kg.search._remote import call_remote, render_search_prds
+        data = call_remote("search_prds", {"query": query})
+        sys.stdout.write(render_search_prds(data))
+    except Exception as e:
+        print("MCP search_prds 失败: %s" % str(e)[:200])
+
+
+def search_prds_local(query):
+    """[已废弃] 原本地逻辑。"""
     prd_index = load_index('prd-index.json', hint='Run: python .qoder/scripts/domain/task/git_sync.py --prd-only')
     if prd_index is None:
         return
@@ -627,7 +654,17 @@ def search_prds(query):
 
 
 def search_style(style_type, platform=None):
-    """Find pages of a given UI type (table/form/detail/modal/dashboard)."""
+    """搜风格 → 走工作台 MCP search_style。"""
+    try:
+        from domain.kg.search._remote import call_remote, render_search_style
+        data = call_remote("search_style", {"style_type": style_type, "platform": platform})
+        sys.stdout.write(render_search_style(data))
+    except Exception as e:
+        print("MCP search_style 失败: %s" % str(e)[:200])
+
+
+def search_style_local(style_type, platform=None):
+    """[已废弃] 原本地逻辑。"""
     si = load_index('ui-style.json', hint='Run: python build_style_index.py')
     if si is None:
         return
@@ -663,7 +700,17 @@ def search_style(style_type, platform=None):
 
 
 def search_field(field_name):
-    """Search field usage in ui-style.json field_map."""
+    """搜字段 → 走工作台 MCP search_field。"""
+    try:
+        from domain.kg.search._remote import call_remote, render_search_field
+        data = call_remote("search_field", {"name": field_name})
+        sys.stdout.write(render_search_field(data))
+    except Exception as e:
+        print("MCP search_field 失败: %s" % str(e)[:200])
+
+
+def search_field_local(field_name):
+    """[已废弃] 原本地逻辑。"""
     si = load_index('ui-style.json', hint='Run: python build_style_index.py')
     if si is None:
         return
@@ -681,7 +728,17 @@ def search_field(field_name):
 
 
 def search_api(query):
-    """Search API endpoints in code-api.json."""
+    """搜 API → 走工作台 MCP search_api。"""
+    try:
+        from domain.kg.search._remote import call_remote, render_search_api
+        data = call_remote("search_api", {"keyword": query})
+        sys.stdout.write(render_search_api(data))
+    except Exception as e:
+        print("MCP search_api 失败: %s" % str(e)[:200])
+
+
+def search_api_local(query):
+    """[已废弃] 原本地逻辑。"""
     ai = load_index('code-api.json', hint='Run: python .qoder/scripts/domain/task/git_sync.py --index-only')
     if ai is None:
         return
@@ -813,26 +870,23 @@ def search_wiki_cli(query):
 
 
 def search_impact(query):
-    """改这个接口/符号影响哪些页面/按钮/模块? (U1: 内存图遍历)"""
-    from domain.kg.graph.graph_traverse import CodeGraph
-    g = CodeGraph(INDEX_DIR)
-    result = g.impact(query)
-    if result:
-        print(result)
-    else:
-        print('未找到匹配 "{}" 的端点。尝试用部分路径, 如 /veh 或 export'.format(query))
-        print('提示: trace-chain 当前覆盖率 31%, 未覆盖的调用不会被列出')
+    """影响分析 → 走工作台 MCP get_impact。"""
+    try:
+        from domain.kg.search._remote import call_remote, render_impact
+        data = call_remote("get_impact", {"entity": query, "depth": 2})
+        sys.stdout.write(render_impact(data))
+    except Exception as e:
+        print("MCP get_impact 失败: %s" % str(e)[:200])
 
 
 def search_context360(query):
-    """符号的360度视图: 端点/函数/处理器/PRD (U1: 内存图遍历)"""
-    from domain.kg.graph.graph_traverse import CodeGraph
-    g = CodeGraph(INDEX_DIR)
-    result = g.context360(query)
-    if result:
-        print(result)
-    else:
-        print('未找到匹配 "{}" 的符号。'.format(query))
+    """360度视图 → 走工作台 MCP context_360。"""
+    try:
+        from domain.kg.search._remote import call_remote, render_context360
+        data = call_remote("context_360", {"symbol": query})
+        sys.stdout.write(render_context360(data))
+    except Exception as e:
+        print("MCP context_360 失败: %s" % str(e)[:200])
 
