tamar-model-client 0.2.3__tar.gz → 0.2.5__tar.gz

{tamar_model_client-0.2.3 → tamar_model_client-0.2.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tamar-model-client
-Version: 0.2.3
+Version: 0.2.5
 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
@@ -54,7 +54,7 @@ Dynamic: summary
 
 ### 🔌 多服务商支持
 - **OpenAI** (GPT-3.5/4, DALL-E)
-- **Google** (Gemini - AI Studio & Vertex AI)
+- **Google** (Gemini - AI Studio & Vertex AI, Imagen 图像生成)
 - **Azure OpenAI** (企业级部署)
 - **Anthropic** (Claude)
 - **DeepSeek** (深度求索)
@@ -114,7 +114,7 @@ tamar_model_client/
 │   └── outputs.py            # 响应模型（ModelResponse, Usage）
 ├── 📁 enums/                  # 枚举定义
 │   ├── providers.py          # AI 服务商（OpenAI, Google, Azure...）
-│   ├── invoke.py             # 调用类型（generation, images...）
+│   ├── invoke.py             # 调用类型（generation, images, image-generation-genai...）
 │   └── channel.py            # 服务通道（openai, vertexai...）
 ├── 📁 core/                   # 核心功能模块
 │   ├── base_client.py        # 客户端基类（熔断、降级、配置）
@@ -288,6 +288,37 @@ else:
     print(f"响应: {vertex_response.content}")
     if vertex_response.usage:
         print(f"Token 使用情况: {vertex_response.usage}")
+
+# Google GenAI 图像生成示例
+from google.genai import types
+
+genai_image_request = ModelRequest(
+    provider=ProviderType.GOOGLE,  # 选择 Google 作为提供商
+    channel=Channel.AI_STUDIO,  # 使用 AI Studio 渠道
+    invoke_type=InvokeType.IMAGE_GENERATION_GENAI,  # 使用 GenAI 图像生成调用类型
+    model="imagen-3.0-generate-001",  # 指定图像生成模型
+    prompt="一只可爱的小猫在花园里玩耍",  # 图像描述提示词
+    user_context=UserContext(
+        user_id="test_user",
+        org_id="test_org",
+        client_type="python-sdk"
+    ),
+    # 使用 Google GenAI 类型构建配置
+    config=types.GenerateImagesConfig(
+        number_of_images=1,
+        aspect_ratio="1:1",
+        safety_filter_level="block_some"
+    )
+)
+
+# 发送图像生成请求并获取响应
+image_response = client.invoke(genai_image_request)
+if image_response.error:
+    print(f"错误: {image_response.error}")
+else:
+    print(f"图像生成成功: {image_response.content}")
+    if image_response.usage:
+        print(f"使用情况: {image_response.usage}")
 ```
 
 ### Azure OpenAI 调用示例
@@ -488,6 +519,78 @@ async def batch_example():
 asyncio.run(batch_example())
 ```
 
