tamar-model-client 0.1.25__tar.gz → 0.1.27__tar.gz

{tamar_model_client-0.1.25 → tamar_model_client-0.1.27}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tamar-model-client
-Version: 0.1.25
+Version: 0.1.27
 Summary: A Python SDK for interacting with the Model Manager gRPC service
 Home-page: http://gitlab.tamaredge.top/project-tap/AgentOS/model-manager-client
 Author: Oscar Ou

{tamar_model_client-0.1.25 → tamar_model_client-0.1.27}/setup.py

@@ -2,7 +2,7 @@ from setuptools import setup, find_packages
 
 setup(
     name="tamar-model-client",
-    version="0.1.25",
+    version="0.1.27",
     description="A Python SDK for interacting with the Model Manager gRPC service",
     author="Oscar Ou",
     author_email="oscar.ou@tamaredge.ai",

{tamar_model_client-0.1.25 → tamar_model_client-0.1.27}/tamar_model_client/async_client.py

@@ -33,7 +33,9 @@ from .core import (
     generate_request_id,
     set_request_id,
     get_protected_logger,
-    MAX_MESSAGE_LENGTH, get_request_id
+    MAX_MESSAGE_LENGTH, 
+    get_request_id,
+    RequestIdManager
 )
 from .core.base_client import BaseClient
 from .core.request_builder import RequestBuilder
@@ -102,6 +104,9 @@ class AsyncTamarModelClient(BaseClient, AsyncHttpFallbackMixin):
         self._last_channel_error_time = None
         self._channel_lock = asyncio.Lock()  # 异步锁
         
+        # === Request ID 管理 ===
+        self._request_id_manager = RequestIdManager()
+        
         # === 增强的重试处理器 ===
         self.retry_handler = EnhancedRetryHandler(
             max_retries=self.max_retries,
@@ -744,10 +749,22 @@ class AsyncTamarModelClient(BaseClient, AsyncHttpFallbackMixin):
                 "user_id": model_request.user_context.user_id or ""
             }
 
-        if not request_id:
+        # 处理 request_id
+        origin_request_id = None
+        if request_id:
+            # 用户提供了 request_id，生成组合 ID
+            request_id, origin_request_id = self._request_id_manager.get_composite_id(request_id)
+        else:
+            # 没有提供，生成新的
             request_id = generate_request_id()
+            
         set_request_id(request_id)
-        metadata = self._build_auth_metadata(request_id)
+        metadata = self._build_auth_metadata(request_id, origin_request_id)
+
+        # 构建日志数据
+        log_data = ResponseHandler.build_log_data(model_request)
+        if origin_request_id:
+            log_data['origin_request_id'] = origin_request_id
 
         # 记录开始日志
         start_time = time.time()
@@ -756,7 +773,7 @@ class AsyncTamarModelClient(BaseClient, AsyncHttpFallbackMixin):
             extra={
                 "log_type": "request",
                 "uri": f"/invoke/{model_request.provider.value}/{model_request.invoke_type.value}",
-                "data": ResponseHandler.build_log_data(model_request)
+                "data": log_data
             })
 
         try:
@@ -794,13 +811,19 @@ class AsyncTamarModelClient(BaseClient, AsyncHttpFallbackMixin):
                 # 记录非流式响应的成功日志
                 duration = time.time() - start_time
                 content_length = len(result.content) if result.content else 0
+                
+                # 构建响应日志数据
+                response_log_data = ResponseHandler.build_log_data(model_request, result)
+                if origin_request_id:
+                    response_log_data['origin_request_id'] = origin_request_id
+                    
                 logger.info(
                     f"✅ Request completed | content_length: {content_length}",
                     extra={
                         "log_type": "response",
                         "uri": f"/invoke/{model_request.provider.value}/{model_request.invoke_type.value}",
                         "duration": duration,
-                        "data": ResponseHandler.build_log_data(model_request, result)
+                        "data": response_log_data
                     }
                 )
                 
