tamar-model-client 0.1.30__py3-none-any.whl → 0.2.1__py3-none-any.whl

tamar_model_client/async_client.py

@@ -598,7 +598,7 @@ class AsyncTamarModelClient(BaseClient, AsyncHttpFallbackMixin):
             
         return delay
 
-    async def _stream(self, request, metadata, invoke_timeout) -> AsyncIterator[ModelResponse]:
+    async def _stream(self, request, metadata, invoke_timeout, request_id=None, origin_request_id=None) -> AsyncIterator[ModelResponse]:
         """
         处理流式响应
         
@@ -606,8 +606,10 @@ class AsyncTamarModelClient(BaseClient, AsyncHttpFallbackMixin):
         
         Args:
             request: gRPC 请求对象
-            metadata: 请求元数据
+            metadata: 请求元数据（为了兼容性保留，但会被忽略）
             invoke_timeout: 总体超时时间
+            request_id: 请求ID
+            origin_request_id: 原始请求ID
             
         Yields:
             ModelResponse: 流式响应的每个数据块
@@ -615,7 +617,12 @@ class AsyncTamarModelClient(BaseClient, AsyncHttpFallbackMixin):
         Raises:
             TimeoutError: 当等待下一个数据块超时时
         """
-        stream_iter = self.stub.Invoke(request, metadata=metadata, timeout=invoke_timeout).__aiter__()
+        # 每次调用时重新生成metadata，确保JWT token是最新的
+        fresh_metadata = self._build_auth_metadata(
+            request_id or get_request_id(), 
+            origin_request_id
+        )
+        stream_iter = self.stub.Invoke(request, metadata=fresh_metadata, timeout=invoke_timeout).__aiter__()
         chunk_timeout = 30.0  # 单个数据块的超时时间
         
         try:
@@ -636,7 +643,7 @@ class AsyncTamarModelClient(BaseClient, AsyncHttpFallbackMixin):
         except Exception as e:
             raise
 
-    async def _stream_with_logging(self, request, metadata, invoke_timeout, start_time, model_request) -> AsyncIterator[
+    async def _stream_with_logging(self, request, metadata, invoke_timeout, start_time, model_request, request_id=None, origin_request_id=None) -> AsyncIterator[
         ModelResponse]:
         """流式响应的包装器，用于记录完整的响应日志并处理重试"""
         total_content = ""
@@ -645,7 +652,7 @@ class AsyncTamarModelClient(BaseClient, AsyncHttpFallbackMixin):
         chunk_count = 0
 
         # 使用重试逻辑获取流生成器
-        stream_generator = self._retry_request_stream(self._stream, request, metadata, invoke_timeout, request_id=get_request_id())
+        stream_generator = self._retry_request_stream(self._stream, request, metadata, invoke_timeout, request_id=request_id or get_request_id(), origin_request_id=origin_request_id)
 
         try:
             async for response in stream_generator:
@@ -719,9 +726,22 @@ class AsyncTamarModelClient(BaseClient, AsyncHttpFallbackMixin):
             )
             raise
 
-    async def _invoke_request(self, request, metadata, invoke_timeout):
-        """执行单个非流式请求"""
-        async for response in self.stub.Invoke(request, metadata=metadata, timeout=invoke_timeout):
+    async def _invoke_request(self, request, metadata, invoke_timeout, request_id=None, origin_request_id=None):
+        """执行单个非流式请求
+        
+        Args:
+            request: gRPC请求对象
+            metadata: 请求元数据（为了兼容性保留，但会被忽略）
+            invoke_timeout: 请求超时时间
+            request_id: 请求ID
+            origin_request_id: 原始请求ID
+        """
+        # 每次调用时重新生成metadata，确保JWT token是最新的
+        fresh_metadata = self._build_auth_metadata(
+            request_id or get_request_id(), 
+            origin_request_id
+        )
+        async for response in self.stub.Invoke(request, metadata=fresh_metadata, timeout=invoke_timeout):
             return ResponseHandler.build_model_response(response)
 