+### 图像生成调用示例
+
+支持 OpenAI DALL-E、Google Vertex AI 和 Google GenAI 图像生成：
+
+```python
+from tamar_model_client import TamarModelClient
+from tamar_model_client.schemas import ModelRequest, UserContext
+from tamar_model_client.enums import ProviderType, InvokeType, Channel
+
+client = TamarModelClient()
+
+# OpenAI DALL-E 图像生成
+openai_image_request = ModelRequest(
+    provider=ProviderType.OPENAI,
+    channel=Channel.OPENAI,
+    invoke_type=InvokeType.IMAGE_GENERATION,
+    model="dall-e-3",
+    prompt="一只穿着西装的猫在办公室里工作",
+    user_context=UserContext(
+        user_id="test_user",
+        org_id="test_org",
+        client_type="python-sdk"
+    ),
+    size="1024x1024",
+    quality="hd",
+    n=1
+)
+
+# Google Vertex AI 图像生成
+vertex_image_request = ModelRequest(
+    provider=ProviderType.GOOGLE,
+    channel=Channel.VERTEXAI,
+    invoke_type=InvokeType.IMAGE_GENERATION,
+    model="imagegeneration@006",
+    prompt="一座美丽的山峰在日出时分",
+    user_context=UserContext(
+        user_id="test_user",
+        org_id="test_org",
+        client_type="python-sdk"
+    ),
+    number_of_images=1,
+    aspect_ratio="1:1",
+    safety_filter_level="block_some"
+)
+
+# Google GenAI 图像生成（新增功能）
+genai_image_request = ModelRequest(
+    provider=ProviderType.GOOGLE,
+    channel=Channel.AI_STUDIO,
+    invoke_type=InvokeType.IMAGE_GENERATION_GENAI,  # 新增的调用类型
+    model="imagen-3.0-generate-001",
+    prompt="科幻风格的城市夜景，霓虹灯闪烁",
+    user_context=UserContext(
+        user_id="test_user",
+        org_id="test_org",
+        client_type="python-sdk"
+    ),
+    config=types.GenerateImagesConfig(
+        number_of_images=1,
+        aspect_ratio="16:9"
+    )
+)
+
+# 发送请求
+for request in [openai_image_request, vertex_image_request, genai_image_request]:
+    response = client.invoke(request)
+    if response.error:
+        print(f"图像生成失败: {response.error}")
+    else:
+        print(f"图像生成成功: {response.content}")
+```
+
 ### 文件输入示例
 
 支持处理图像等文件输入（需使用支持多模态的模型，如 gemini-2.0-flash）：

{tamar_model_client-0.2.3 → tamar_model_client-0.2.5}/README.md

@@ -22,7 +22,7 @@
 
 ### 🔌 多服务商支持
 - **OpenAI** (GPT-3.5/4, DALL-E)
-- **Google** (Gemini - AI Studio & Vertex AI)
+- **Google** (Gemini - AI Studio & Vertex AI, Imagen 图像生成)
 - **Azure OpenAI** (企业级部署)
 - **Anthropic** (Claude)
 - **DeepSeek** (深度求索)
@@ -82,7 +82,7 @@ tamar_model_client/
 │   └── outputs.py            # 响应模型（ModelResponse, Usage）
 ├── 📁 enums/                  # 枚举定义
 │   ├── providers.py          # AI 服务商（OpenAI, Google, Azure...）
-│   ├── invoke.py             # 调用类型（generation, images...）
+│   ├── invoke.py             # 调用类型（generation, images, image-generation-genai...）
 │   └── channel.py            # 服务通道（openai, vertexai...）
 ├── 📁 core/                   # 核心功能模块
 │   ├── base_client.py        # 客户端基类（熔断、降级、配置）
@@ -256,6 +256,37 @@ else:
     print(f"响应: {vertex_response.content}")
     if vertex_response.usage:
         print(f"Token 使用情况: {vertex_response.usage}")
+
+# Google GenAI 图像生成示例
+from google.genai import types
+
+genai_image_request = ModelRequest(
+    provider=ProviderType.GOOGLE,  # 选择 Google 作为提供商
+    channel=Channel.AI_STUDIO,  # 使用 AI Studio 渠道
+    invoke_type=InvokeType.IMAGE_GENERATION_GENAI,  # 使用 GenAI 图像生成调用类型
+    model="imagen-3.0-generate-001",  # 指定图像生成模型
+    prompt="一只可爱的小猫在花园里玩耍",  # 图像描述提示词
+    user_context=UserContext(
+        user_id="test_user",
+        org_id="test_org",
+        client_type="python-sdk"
+    ),
+    # 使用 Google GenAI 类型构建配置
+    config=types.GenerateImagesConfig(
+        number_of_images=1,
+        aspect_ratio="1:1",
+        safety_filter_level="block_some"
+    )
+)
+
+# 发送图像生成请求并获取响应
+image_response = client.invoke(genai_image_request)
+if image_response.error:
+    print(f"错误: {image_response.error}")
+else:
+    print(f"图像生成成功: {image_response.content}")
+    if image_response.usage:
+        print(f"使用情况: {image_response.usage}")
 ```
 
 ### Azure OpenAI 调用示例
@@ -456,6 +487,78 @@ async def batch_example():
 asyncio.run(batch_example())
 ```
 