@@ -813,15 +836,18 @@ class AsyncTamarModelClient(BaseClient, AsyncHttpFallbackMixin):
         except (ConnectionError, grpc.RpcError) as e:
             duration = time.time() - start_time
             error_message = f"❌ Invoke gRPC failed: {str(e)}"
+            
+            # 构建错误日志数据
+            error_log_data = ResponseHandler.build_log_data(model_request, error=e)
+            if origin_request_id:
+                error_log_data['origin_request_id'] = origin_request_id
+                
             logger.error(error_message, exc_info=True,
                          extra={
                              "log_type": "response",
                              "uri": f"/invoke/{model_request.provider.value}/{model_request.invoke_type.value}",
                              "duration": duration,
-                             "data": ResponseHandler.build_log_data(
-                                 model_request,
-                                 error=e
-                             )
+                             "data": error_log_data
                          })
             
             # 记录 channel 错误
@@ -875,10 +901,27 @@ class AsyncTamarModelClient(BaseClient, AsyncHttpFallbackMixin):
                 "user_id": batch_request_model.user_context.user_id or ""
             }
 
-        if not request_id:
+        # 处理 request_id
+        origin_request_id = None
+        if request_id:
+            # 用户提供了 request_id，生成组合 ID
+            request_id, origin_request_id = self._request_id_manager.get_composite_id(request_id)
+        else:
+            # 没有提供，生成新的
             request_id = generate_request_id()
+            
         set_request_id(request_id)
-        metadata = self._build_auth_metadata(request_id)
+        metadata = self._build_auth_metadata(request_id, origin_request_id)
+
+        # 构建日志数据
+        batch_log_data = {
+            "batch_size": len(batch_request_model.items),
+            "org_id": batch_request_model.user_context.org_id,
+            "user_id": batch_request_model.user_context.user_id,
+            "client_type": batch_request_model.user_context.client_type
+        }
+        if origin_request_id:
+            batch_log_data['origin_request_id'] = origin_request_id
 
         # 记录开始日志
         start_time = time.time()
@@ -887,12 +930,7 @@ class AsyncTamarModelClient(BaseClient, AsyncHttpFallbackMixin):
             extra={
                 "log_type": "request",
                 "uri": "/batch_invoke",
-                "data": {
-                    "batch_size": len(batch_request_model.items),
-                    "org_id": batch_request_model.user_context.org_id,
-                    "user_id": batch_request_model.user_context.user_id,
-                    "client_type": batch_request_model.user_context.client_type
-                }
+                "data": batch_log_data
             })
 
         try:

{tamar_model_client-0.1.25 → tamar_model_client-0.1.27}/tamar_model_client/core/__init__.py

@@ -22,6 +22,8 @@ from .logging_setup import (
     MAX_MESSAGE_LENGTH
 )
 
+from .request_id_manager import RequestIdManager
+
 __all__ = [
     # Utils
     'is_effective_value',
@@ -37,4 +39,6 @@ __all__ = [
     'get_protected_logger',
     'reset_logger_config',
     'MAX_MESSAGE_LENGTH',
+    # Request ID Management
+    'RequestIdManager',
 ]

{tamar_model_client-0.1.25 → tamar_model_client-0.1.27}/tamar_model_client/core/base_client.py

@@ -74,7 +74,7 @@ class BaseClient(ABC):
         
         # === 重试配置 ===
         self.max_retries = max_retries if max_retries is not None else int(
-            os.getenv("MODEL_MANAGER_SERVER_GRPC_MAX_RETRIES", 3))
+            os.getenv("MODEL_MANAGER_SERVER_GRPC_MAX_RETRIES", 6))
         self.retry_delay = retry_delay if retry_delay is not None else float(
             os.getenv("MODEL_MANAGER_SERVER_GRPC_RETRY_DELAY", 1.0))
         
@@ -131,7 +131,7 @@ class BaseClient(ABC):
             
         return options
     
-    def _build_auth_metadata(self, request_id: str) -> list:
+    def _build_auth_metadata(self, request_id: str, origin_request_id: Optional[str] = None) -> list:
         """
         构建认证元数据
         