     async def invoke(self, model_request: ModelRequest, timeout: Optional[float] = None,
@@ -817,13 +837,13 @@ class AsyncTamarModelClient(BaseClient, AsyncHttpFallbackMixin):
             invoke_timeout = timeout or self.default_invoke_timeout
             if model_request.stream:
                 # 对于流式响应，直接返回带日志记录的包装器
-                return self._stream_with_logging(request, metadata, invoke_timeout, start_time, model_request)
+                return self._stream_with_logging(request, metadata, invoke_timeout, start_time, model_request, request_id, origin_request_id)
             else:
                 # 存储model_request和origin_request_id供重试方法使用
                 self._current_model_request = model_request
                 self._current_origin_request_id = origin_request_id
                 try:
-                    result = await self._retry_request(self._invoke_request, request, metadata, invoke_timeout, request_id=request_id)
+                    result = await self._retry_request(self._invoke_request, request, metadata, invoke_timeout, request_id=request_id, origin_request_id=origin_request_id)
                 finally:
                     # 清理临时存储
                     if hasattr(self, '_current_model_request'):

tamar_model_client/auth.py

@@ -1,14 +1,43 @@
 import time
 import jwt
+from typing import Optional
 
 
 # JWT 处理类
 class JWTAuthHandler:
     def __init__(self, secret_key: str):
         self.secret_key = secret_key
+        self._token_cache: Optional[str] = None
+        self._token_exp_time: Optional[int] = None
 
     def encode_token(self, payload: dict, expires_in: int = 3600) -> str:
         """生成带过期时间的 JWT Token"""
         payload = payload.copy()
-        payload["exp"] = int(time.time()) + expires_in
-        return jwt.encode(payload, self.secret_key, algorithm="HS256")
+        exp_time = int(time.time()) + expires_in
+        payload["exp"] = exp_time
+        token = jwt.encode(payload, self.secret_key, algorithm="HS256")
+        
+        # 缓存token和过期时间
+        self._token_cache = token
+        self._token_exp_time = exp_time
+        
+        return token
+    
+    def is_token_expiring_soon(self, buffer_seconds: int = 60) -> bool:
+        """检查token是否即将过期
+        
+        Args:
+            buffer_seconds: 提前多少秒认为token即将过期，默认60秒
+            
+        Returns:
+            bool: True表示token即将过期或已过期
+        """
+        if not self._token_exp_time:
+            return True
+        
+        current_time = int(time.time())
+        return current_time >= (self._token_exp_time - buffer_seconds)
+    
+    def get_cached_token(self) -> Optional[str]:
+        """获取缓存的token"""
+        return self._token_cache

tamar_model_client/core/base_client.py

@@ -165,10 +165,27 @@ class BaseClient(ABC):
             metadata.append(("x-origin-request-id", origin_request_id))
 
         if self.jwt_handler:
-            self.jwt_token = self.jwt_handler.encode_token(
-                self.default_payload,
-                expires_in=self.token_expires_in
-            )
+            # 检查token是否即将过期，如果是则刷新
+            if self.jwt_handler.is_token_expiring_soon():
+                self.jwt_token = self.jwt_handler.encode_token(
+                    self.default_payload,
+                    expires_in=self.token_expires_in
+                )
+            else:
+                # 使用缓存的token
+                cached_token = self.jwt_handler.get_cached_token()
+                if cached_token:
+                    self.jwt_token = cached_token
+                else:
+                    # 如果没有缓存，生成新token
+                    self.jwt_token = self.jwt_handler.encode_token(
+                        self.default_payload,
+                        expires_in=self.token_expires_in
+                    )
+            
+            metadata.append(("authorization", f"Bearer {self.jwt_token}"))
+        elif self.jwt_token:
+            # 使用用户提供的预生成token
             metadata.append(("authorization", f"Bearer {self.jwt_token}"))
 
