sycommon-python-lib 0.2.5a6__tar.gz → 0.2.5a8__tar.gz

{sycommon_python_lib-0.2.5a6 → sycommon_python_lib-0.2.5a8}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sycommon-python-lib
-Version: 0.2.5a6
+Version: 0.2.5a8
 Summary: Add your description here
 Requires-Python: >=3.11
 Description-Content-Type: text/markdown

{sycommon_python_lib-0.2.5a6 → sycommon_python_lib-0.2.5a8}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "sycommon-python-lib"
-version = "0.2.5a6"
+version = "0.2.5a8"
 description = "Add your description here"
 readme = "README.md"
 requires-python = ">=3.11"

{sycommon_python_lib-0.2.5a6 → sycommon_python_lib-0.2.5a8}/src/sycommon/agent/mcp/tool_loader.py

@@ -73,7 +73,7 @@ def wrap_tool_with_error_handler(
 
 
 def _make_user_id_interceptor(user_id: str):
-    """创建 MCP 工具调用拦截器，注入 X-User-Id 和 X-OA-Session-Id header"""
+    """创建 MCP 工具调用拦截器，注入 X-User-Id header"""
     from langchain_mcp_adapters.interceptors import MCPToolCallRequest, MCPToolCallResult
     from typing import Callable, Awaitable
 
@@ -83,16 +83,6 @@ def _make_user_id_interceptor(user_id: str):
     ) -> MCPToolCallResult:
         headers = dict(request.headers or {})
         headers["X-User-Id"] = user_id
-
-        # 自动获取 OA session_id 注入 header
-        try:
-            from sycommon.auth.oa_cache import oa_login_with_cache
-            success, _, session_id = await oa_login_with_cache(user_id)
-            if success and session_id:
-                headers["X-OA-Session-Id"] = session_id
-        except Exception:
-            pass
-
         request = request.override(headers=headers)
         return await handler(request)
 
@@ -132,7 +122,6 @@ async def load_mcp_tools(
     for config in enabled_configs:
         try:
             from langchain_mcp_adapters.client import MultiServerMCPClient
-            from langchain_mcp_adapters.tools import load_mcp_tools as _load_tools
 
             key = sanitize_name(config.name)
             client_config = {
@@ -144,18 +133,16 @@ async def load_mcp_tools(
             }
 
             client = MultiServerMCPClient(client_config)
-            session = await client.session_for(key)
-            await session.initialize()
 
-            from langchain_mcp_adapters.tools import _list_all_tools
-            raw_tools = await _list_all_tools(session)
+            async with client.session(key) as session:
+                from langchain_mcp_adapters.tools import _list_all_tools
+                raw_tools = await _list_all_tools(session)
 
             original_names = []
             for tool in raw_tools:
                 original_name = tool.name
                 original_names.append(original_name)
 
-                # 使用 convert_mcp_tool_to_langchain_tool 带 interceptors
                 from langchain_mcp_adapters.tools import convert_mcp_tool_to_langchain_tool
                 lc_tool = convert_mcp_tool_to_langchain_tool(
                     session,

{sycommon_python_lib-0.2.5a6 → sycommon_python_lib-0.2.5a8}/src/sycommon/auth/wecom_ldap_service.py

@@ -3,7 +3,7 @@
 
 通过企微 userid 关联 LDAP 域账号，获取邮箱、部门、职位等完整信息。
 
-关联策略：中文名优先 + 英文名兜底，匹配率约 95.8%。
+关联策略：中文名 → 英文名+中文姓 → 英文名前缀，匹配率约 98%。
 
 使用示例::
 