@@ -140,12 +140,17 @@ class BaseClient(ABC):
         
         Args:
             request_id: 当前请求的唯一标识符
+            origin_request_id: 原始请求ID（可选）
             
         Returns:
             list: gRPC元数据列表，包含请求ID和认证令牌
         """
         metadata = [("x-request-id", request_id)]  # 将 request_id 添加到 headers
         
+        # 如果有原始请求ID，也添加到 headers
+        if origin_request_id:
+            metadata.append(("x-origin-request-id", origin_request_id))
+        
         if self.jwt_handler:
             self.jwt_token = self.jwt_handler.encode_token(
                 self.default_payload, 

tamar_model_client-0.1.27/tamar_model_client/core/request_id_manager.py

@@ -0,0 +1,112 @@
+"""
+Request ID 管理器
+
+管理 request_id 的生成和追踪，支持为同一个原始 request_id 生成带序号的组合 ID。
+"""
+
+import threading
+import time
+from typing import Dict, Tuple, Optional
+
+
+class RequestIdManager:
+    """
+    管理 request_id 的生成和追踪
+    
+    为同一个原始 request_id 生成带序号的组合 ID，如：
+    - 原始 ID: abc-123
+    - 第一次调用: abc-123-1
+    - 第二次调用: abc-123-2
+    
+    包含自动清理机制，避免内存泄漏。
+    """
+    
+    def __init__(self, ttl_seconds: int = 3600):
+        """
+        初始化 RequestIdManager
+        
+        Args:
+            ttl_seconds: 计数器的生存时间（秒），默认 1 小时
+        """
+        self._counters: Dict[str, Dict[str, any]] = {}  # {origin_id: {'count': int, 'last_used': float}}
+        self._lock = threading.Lock()
+        self._ttl = ttl_seconds
+        self._last_cleanup = time.time()
+        self._cleanup_interval = 300  # 每 5 分钟执行一次清理
+        
+    def get_composite_id(self, origin_id: str) -> Tuple[str, str]:
+        """
+        获取组合的 request_id
+        
+        Args:
+            origin_id: 原始的 request_id
+            
+        Returns:
+            tuple: (composite_request_id, origin_request_id)
+                - composite_request_id: 带序号的组合 ID，如 "abc-123-1"
+                - origin_request_id: 原始 ID，如 "abc-123"
+        """
+        with self._lock:
+            current_time = time.time()
+            
+            # 定期清理过期的计数器
+            if current_time - self._last_cleanup > self._cleanup_interval:
+                self._cleanup_expired(current_time)
+                self._last_cleanup = current_time
+            
+            # 获取或初始化计数器
+            if origin_id not in self._counters:
+                self._counters[origin_id] = {
+                    'count': 0,
+                    'last_used': current_time
+                }
+            
+            # 递增计数
+            self._counters[origin_id]['count'] += 1
+            self._counters[origin_id]['last_used'] = current_time
+            
+            composite_id = f"{origin_id}-{self._counters[origin_id]['count']}"
+            return composite_id, origin_id
+    
+    def _cleanup_expired(self, current_time: float):
+        """
+        清理过期的计数器
+        
+        Args:
+            current_time: 当前时间戳
+        """
+        expired_ids = [
+            origin_id for origin_id, info in self._counters.items()
+            if current_time - info['last_used'] > self._ttl
+        ]
+        
+        for origin_id in expired_ids:
+            del self._counters[origin_id]
+            
+        # 如果计数器数量过多，保留最近使用的一半
+        if len(self._counters) > 1000:
+            sorted_items = sorted(
+                self._counters.items(),
+                key=lambda x: x[1]['last_used'],
+                reverse=True
+            )[:500]
+            self._counters = dict(sorted_items)
+    
+    def get_stats(self) -> Dict[str, any]:
+        """
+        获取统计信息
+        
+        Returns:
+            dict: 包含计数器数量等统计信息
+        """
+        with self._lock:
+            return {
+                'total_counters': len(self._counters),
+                'ttl_seconds': self._ttl,
+                'cleanup_interval': self._cleanup_interval
+            }
+    
+    def clear(self):
+        """清空所有计数器"""
+        with self._lock:
+            self._counters.clear()