         return metadata

tamar_model_client/core/http_fallback.py

@@ -322,14 +322,35 @@ class AsyncHttpFallbackMixin:
     """
     
     async def _ensure_http_client(self) -> None:
-        """Ensure async HTTP client is initialized"""
+        """Ensure async HTTP client is initialized in the correct event loop"""
+        import asyncio
+        import aiohttp
+        
+        # Get current event loop
+        current_loop = asyncio.get_running_loop()
+        
+        # Check if we need to recreate the session
+        need_new_session = False
+        
         if not hasattr(self, '_http_session') or not self._http_session:
-            import aiohttp
+            need_new_session = True
+        elif hasattr(self, '_http_session_loop') and self._http_session_loop != current_loop:
+            # Session was created in a different event loop
+            logger.warning("🔄 HTTP session bound to different event loop, recreating...")
+            # Close old session if possible
+            try:
+                await self._http_session.close()
+            except Exception as e:
+                logger.debug(f"Error closing old session: {e}")
+            need_new_session = True
+            
+        if need_new_session:
             self._http_session = aiohttp.ClientSession(
                 headers={
                     'User-Agent': 'AsyncTamarModelClient/1.0'
                 }
             )
+            self._http_session_loop = current_loop
             
             # Note: JWT token will be set per request in headers
     
@@ -528,4 +549,6 @@ class AsyncHttpFallbackMixin:
         """Clean up HTTP session"""
         if hasattr(self, '_http_session') and self._http_session:
             await self._http_session.close()
-            self._http_session = None
+            self._http_session = None
+        if hasattr(self, '_http_session_loop'):
+            self._http_session_loop = None

tamar_model_client/schemas/inputs.py

@@ -137,10 +137,12 @@ class OpenAIImagesInput(BaseModel):
     n: Optional[int] | NotGiven = NOT_GIVEN
     output_compression: Optional[int] | NotGiven = NOT_GIVEN
     output_format: Optional[Literal["png", "jpeg", "webp"]] | NotGiven = NOT_GIVEN
+    partial_images: Optional[int] | NotGiven = NOT_GIVEN,
     quality: Literal["standard", "hd"] | NotGiven = NOT_GIVEN
     response_format: Optional[Literal["url", "b64_json"]] | NotGiven = NOT_GIVEN
     size: Optional[Literal[
         "auto", "1024x1024", "1536x1024", "1024x1536", "256x256", "512x512", "1792x1024", "1024x1792"]] | NotGiven = NOT_GIVEN
+    stream: Optional[Literal[False]] | Literal[True] | NotGiven = NOT_GIVEN,
     style: Optional[Literal["vivid", "natural"]] | NotGiven = NOT_GIVEN
     user: str | NotGiven = NOT_GIVEN
     extra_headers: Headers | None = None
@@ -157,15 +159,18 @@ class OpenAIImagesEditInput(BaseModel):
     image: Union[FileTypes, List[FileTypes]]
     prompt: str
     background: Optional[Literal["transparent", "opaque", "auto"]] | NotGiven = NOT_GIVEN
+    input_fidelity: Optional[Literal["high", "low"]] | NotGiven = NOT_GIVEN,
     mask: FileTypes | NotGiven = NOT_GIVEN
     model: Union[str, ImageModel, None] | NotGiven = NOT_GIVEN
     n: Optional[int] | NotGiven = NOT_GIVEN
+    output_compression: Optional[int] | NotGiven = NOT_GIVEN,
+    output_format: Optional[Literal["png", "jpeg", "webp"]] | NotGiven = NOT_GIVEN,
+    partial_images: Optional[int] | NotGiven = NOT_GIVEN,
     quality: Optional[Literal["standard", "low", "medium", "high", "auto"]] | NotGiven = NOT_GIVEN
     response_format: Optional[Literal["url", "b64_json"]] | NotGiven = NOT_GIVEN
     size: Optional[Literal["256x256", "512x512", "1024x1024", "1536x1024", "1024x1536", "auto"]] | NotGiven = NOT_GIVEN
+    stream: Optional[Literal[False]] | Literal[True] | NotGiven = NOT_GIVEN,
     user: str | NotGiven = NOT_GIVEN
-    # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs.
-    # The extra values given here take precedence over values defined on the client or passed to this method.
     extra_headers: Headers | None = None
     extra_query: Query | None = None
     extra_body: Body | None = None

tamar_model_client/sync_client.py

@@ -742,14 +742,16 @@ class TamarModelClient(BaseClient, HttpFallbackMixin):
         else:
             raise TamarModelException("Unknown streaming error occurred")
 