+### 图像生成调用示例
+
+支持 OpenAI DALL-E、Google Vertex AI 和 Google GenAI 图像生成：
+
+```python
+from tamar_model_client import TamarModelClient
+from tamar_model_client.schemas import ModelRequest, UserContext
+from tamar_model_client.enums import ProviderType, InvokeType, Channel
+
+client = TamarModelClient()
+
+# OpenAI DALL-E 图像生成
+openai_image_request = ModelRequest(
+    provider=ProviderType.OPENAI,
+    channel=Channel.OPENAI,
+    invoke_type=InvokeType.IMAGE_GENERATION,
+    model="dall-e-3",
+    prompt="一只穿着西装的猫在办公室里工作",
+    user_context=UserContext(
+        user_id="test_user",
+        org_id="test_org",
+        client_type="python-sdk"
+    ),
+    size="1024x1024",
+    quality="hd",
+    n=1
+)
+
+# Google Vertex AI 图像生成
+vertex_image_request = ModelRequest(
+    provider=ProviderType.GOOGLE,
+    channel=Channel.VERTEXAI,
+    invoke_type=InvokeType.IMAGE_GENERATION,
+    model="imagegeneration@006",
+    prompt="一座美丽的山峰在日出时分",
+    user_context=UserContext(
+        user_id="test_user",
+        org_id="test_org",
+        client_type="python-sdk"
+    ),
+    number_of_images=1,
+    aspect_ratio="1:1",
+    safety_filter_level="block_some"
+)
+
+# Google GenAI 图像生成（新增功能）
+genai_image_request = ModelRequest(
+    provider=ProviderType.GOOGLE,
+    channel=Channel.AI_STUDIO,
+    invoke_type=InvokeType.IMAGE_GENERATION_GENAI,  # 新增的调用类型
+    model="imagen-3.0-generate-001",
+    prompt="科幻风格的城市夜景，霓虹灯闪烁",
+    user_context=UserContext(
+        user_id="test_user",
+        org_id="test_org",
+        client_type="python-sdk"
+    ),
+    config=types.GenerateImagesConfig(
+        number_of_images=1,
+        aspect_ratio="16:9"
+    )
+)
+
+# 发送请求
+for request in [openai_image_request, vertex_image_request, genai_image_request]:
+    response = client.invoke(request)
+    if response.error:
+        print(f"图像生成失败: {response.error}")
+    else:
+        print(f"图像生成成功: {response.content}")
+```
+
 ### 文件输入示例
 
 支持处理图像等文件输入（需使用支持多模态的模型，如 gemini-2.0-flash）：

{tamar_model_client-0.2.3 → tamar_model_client-0.2.5}/setup.py

@@ -2,7 +2,7 @@ from setuptools import setup, find_packages
 
 setup(
     name="tamar-model-client",
-    version="0.2.3",
+    version="0.2.5",
     description="A Python SDK for interacting with the Model Manager gRPC service",
     author="Oscar Ou",
     author_email="oscar.ou@tamaredge.ai",

{tamar_model_client-0.2.3 → tamar_model_client-0.2.5}/tamar_model_client/core/request_builder.py