@@ -87,18 +87,21 @@ class WeComLDAPUser(BaseModel):
 
 
 class _LDAPIndex:
-    """LDAP 用户索引（按 cn 和 sAMAccountName 前缀建索引）"""
+    """LDAP 用户索引（按 cn、英文名+姓拼音、sAMAccountName 前缀建索引）"""
 
     def __init__(self):
         # cn(中文名) -> LDAPFullUser
         self.by_cn: Dict[str, LDAPFullUser] = {}
-        # sAMAccountName 前缀(小写) -> [cn]
-        self.by_sam_prefix: Dict[str, List[str]] = defaultdict(list)
+        # sAMAccountName 前缀(小写) -> [LDAPFullUser]
+        self.by_sam_prefix: Dict[str, List[LDAPFullUser]] = defaultdict(list)
+        # "en_prefix.sn_pinyin" -> [LDAPFullUser]
+        self.by_en_sn: Dict[str, List[LDAPFullUser]] = defaultdict(list)
 
     def build(self, users: List[dict]):
         """从 LDAP 查询结果构建索引"""
         self.by_cn.clear()
         self.by_sam_prefix.clear()
+        self.by_en_sn.clear()
 
         for u in users:
             cn = u.get("cn", "").strip()
@@ -126,14 +129,31 @@ class _LDAPIndex:
 
             # sAMAccountName 如 "sussie.xia" -> 前缀 "sussie"
             prefix = sam.split(".")[0].lower()
-            self.by_sam_prefix[prefix].append(cn)
+            self.by_sam_prefix[prefix].append(ldap_user)
 
-    def lookup(self, chinese_name: str, english_name: str = "") -> tuple:
+            # 英文名+姓(拼音)索引: "sussie.xiao"
+            sn = u.get("sn", "").strip()
+            if prefix and sn:
+                sn_pinyin = self._to_pinyin(sn)
+                if sn_pinyin:
+                    self.by_en_sn[f"{prefix}.{sn_pinyin}"].append(ldap_user)
+
+    @staticmethod
+    def _to_pinyin(text: str) -> str:
+        """中文转拼音（小写）"""
+        try:
+            from pypinyin import pinyin, Style
+            return "".join(p[0] for p in pinyin(text, style=Style.NORMAL)).lower()
+        except Exception:
+            return ""
+
+    def lookup(self, chinese_name: str, english_name: str = "", surname: str = "", position: str = "") -> tuple:
         """查找匹配的 LDAP 用户
 
         Args:
-            chinese_name: 中文名
-            english_name: 英文名（来自企微 extattr）
+            chinese_name: 中文名（企微 name）
+            english_name: 英文名（企微 extattr.英文名）
+            surname: 中文姓（取 chinese_name 首字）
 
         Returns:
             (LDAPFullUser | None, match_method | None)
@@ -142,16 +162,29 @@ class _LDAPIndex:
         if chinese_name in self.by_cn:
             return self.by_cn[chinese_name], "chinese_name"
 
-        # 兜底：英文名前缀匹配
-        if english_name:
-            prefix = english_name.lower().split()[0]
-            candidates = self.by_sam_prefix.get(prefix, [])
+        en_first = english_name.lower().split()[0] if english_name else ""
+
+        # 第二优先级：英文名 + 中文姓拼音
+        if en_first and surname:
+            surname_pinyin = self._to_pinyin(surname)
+            if surname_pinyin:
+                key = f"{en_first}.{surname_pinyin}"
+                candidates = self.by_en_sn.get(key, [])
+                if len(candidates) == 1:
+                    return candidates[0], "english_name_surname"
+                elif len(candidates) > 1:
+                    pass
+
+        # 第三优先级：英文名前缀匹配（唯一）
+        if en_first:
+            candidates = self.by_sam_prefix.get(en_first, [])
             if len(candidates) == 1:
-                # 唯一匹配，直接返回
-                return self.by_cn[candidates[0]], "english_name"
-            elif len(candidates) > 1:
-                # 重名，无法确定，不匹配
-                pass
+                return candidates[0], "english_name"
+            elif len(candidates) > 1 and position:
+                # 多候选人时用职位消歧
+                pos_hits = [c for c in candidates if c.title and c.title in position]
+                if len(pos_hits) == 1:
+                    return pos_hits[0], "english_name_position"
 
         return None, None
 
@@ -297,6 +330,7 @@ class WeComLDAPService(metaclass=SingletonMeta):
                         attributes=[
                             "cn",
                             "sAMAccountName",
+                            "sn",
                             "mail",
                             "department",
                             "title",
@@ -320,6 +354,7 @@ class WeComLDAPService(metaclass=SingletonMeta):
                             {
                                 "cn": cn,
                                 "sAMAccountName": sam,
+                                "sn": str(entry.sn).strip() if entry.sn else "",
                                 "mail": str(entry.mail).strip()
                                 if entry.mail
                                 else "",
@@ -389,7 +424,15 @@ class WeComLDAPService(metaclass=SingletonMeta):
         english_name = self._parse_wecom_english_name(user_data)
         chinese_name = user_data.get("name", "")
 