-    def _stream(self, request, metadata, invoke_timeout) -> Iterator[ModelResponse]:
+    def _stream(self, request, metadata, invoke_timeout, request_id=None, origin_request_id=None) -> Iterator[ModelResponse]:
         """
         处理流式响应
         
         Args:
             request: gRPC 请求对象
-            metadata: 请求元数据
+            metadata: 请求元数据（为了兼容性保留，但会被忽略）
             invoke_timeout: 总体超时时间
+            request_id: 请求ID
+            origin_request_id: 原始请求ID
             
         Yields:
             ModelResponse: 流式响应的每个数据块
@@ -757,6 +759,11 @@ class TamarModelClient(BaseClient, HttpFallbackMixin):
         Raises:
             TimeoutError: 当等待下一个数据块超时时
         """
+        # 每次调用时重新生成metadata，确保JWT token是最新的
+        fresh_metadata = self._build_auth_metadata(
+            request_id or get_request_id(), 
+            origin_request_id
+        )
         import threading
         import queue
         
@@ -767,7 +774,7 @@ class TamarModelClient(BaseClient, HttpFallbackMixin):
         def fetch_responses():
             """在单独线程中获取流式响应"""
             try:
-                for response in self.stub.Invoke(request, metadata=metadata, timeout=invoke_timeout):
+                for response in self.stub.Invoke(request, metadata=fresh_metadata, timeout=invoke_timeout):
                     response_queue.put(response)
                 response_queue.put(None)  # 标记流结束
             except Exception as e:
@@ -799,7 +806,7 @@ class TamarModelClient(BaseClient, HttpFallbackMixin):
             except queue.Empty:
                 raise TimeoutError(f"流式响应在等待下一个数据块时超时 ({chunk_timeout}s)")
 
-    def _stream_with_logging(self, request, metadata, invoke_timeout, start_time, model_request) -> Iterator[
+    def _stream_with_logging(self, request, metadata, invoke_timeout, start_time, model_request, request_id=None, origin_request_id=None) -> Iterator[
         ModelResponse]:
         """流式响应的包装器，用于记录完整的响应日志并处理重试"""
         total_content = ""
@@ -808,7 +815,7 @@ class TamarModelClient(BaseClient, HttpFallbackMixin):
         chunk_count = 0
 
         try:
-            for response in self._stream(request, metadata, invoke_timeout):
+            for response in self._stream(request, metadata, invoke_timeout, request_id, origin_request_id):
                 chunk_count += 1
                 if response.content:
                     total_content += response.content
@@ -879,9 +886,22 @@ class TamarModelClient(BaseClient, HttpFallbackMixin):
             )
             raise
 
-    def _invoke_request(self, request, metadata, invoke_timeout):
-        """执行单个非流式请求"""
-        response = self.stub.Invoke(request, metadata=metadata, timeout=invoke_timeout)
+    def _invoke_request(self, request, metadata, invoke_timeout, request_id=None, origin_request_id=None):
+        """执行单个非流式请求
+        
+        Args:
+            request: gRPC请求对象
+            metadata: 请求元数据（为了兼容性保留，但会被忽略）
+            invoke_timeout: 请求超时时间
+            request_id: 请求ID
+            origin_request_id: 原始请求ID
+        """
+        # 每次调用时重新生成metadata，确保JWT token是最新的
+        fresh_metadata = self._build_auth_metadata(
+            request_id or get_request_id(), 
+            origin_request_id
+        )
+        response = self.stub.Invoke(request, metadata=fresh_metadata, timeout=invoke_timeout)
         for response in response:
             return ResponseHandler.build_model_response(response)
 