@@ -17,6 +17,7 @@ from ..schemas.inputs import (
     UserContext,
     GoogleGenAiInput,
     GoogleVertexAIImagesInput,
+    GoogleGenAIImagesInput,
     OpenAIResponsesInput,
     OpenAIChatCompletionsInput,
     OpenAIImagesInput,
@@ -32,7 +33,7 @@ class RequestBuilder:
     负责将高级的 ModelRequest 对象转换为 gRPC 协议所需的请求对象，
     包括参数验证、序列化和提供商特定的字段处理。
     """
-    
+
     @staticmethod
     def get_allowed_fields(provider: ProviderType, invoke_type: InvokeType) -> Set[str]:
         """
@@ -61,11 +62,13 @@ class RequestBuilder:
                 return set(OpenAIImagesInput.model_fields.keys())
             case ((ProviderType.OPENAI | ProviderType.AZURE), InvokeType.IMAGE_EDIT_GENERATION):
                 return set(OpenAIImagesEditInput.model_fields.keys())
+            case (ProviderType.GOOGLE, InvokeType.IMAGE_GENERATION_GENAI):
+                return set(GoogleGenAIImagesInput.model_fields.keys())
             case _:
                 raise ValueError(
                     f"Unsupported provider/invoke_type combination: {provider} + {invoke_type}"
                 )
-    
+
     @staticmethod
     def build_grpc_extra_fields(model_request: ModelRequest) -> Dict[str, Any]:
         """
@@ -88,31 +91,31 @@ class RequestBuilder:
                 model_request.provider,
                 model_request.invoke_type
             )
-            
+
             # 将 ModelRequest 转换为字典，只包含已设置的字段
             model_request_dict = model_request.model_dump(exclude_unset=True)
-            
+
             # 构建 gRPC 请求参数
             grpc_request_kwargs = {}
             for field in allowed_fields:
                 if field in model_request_dict:
                     value = model_request_dict[field]
-                    
+
                     # 跳过无效的值
                     if not is_effective_value(value):
                         continue
-                    
+
                     # 序列化不支持的类型
                     grpc_request_kwargs[field] = serialize_value(value)
-            
+
             # 清理序列化后的参数中的 None 值
             grpc_request_kwargs = remove_none_from_dict(grpc_request_kwargs)
-            
+
             return grpc_request_kwargs
-            
+
         except Exception as e:
             raise ValueError(f"构建请求失败: {str(e)}") from e
-    
+
     @staticmethod
     def build_single_request(model_request: ModelRequest) -> model_service_pb2.ModelRequestItem:
         """
@@ -129,7 +132,7 @@ class RequestBuilder:
         """
         # 构建额外字段
         extra_fields = RequestBuilder.build_grpc_extra_fields(model_request)
-        
+
         # 创建 gRPC 请求对象
         return model_service_pb2.ModelRequestItem(
             provider=model_request.provider.value,
@@ -141,11 +144,11 @@ class RequestBuilder:
             client_type=model_request.user_context.client_type or "",
             extra=extra_fields
         )
-    
+
     @staticmethod
     def build_batch_request_item(
-        batch_item: "BatchModelRequestItem", 
-        user_context: "UserContext"
+            batch_item: "BatchModelRequestItem",
+            user_context: "UserContext"
     ) -> model_service_pb2.ModelRequestItem:
         """
         构建批量请求中的单个项目
@@ -159,7 +162,7 @@ class RequestBuilder:
         """
         # 构建额外字段
         extra_fields = RequestBuilder.build_grpc_extra_fields(batch_item)
-        
+
         # 添加 custom_id 如果存在
         if hasattr(batch_item, 'custom_id') and batch_item.custom_id:
             request_item = model_service_pb2.ModelRequestItem(
@@ -184,13 +187,13 @@ class RequestBuilder:
                 client_type=user_context.client_type or "",
                 extra=extra_fields
             )
-            
+
         # 添加 priority 如果存在
         if hasattr(batch_item, 'priority') and batch_item.priority is not None:
             request_item.priority = batch_item.priority
-            
+
         return request_item
-    
+
     @staticmethod
     def build_batch_request(batch_request: BatchModelRequest) -> model_service_pb2.ModelRequest:
         """
@@ -206,16 +209,16 @@ class RequestBuilder:
             ValueError: 当构建请求失败时
         """
         items = []
-        
+
         for batch_item in batch_request.items:
             # 为每个请求项构建 gRPC 对象，传入 user_context
             request_item = RequestBuilder.build_batch_request_item(
-                batch_item, 
+                batch_item,
                 batch_request.user_context
             )
             items.append(request_item)
-        
+
         # 创建批量请求对象
         return model_service_pb2.ModelRequest(
             items=items
-        )
+        )

{tamar_model_client-0.2.3 → tamar_model_client-0.2.5}/tamar_model_client/enums/invoke.py