-        ldap_user, match_method = ldap_index.lookup(chinese_name, english_name)
+        # 提取中文姓（取中文名首字）
+        import re
+        surname = ""
+        if chinese_name and re.search(r'[一-鿿]', chinese_name[0]):
+            surname = chinese_name[0]
+
+        ldap_user, match_method = ldap_index.lookup(
+            chinese_name, english_name, surname, user_data.get("position", "")
+        )
 
         return WeComLDAPUser(
             wecom_userid=user_data.get("userid", ""),

{sycommon_python_lib-0.2.5a6 → sycommon_python_lib-0.2.5a8}/src/sycommon/llm/struct_token.py

@@ -10,10 +10,33 @@ from sycommon.llm.llm_logger import LLMLogger
 from langfuse import Langfuse, LangfuseSpan, propagate_attributes
 from typing import Dict, List, Optional, Any
 import tiktoken
-import threading
+import os
 import warnings
 warnings.filterwarnings("ignore", message="Pydantic serializer warnings")
 
+# 模块级 encoding 单例，所有实例共享，只初始化一次
+_cached_encoding = None
+_encoding_initialized = False
+
+
+def _get_shared_encoding(model_name: str):
+    """获取模块级共享的 tiktoken encoding，只初始化一次"""
+    global _cached_encoding, _encoding_initialized
+    if _encoding_initialized:
+        return _cached_encoding
+
+    _encoding_initialized = True
+    try:
+        _cached_encoding = tiktoken.encoding_for_model(model_name)
+    except KeyError:
+        try:
+            _cached_encoding = tiktoken.get_encoding("cl100k_base")
+        except Exception as e:
+            SYLogger.error(f"tiktoken 初始化失败，将使用字符数近似估算: {e}")
+    except Exception as e:
+        SYLogger.error(f"tiktoken 初始化失败，将使用字符数近似估算: {e}")
+    return _cached_encoding
+
 
 class StructuredRunnableWithToken(Runnable):
     """
@@ -49,29 +72,8 @@ class StructuredRunnableWithToken(Runnable):
         self.max_compression_attempts = max_compression_attempts  # 最大压缩尝试次数
         self.is_native_mode = is_native_mode  # 是否为原生模式
 
-        # 初始化 Tokenizer（5秒超时保护）
-        self.encoding = None
-
-        def _init_encoding():
-            nonlocal encoding_result
-            try:
-                encoding_result = tiktoken.encoding_for_model(model_name)
-            except KeyError:
-                try:
-                    encoding_result = tiktoken.get_encoding("cl100k_base")
-                except Exception as e:
-                    SYLogger.error(f"tiktoken get_encoding 失败: {e}")
-            except Exception as e:
-                SYLogger.error(f"tiktoken encoding_for_model 失败: {e}")
-
-        encoding_result = None
-        t = threading.Thread(target=_init_encoding, daemon=True)
-        t.start()
-        t.join(timeout=5)
-        if t.is_alive():
-            SYLogger.error("tiktoken 初始化超时(5s)，将使用字符数近似估算")
-        elif encoding_result is not None:
-            self.encoding = encoding_result
+        # 初始化 Tokenizer（模块级单例，只初始化一次）
+        self.encoding = _get_shared_encoding(model_name)
 
     def _encode_text(self, text: str) -> int:
         """计算文本的 token 数量，encoding 不可用时按字符数近似估算"""

{sycommon_python_lib-0.2.5a6 → sycommon_python_lib-0.2.5a8}/src/sycommon_python_lib.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sycommon-python-lib
-Version: 0.2.5a6
+Version: 0.2.5a8
 Summary: Add your description here
 Requires-Python: >=3.11
 Description-Content-Type: text/markdown