@@ -980,14 +1000,14 @@ class TamarModelClient(BaseClient, HttpFallbackMixin):
                 return self._retry_request_stream(
                     self._stream_with_logging,
                     request, metadata, invoke_timeout, start_time, model_request,
-                    request_id=request_id
+                    request_id=request_id, origin_request_id=origin_request_id
                 )
             else:
                 # 存储model_request和origin_request_id供重试方法使用
                 self._current_model_request = model_request
                 self._current_origin_request_id = origin_request_id
                 try:
-                    result = self._retry_request(self._invoke_request, request, metadata, invoke_timeout, request_id=request_id)
+                    result = self._retry_request(self._invoke_request, request, metadata, invoke_timeout, request_id=request_id, origin_request_id=origin_request_id)
                 finally:
                     # 清理临时存储
                     if hasattr(self, '_current_model_request'):

{tamar_model_client-0.1.30.dist-info → tamar_model_client-0.2.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tamar-model-client
-Version: 0.1.30
+Version: 0.2.1
 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-0.1.30.dist-info → tamar_model_client-0.2.1.dist-info}/RECORD

@@ -1,16 +1,16 @@
 tamar_model_client/__init__.py,sha256=4DEIUGlLTeiaECjJQbGYik7C0JO6hHwwfbLYpYpMdzg,444
-tamar_model_client/async_client.py,sha256=J3787otYGW2ycizLC5HDPibm_USNy4oKI5QQUo3L-aE,45328
-tamar_model_client/auth.py,sha256=gbwW5Aakeb49PMbmYvrYlVx1mfyn1LEDJ4qQVs-9DA4,438
+tamar_model_client/async_client.py,sha256=047a3Ts6Qe2Wcs4xyxtG71kvYngTuSHFzB3V8D97_ec,46431
+tamar_model_client/auth.py,sha256=DrtnFpG0ZKFUnTnV_Y-FuLRiC2kobcgg0W5Gr1ywg1k,1398
 tamar_model_client/circuit_breaker.py,sha256=Y3AVp7WzVYU-ubcmovKsJ8DRJbbO4G7vdZgSjnwcWJQ,5550
 tamar_model_client/error_handler.py,sha256=y7EipcqkXbCecSAOsnoSP3SH7hvZSNF_NUHooTi3hP0,18364
 tamar_model_client/exceptions.py,sha256=EOr4JMYI7hVszRvNYJ1JqsUNpVmd16T2KpJ0MkFTsUE,13073
 tamar_model_client/json_formatter.py,sha256=XT8XPMKKM2M22tuYR2e1rvWHcpz3UD9iLLgGPsGOjCI,2410
 tamar_model_client/logging_icons.py,sha256=MRTZ1Xvkep9ce_jdltj54_XZUXvIpQ95soRNmLdJ4qw,1837
-tamar_model_client/sync_client.py,sha256=vBxVvDFeY_Sd7JRLJwkdOcm6sCxmGaDW0tyCspp-n7E,52671
+tamar_model_client/sync_client.py,sha256=FbyjuyDRiXklSS_l5h5fwNxvABI-hpLGiIWAXqhPHoI,53760
 tamar_model_client/utils.py,sha256=Kn6pFz9GEC96H4eejEax66AkzvsrXI3WCSDtgDjnVTI,5238
 tamar_model_client/core/__init__.py,sha256=RMiZjV1S4csWPLxB_JfdOea8fYPz97Oj3humQSBw1OI,1054