@@ -9,3 +9,4 @@ class InvokeType(str, Enum):
     GENERATION = "generation"  # 生成类，默认的值
     IMAGE_GENERATION = "image-generation"
     IMAGE_EDIT_GENERATION = "image-edit-generation"
+    IMAGE_GENERATION_GENAI = "image-generation-genai"  # GenAI SDK图像生成

{tamar_model_client-0.2.3 → tamar_model_client-0.2.5}/tamar_model_client/schemas/inputs.py

@@ -11,8 +11,9 @@ from openai.types.chat import ChatCompletionMessageParam, ChatCompletionAudioPar
     ChatCompletionToolParam
 from openai.types.responses import ResponseInputParam, ResponseIncludable, ResponseTextConfigParam, \
     response_create_params, ToolParam
+from openai.types.responses.response_prompt_param import ResponsePromptParam
 from pydantic import BaseModel, model_validator, field_validator
-from typing import List, Optional, Union, Iterable, Dict, Literal, IO
+from typing import List, Optional, Union, Iterable, Dict, Literal
 
 from tamar_model_client.enums import ProviderType, InvokeType
 from tamar_model_client.enums.channel import Channel
@@ -58,25 +59,44 @@ class GoogleVertexAIImagesInput(BaseModel):
     }
 
 
+class GoogleGenAIImagesInput(BaseModel):
+    model: str
+    prompt: str
+    config: Optional[types.GenerateImagesConfigOrDict] = None
+
+    model_config = {
+        "arbitrary_types_allowed": True
+    }
+
+
 class OpenAIResponsesInput(BaseModel):
-    input: Union[str, ResponseInputParam]
-    model: ResponsesModel
+    background: Optional[bool] | NotGiven = NOT_GIVEN
     include: Optional[List[ResponseIncludable]] | NotGiven = NOT_GIVEN
+    input: Union[str, ResponseInputParam] | NotGiven = NOT_GIVEN
     instructions: Optional[str] | NotGiven = NOT_GIVEN
     max_output_tokens: Optional[int] | NotGiven = NOT_GIVEN
+    max_tool_calls: Optional[int] | NotGiven = NOT_GIVEN
     metadata: Optional[Metadata] | NotGiven = NOT_GIVEN
+    model: ResponsesModel | NotGiven = NOT_GIVEN
     parallel_tool_calls: Optional[bool] | NotGiven = NOT_GIVEN
     previous_response_id: Optional[str] | NotGiven = NOT_GIVEN
+    prompt: Optional[ResponsePromptParam] | NotGiven = NOT_GIVEN
+    prompt_cache_key: str | NotGiven = NOT_GIVEN
     reasoning: Optional[Reasoning] | NotGiven = NOT_GIVEN
+    safety_identifier: str | NotGiven = NOT_GIVEN
+    service_tier: Optional[Literal["auto", "default", "flex", "scale", "priority"]] | NotGiven = NOT_GIVEN
     store: Optional[bool] | NotGiven = NOT_GIVEN
     stream: Optional[Literal[False]] | Literal[True] | NotGiven = NOT_GIVEN
+    stream_options: Optional[response_create_params.StreamOptions] | NotGiven = NOT_GIVEN
     temperature: Optional[float] | NotGiven = NOT_GIVEN
     text: ResponseTextConfigParam | NotGiven = NOT_GIVEN
     tool_choice: response_create_params.ToolChoice | NotGiven = NOT_GIVEN
     tools: Iterable[ToolParam] | NotGiven = NOT_GIVEN
+    top_logprobs: Optional[int] | NotGiven = NOT_GIVEN
     top_p: Optional[float] | NotGiven = NOT_GIVEN
     truncation: Optional[Literal["auto", "disabled"]] | NotGiven = NOT_GIVEN
     user: str | NotGiven = NOT_GIVEN
+    verbosity: Optional[Literal["low", "medium", "high"]] | NotGiven = NOT_GIVEN
     extra_headers: Headers | None = None
     extra_query: Query | None = None
     extra_body: Body | None = None
@@ -104,10 +124,12 @@ class OpenAIChatCompletionsInput(BaseModel):
     parallel_tool_calls: bool | NotGiven = NOT_GIVEN
     prediction: Optional[ChatCompletionPredictionContentParam] | NotGiven = NOT_GIVEN
     presence_penalty: Optional[float] | NotGiven = NOT_GIVEN
+    prompt_cache_key: str | NotGiven = NOT_GIVEN
     reasoning_effort: Optional[ReasoningEffort] | NotGiven = NOT_GIVEN
     response_format: completion_create_params.ResponseFormat | NotGiven = NOT_GIVEN
+    safety_identifier: str | NotGiven = NOT_GIVEN
     seed: Optional[int] | NotGiven = NOT_GIVEN
-    service_tier: Optional[Literal["auto", "default"]] | NotGiven = NOT_GIVEN
+    service_tier: Optional[Literal["auto", "default", "flex", "scale", "priority"]] | NotGiven = NOT_GIVEN
     stop: Union[Optional[str], List[str], None] | NotGiven = NOT_GIVEN
     store: Optional[bool] | NotGiven = NOT_GIVEN
     stream: Optional[Literal[False]] | Literal[True] | NotGiven = NOT_GIVEN
@@ -118,6 +140,7 @@ class OpenAIChatCompletionsInput(BaseModel):
     top_logprobs: Optional[int] | NotGiven = NOT_GIVEN
     top_p: Optional[float] | NotGiven = NOT_GIVEN
     user: str | NotGiven = NOT_GIVEN
+    verbosity: Optional[Literal["low", "medium", "high"]] | NotGiven = NOT_GIVEN
     web_search_options: completion_create_params.WebSearchOptions | NotGiven = NOT_GIVEN
     extra_headers: Headers | None = None
     extra_query: Query | None = None
@@ -166,7 +189,7 @@ class OpenAIImagesEditInput(BaseModel):
     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", "hd", "low", "medium", "high", "auto"]] | 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
@@ -191,33 +214,44 @@ class ModelRequestInput(BaseRequest):
     # 合并 model 字段
     model: Optional[Union[str, ResponsesModel, ChatModel, ImageModel]] = None
 