 
 def print_usage():

package/templates/qoder/scripts/domain/kg/storage/kg_duckdb.py

@@ -29,6 +29,7 @@ Schema:
 import json
 import os
 import sys
+import time
 
 import duckdb
 
@@ -102,6 +103,19 @@ def init_schema(con):
             value TEXT
         )
     """)
+    # ── 文件签名表 (P0: fingerprint 三级增量) ──
+    # 记录每个被索引文件的 content_hash (全文) 和 structural_hash (去注释去空白)。
+    # 增量构建前比对: NONE=完全没变; COSMETIC=只改注释/空白(跳过重算);
+    # STRUCTURAL=真改结构(正常重算)。图谱只记拓扑不记实现 → COSMETIC 跳过 100% 安全。
+    con.execute("""
+        CREATE TABLE IF NOT EXISTS file_signatures (
+            file_path TEXT PRIMARY KEY,
+            content_hash TEXT,
+            structural_hash TEXT,
+            level TEXT,
+            updated_at TEXT
+        )
+    """)
     # ── 搜索层表 ──
     con.execute("CREATE TABLE IF NOT EXISTS keywords(keyword TEXT, file_path TEXT)")
     con.execute("CREATE TABLE IF NOT EXISTS endpoints(endpoint TEXT PRIMARY KEY, controller_file TEXT)")
@@ -182,6 +196,8 @@ def init_schema(con):
     con.execute("CREATE INDEX IF NOT EXISTS idx_alias ON aliases(alias)")
     con.execute("CREATE INDEX IF NOT EXISTS idx_edge_from ON edges(from_id)")
     con.execute("CREATE INDEX IF NOT EXISTS idx_edge_to ON edges(to_id)")
+    # P1: 复合索引加速按类型的边查询 (calls/imports 边量大, multi_hop 按类型过滤)
+    con.execute("CREATE INDEX IF NOT EXISTS idx_edge_from_type ON edges(from_id, edge_type)")
     con.execute("CREATE INDEX IF NOT EXISTS idx_kw ON keywords(keyword)")
     con.execute("CREATE INDEX IF NOT EXISTS idx_page_type ON pages(page_type)")
     con.execute("CREATE INDEX IF NOT EXISTS idx_field ON fields(field)")
@@ -192,6 +208,7 @@ def init_schema(con):
     con.execute("CREATE INDEX IF NOT EXISTS idx_learn_dev ON learning_events(dev)")
     con.execute("CREATE INDEX IF NOT EXISTS idx_learn_pat_cat ON learning_patterns(category)")
     con.execute("CREATE INDEX IF NOT EXISTS idx_wf_mod ON workflows(module)")
+    con.execute("CREATE INDEX IF NOT EXISTS idx_fsig ON file_signatures(file_path)")
 
 
 def _sql_escape(val):
@@ -244,6 +261,32 @@ def delete_edges_from(con, from_id, edge_type=None):
         con.execute("DELETE FROM edges WHERE from_id = ?", [from_id])
 
 
+def upsert_file_signature(con, file_path, content_hash, structural_hash, level):
+    """Insert or replace a file's signature (P0: fingerprint incremental).
+
+    Called after classifying a changed file; stored so the next incremental
+    run can compare and skip COSMETIC-only changes.
+    """
+    con.execute(
+        "INSERT OR REPLACE INTO file_signatures VALUES (?, ?, ?, ?, ?)",
+        [file_path, content_hash, structural_hash, level, str(int(time.time()))]
+    )
+
+
+def get_file_signature(con, file_path):
+    """Read a file's stored signature, or None if never seen."""
+    row = con.execute(
+        "SELECT content_hash, structural_hash FROM file_signatures WHERE file_path = ?",
+        [file_path]
+    ).fetchone()
+    return {'content_hash': row[0], 'structural_hash': row[1]} if row else None
+
+
+def clear_file_signatures(con):
+    """Wipe all signatures (full rebuild resets the baseline)."""
+    con.execute("DELETE FROM file_signatures")
+
+
 def resolve_entity(con, alias):
     """Resolve any alias to its canonical entity ID."""
     row = con.execute(