-tamar_model_client/core/base_client.py,sha256=XUbMDM6B3ZtAnAdgSDF-bdLgACY46igCcMVoiMC3faQ,13056
-tamar_model_client/core/http_fallback.py,sha256=ULmHXfKPwP4T32xo7yQV_z2bGaI_L71BQIcylFs8dTM,21243
+tamar_model_client/core/base_client.py,sha256=spb4zjDuPczqnXNlDcIq_bDQ09TOpxeeuX7IxpTS_38,13859
+tamar_model_client/core/http_fallback.py,sha256=2N7-N_TZrtffDjuv9s3-CD8Xy7qw9AuI5xeWGUnGQ0w,22217
 tamar_model_client/core/logging_setup.py,sha256=-MXzTR4Ax50H16cbq1jCXbxgayf5fZ0U3o0--fMmxD8,6692
 tamar_model_client/core/request_builder.py,sha256=yi8iy2Ps2m4d1YwIFiQLRxTvxQxgEGV576aXnNYRl7E,8507
 tamar_model_client/core/request_id_manager.py,sha256=S-Mliaby9zN_bx-B85FvVnttal-w0skkjy2ZvWoQ5vw,3689
@@ -24,7 +24,7 @@ tamar_model_client/generated/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NM
 tamar_model_client/generated/model_service_pb2.py,sha256=RI6wNSmgmylzWPedFfPxx938UzS7kcPR58YTzYshcL8,3066
 tamar_model_client/generated/model_service_pb2_grpc.py,sha256=k4tIbp3XBxdyuOVR18Ung_4SUryONB51UYf_uUEl6V4,5145
 tamar_model_client/schemas/__init__.py,sha256=AxuI-TcvA4OMTj2FtK4wAItvz9LrK_293pu3cmMLE7k,394
-tamar_model_client/schemas/inputs.py,sha256=dz1m8NbUIxA99JXZc8WlyzbKpDuz1lEzx3VghC33zYI,14625
+tamar_model_client/schemas/inputs.py,sha256=vrBym9ywj0Zp9vegp-t6EWDnBiI_bAIQVWYhOY7Vh1A,14892
 tamar_model_client/schemas/outputs.py,sha256=M_fcqUtXPJnfiLabHlyA8BorlC5pYkf5KLjXO1ysKIQ,1031
 tests/__init__.py,sha256=kbmImddLDwdqlkkmkyKtl4bQy_ipe-R8eskpaBylU9w,38
 tests/stream_hanging_analysis.py,sha256=W3W48IhQbNAR6-xvMpoWZvnWOnr56CTaH4-aORNBuD4,14807
@@ -32,7 +32,7 @@ tests/test_circuit_breaker.py,sha256=nhEBnyXFjIYjRWlUdu7Z9PnPq48ypbBK6fxN6deHedw
 tests/test_google_azure_final.py,sha256=Cx2lfnoj48_7pUjpCYbrx6OLJF4cI79McV24_EYt_8s,55093
 tests/test_logging_issue.py,sha256=JTMbotfHpAEPMBj73pOwxPn-Zn4QVQJX6scMz48FRDQ,2427
 tests/test_simple.py,sha256=Xf0U-J9_xn_LzUsmYu06suK0_7DrPeko8OHoHldsNxE,7169
-tamar_model_client-0.1.30.dist-info/METADATA,sha256=QrJLOVUNNXKGNv4ZJRPKDFLP9d9JUlfrqQNN-FQuAvA,41310
-tamar_model_client-0.1.30.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-tamar_model_client-0.1.30.dist-info/top_level.txt,sha256=f1I-S8iWN-cgv4gB8gxRg9jJOTJMumvm4oGKVPfGg6A,25
-tamar_model_client-0.1.30.dist-info/RECORD,,
+tamar_model_client-0.2.1.dist-info/METADATA,sha256=sB0xQbFo6EAeT8ALLXbE9HEB4sQz9u8fdJMswLKMnAQ,41309
+tamar_model_client-0.2.1.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+tamar_model_client-0.2.1.dist-info/top_level.txt,sha256=f1I-S8iWN-cgv4gB8gxRg9jJOTJMumvm4oGKVPfGg6A,25
+tamar_model_client-0.2.1.dist-info/RECORD,,