-    # OpenAI Responses Input
+    # OpenAI Responses Input（合并）
     input: Optional[Union[str, ResponseInputParam]] = None
     include: Optional[Union[List[ResponseIncludable], NotGiven]] = NOT_GIVEN
     instructions: Optional[Union[str, NotGiven]] = NOT_GIVEN
     max_output_tokens: Optional[Union[int, NotGiven]] = NOT_GIVEN
+    max_tool_calls: Optional[Union[int, NotGiven]] = NOT_GIVEN
     metadata: Optional[Union[Metadata, NotGiven]] = NOT_GIVEN
     parallel_tool_calls: Optional[Union[bool, NotGiven]] = NOT_GIVEN
     previous_response_id: Optional[Union[str, NotGiven]] = NOT_GIVEN
+    # prompt 同名字段合并：Responses 的 ResponsePromptParam + 图片生成的 str
+    prompt: Optional[Union[str, ResponsePromptParam, NotGiven]] = NOT_GIVEN
+    prompt_cache_key: Optional[Union[str, NotGiven]] = NOT_GIVEN
     reasoning: Optional[Union[Reasoning, NotGiven]] = NOT_GIVEN
+    safety_identifier: Optional[Union[str, NotGiven]] = NOT_GIVEN
+    service_tier: Optional[Union[Literal["auto", "default", "flex", "scale", "priority"], NotGiven]] = NOT_GIVEN
     store: Optional[Union[bool, NotGiven]] = NOT_GIVEN
     stream: Optional[Union[Literal[False], Literal[True], NotGiven]] = NOT_GIVEN
+    # 合并两套 stream_options 类型
+    stream_options: Optional[
+        Union[response_create_params.StreamOptions, ChatCompletionStreamOptionsParam, NotGiven]] = NOT_GIVEN
     temperature: Optional[Union[float, NotGiven]] = NOT_GIVEN
     text: Optional[Union[ResponseTextConfigParam, NotGiven]] = NOT_GIVEN
     tool_choice: Optional[
         Union[response_create_params.ToolChoice, ChatCompletionToolChoiceOptionParam, NotGiven]
     ] = NOT_GIVEN
     tools: Optional[Union[Iterable[ToolParam], Iterable[ChatCompletionToolParam], NotGiven]] = NOT_GIVEN
+    top_logprobs: Optional[Union[int, NotGiven]] = NOT_GIVEN
     top_p: Optional[Union[float, NotGiven]] = NOT_GIVEN
     truncation: Optional[Union[Literal["auto", "disabled"], NotGiven]] = NOT_GIVEN
     user: Optional[Union[str, NotGiven]] = NOT_GIVEN
+    verbosity: Optional[Union[Literal["low", "medium", "high"], NotGiven]] = NOT_GIVEN
 