{tamar_model_client-0.1.25 → tamar_model_client-0.1.27}/tamar_model_client/exceptions.py

@@ -77,7 +77,7 @@ RETRY_POLICY = {
     grpc.StatusCode.CANCELLED: {
         'retryable': True,
         'backoff': 'linear',        # 线性退避，网络问题通常不需要指数退避
-        'max_attempts': 2,          # 最大重试次数（不包括初始请求），总共会尝试3次
+        'max_attempts': 5,          # 最大重试次数（不包括初始请求），总共会尝试6次
         'check_details': False      # 不检查详细信息，统一重试
     },
     grpc.StatusCode.ABORTED: {

{tamar_model_client-0.1.25 → tamar_model_client-0.1.27}/tamar_model_client/sync_client.py

@@ -32,7 +32,9 @@ from .core import (
     generate_request_id,
     set_request_id,
     get_protected_logger,
-    MAX_MESSAGE_LENGTH, get_request_id
+    MAX_MESSAGE_LENGTH, 
+    get_request_id,
+    RequestIdManager
 )
 from .core.base_client import BaseClient
 from .core.request_builder import RequestBuilder
@@ -99,6 +101,9 @@ class TamarModelClient(BaseClient, HttpFallbackMixin):
         self._channel_error_count = 0
         self._last_channel_error_time = None
         self._channel_lock = threading.Lock()  # 线程安全的channel操作
+        
+        # === Request ID 管理 ===
+        self._request_id_manager = RequestIdManager()
 
     def close(self):
         """
@@ -808,10 +813,22 @@ class TamarModelClient(BaseClient, HttpFallbackMixin):
                 "user_id": model_request.user_context.user_id or ""
             }
 
-        if not request_id:
+        # 处理 request_id
+        origin_request_id = None
+        if request_id:
+            # 用户提供了 request_id，生成组合 ID
+            request_id, origin_request_id = self._request_id_manager.get_composite_id(request_id)
+        else:
+            # 没有提供，生成新的
             request_id = generate_request_id()
+            
         set_request_id(request_id)
-        metadata = self._build_auth_metadata(request_id)
+        metadata = self._build_auth_metadata(request_id, origin_request_id)
+
+        # 构建日志数据
+        log_data = ResponseHandler.build_log_data(model_request)
+        if origin_request_id:
+            log_data['origin_request_id'] = origin_request_id
 
         # 记录开始日志
         start_time = time.time()
@@ -820,7 +837,7 @@ class TamarModelClient(BaseClient, HttpFallbackMixin):
             extra={
                 "log_type": "request",
                 "uri": f"/invoke/{model_request.provider.value}/{model_request.invoke_type.value}",
-                "data": ResponseHandler.build_log_data(model_request)
+                "data": log_data
             })
 
         try:
@@ -862,13 +879,19 @@ class TamarModelClient(BaseClient, HttpFallbackMixin):
                 # 记录非流式响应的成功日志
                 duration = time.time() - start_time
                 content_length = len(result.content) if result.content else 0
+                
+                # 构建响应日志数据
+                response_log_data = ResponseHandler.build_log_data(model_request, result)
+                if origin_request_id:
+                    response_log_data['origin_request_id'] = origin_request_id
+                    
                 logger.info(
                     f"✅ Request completed | content_length: {content_length}",
                     extra={
                         "log_type": "response",
                         "uri": f"/invoke/{model_request.provider.value}/{model_request.invoke_type.value}",
                         "duration": duration,
-                        "data": ResponseHandler.build_log_data(model_request, result)
+                        "data": response_log_data
                     }
                 )
                 