     extra_headers: Optional[Union[Headers, None]] = None
     extra_query: Optional[Union[Query, None]] = None
     extra_body: Optional[Union[Body, None]] = None
     timeout: Optional[Union[float, httpx.Timeout, None, NotGiven]] = NOT_GIVEN
 
-    # OpenAI Chat Completions Input
+    # OpenAI Chat Completions Input（合并）
     messages: Optional[Iterable[ChatCompletionMessageParam]] = None
     audio: Optional[Union[ChatCompletionAudioParam, NotGiven]] = NOT_GIVEN
     frequency_penalty: Optional[Union[float, NotGiven]] = NOT_GIVEN
@@ -226,34 +260,32 @@ class ModelRequestInput(BaseRequest):
     logit_bias: Optional[Union[Dict[str, int], NotGiven]] = NOT_GIVEN
     logprobs: Optional[Union[bool, NotGiven]] = NOT_GIVEN
     max_completion_tokens: Optional[Union[int, NotGiven]] = NOT_GIVEN
+    max_tokens: Optional[Union[int, NotGiven]] = NOT_GIVEN
     modalities: Optional[Union[List[Literal["text", "audio"]], NotGiven]] = NOT_GIVEN
-    n: Optional[Union[int, NotGiven]] = NOT_GIVEN
+    n: Optional[Union[int, NotGiven]] = NOT_GIVEN  # 复用给 Chat 和 Images
     prediction: Optional[Union[ChatCompletionPredictionContentParam, NotGiven]] = NOT_GIVEN
     presence_penalty: Optional[Union[float, NotGiven]] = NOT_GIVEN
     reasoning_effort: Optional[Union[ReasoningEffort, NotGiven]] = NOT_GIVEN
     response_format: Optional[
-        Union[Literal["url", "b64_json"], completion_create_params.ResponseFormat, NotGiven]] = NOT_GIVEN
-    seed: Optional[Union[int, NotGiven]] = NOT_GIVEN
-    service_tier: Optional[Union[Literal["auto", "default"], NotGiven]] = NOT_GIVEN
-    stop: Optional[Union[Optional[str], List[str], None, NotGiven]] = NOT_GIVEN
-    top_logprobs: Optional[Union[int, NotGiven]] = NOT_GIVEN
+        Union[completion_create_params.ResponseFormat, Literal["url", "b64_json"], NotGiven]
+    ] = NOT_GIVEN
+    seed: Optional[Union[int, NotGiven]] = NOT_GIVEN  # Chat/Vertex Images 共用
     web_search_options: Optional[Union[completion_create_params.WebSearchOptions, NotGiven]] = NOT_GIVEN
-    stream_options: Optional[Union[ChatCompletionStreamOptionsParam, NotGiven]] = NOT_GIVEN
 
     # Google GenAI Input
     contents: Optional[Union[types.ContentListUnion, types.ContentListUnionDict]] = None
-    config: Optional[types.GenerateContentConfigOrDict] = None
+    config: Optional[Union[types.GenerateContentConfigOrDict, types.GenerateImagesConfigOrDict]] = None
 