@@ -881,15 +904,18 @@ class TamarModelClient(BaseClient, HttpFallbackMixin):
         except (ConnectionError, grpc.RpcError) as e:
             duration = time.time() - start_time
             error_message = f"❌ Invoke gRPC failed: {str(e)}"
+            
+            # 构建错误日志数据
+            error_log_data = ResponseHandler.build_log_data(model_request, error=e)
+            if origin_request_id:
+                error_log_data['origin_request_id'] = origin_request_id
+                
             logger.error(error_message, exc_info=True,
                          extra={
                              "log_type": "response",
                              "uri": f"/invoke/{model_request.provider.value}/{model_request.invoke_type.value}",
                              "duration": duration,
-                             "data": ResponseHandler.build_log_data(
-                                 model_request,
-                                 error=e
-                             )
+                             "data": error_log_data
                          })
             
             # 记录 channel 错误
@@ -943,10 +969,27 @@ class TamarModelClient(BaseClient, HttpFallbackMixin):
                 "user_id": batch_request_model.user_context.user_id or ""
             }
 
-        if not request_id:
+        # 处理 request_id
+        origin_request_id = None
+        if request_id:
+            # 用户提供了 request_id，生成组合 ID
+            request_id, origin_request_id = self._request_id_manager.get_composite_id(request_id)
+        else:
+            # 没有提供，生成新的
             request_id = generate_request_id()
+            
         set_request_id(request_id)
-        metadata = self._build_auth_metadata(request_id)
+        metadata = self._build_auth_metadata(request_id, origin_request_id)
+
+        # 构建日志数据
+        batch_log_data = {
+            "batch_size": len(batch_request_model.items),
+            "org_id": batch_request_model.user_context.org_id,
+            "user_id": batch_request_model.user_context.user_id,
+            "client_type": batch_request_model.user_context.client_type
+        }
+        if origin_request_id:
+            batch_log_data['origin_request_id'] = origin_request_id
 
         # 记录开始日志
         start_time = time.time()
@@ -955,12 +998,7 @@ class TamarModelClient(BaseClient, HttpFallbackMixin):
             extra={
                 "log_type": "request",
                 "uri": "/batch_invoke",
-                "data": {
-                    "batch_size": len(batch_request_model.items),
-                    "org_id": batch_request_model.user_context.org_id,
-                    "user_id": batch_request_model.user_context.user_id,
-                    "client_type": batch_request_model.user_context.client_type
-                }
+                "data": batch_log_data
             })
 
         try:

{tamar_model_client-0.1.25 → tamar_model_client-0.1.27}/tamar_model_client.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tamar-model-client
-Version: 0.1.25
+Version: 0.1.27
 Summary: A Python SDK for interacting with the Model Manager gRPC service
 Home-page: http://gitlab.tamaredge.top/project-tap/AgentOS/model-manager-client
 Author: Oscar Ou

{tamar_model_client-0.1.25 → tamar_model_client-0.1.27}/tamar_model_client.egg-info/SOURCES.txt

@@ -20,6 +20,7 @@ tamar_model_client/core/base_client.py
 tamar_model_client/core/http_fallback.py
 tamar_model_client/core/logging_setup.py
 tamar_model_client/core/request_builder.py
+tamar_model_client/core/request_id_manager.py
 tamar_model_client/core/response_handler.py
 tamar_model_client/core/utils.py
 tamar_model_client/enums/__init__.py

{tamar_model_client-0.1.25 → tamar_model_client-0.1.27}/tests/test_google_azure_final.py

@@ -645,10 +645,10 @@ async def main():
         # await asyncio.wait_for(test_batch_requests(), timeout=120.0)
 
         # 同步并发测试
-        test_concurrent_requests(150)  # 测试150个并发请求
+        test_concurrent_requests(25)  # 测试150个并发请求
 
         # 异步并发测试
-        await test_async_concurrent_requests(150)  # 测试150个异步并发请求
+        await test_async_concurrent_requests(25)  # 测试150个异步并发请求
 
         print("\n✅ 测试完成")