-    # OpenAIImagesInput + OpenAIImagesEditInput + GoogleVertexAIImagesInput 合并字段
+    # Images（OpenAI Images / Images Edit / Google Vertex Images 合并）
     image: Optional[Union[FileTypes, List[FileTypes]]] = None
-    prompt: Optional[str] = None
-    background: Optional[Literal["transparent", "opaque", "auto"]] | NotGiven = NOT_GIVEN
-    moderation: Optional[Literal["low", "auto"]] | NotGiven = NOT_GIVEN
-    input_fidelity: Optional[Literal["high", "low"]] | 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
-    mask: FileTypes | NotGiven = NOT_GIVEN
+    # background 同名字段合并：Responses 的 bool（后台任务）+ Images 的透明度枚举
+    background: Optional[Union[bool, Literal["transparent", "opaque", "auto"], NotGiven]] = NOT_GIVEN
+    moderation: Optional[Union[Literal["low", "auto"], NotGiven]] = NOT_GIVEN
+    input_fidelity: Optional[Union[Literal["high", "low"], NotGiven]] = NOT_GIVEN
+    output_compression: Optional[Union[int, NotGiven]] = NOT_GIVEN
+    output_format: Optional[Union[Literal["png", "jpeg", "webp"], NotGiven]] = NOT_GIVEN
+    partial_images: Optional[Union[int, NotGiven]] = NOT_GIVEN
+    mask: Union[FileTypes, NotGiven] = NOT_GIVEN
     negative_prompt: Optional[str] = None
     aspect_ratio: Optional[Literal["1:1", "9:16", "16:9", "4:3", "3:4"]] = None
     guidance_scale: Optional[float] = None
@@ -262,10 +294,14 @@ class ModelRequestInput(BaseRequest):
     add_watermark: Optional[bool] = None
     safety_filter_level: Optional[Literal["block_most", "block_some", "block_few", "block_fewest"]] = None
     person_generation: Optional[Literal["dont_allow", "allow_adult", "allow_all"]] = None
-    quality: Optional[Literal["standard", "hd", "low", "medium", "high", "auto"]] | NotGiven = NOT_GIVEN
-    size: Optional[Literal[
-        "auto", "1024x1024", "1536x1024", "1024x1536", "256x256", "512x512", "1792x1024", "1024x1792"]] | NotGiven = NOT_GIVEN
-    style: Optional[Literal["vivid", "natural"]] | NotGiven = NOT_GIVEN
+    quality: Optional[Union[Literal["standard", "hd", "low", "medium", "high", "auto"], NotGiven]] = NOT_GIVEN
+    size: Optional[
+        Union[
+            Literal["auto", "1024x1024", "1536x1024", "1024x1536", "256x256", "512x512", "1792x1024", "1024x1792"],
+            NotGiven,
+        ]
+    ] = NOT_GIVEN
+    style: Optional[Union[Literal["vivid", "natural"], NotGiven]] = NOT_GIVEN
     number_of_images: Optional[int] = None  # Google 用法
 
     model_config = {

{tamar_model_client-0.2.3 → tamar_model_client-0.2.5}/tamar_model_client/utils.py

@@ -3,6 +3,7 @@ from pydantic import BaseModel
 from typing import Any
 import os, mimetypes
 
+
 def convert_file_field(value: Any) -> Any:
     def is_file_like(obj):
         return hasattr(obj, "read") and callable(obj.read)
@@ -55,7 +56,7 @@ def validate_fields_by_provider_and_invoke_type(
     """
     from tamar_model_client.enums import ProviderType, InvokeType
     from tamar_model_client.schemas.inputs import GoogleGenAiInput, OpenAIResponsesInput, OpenAIChatCompletionsInput, \
-        OpenAIImagesInput, OpenAIImagesEditInput, GoogleVertexAIImagesInput
+        OpenAIImagesInput, OpenAIImagesEditInput, GoogleVertexAIImagesInput, GoogleGenAIImagesInput
 
     google_allowed = extra_allowed_fields | set(GoogleGenAiInput.model_fields)
     openai_responses_allowed = extra_allowed_fields | set(OpenAIResponsesInput.model_fields)
@@ -63,9 +64,11 @@ def validate_fields_by_provider_and_invoke_type(
     openai_images_allowed = extra_allowed_fields | set(OpenAIImagesInput.model_fields)
     openai_images_edit_allowed = extra_allowed_fields | set(OpenAIImagesEditInput.model_fields)
     google_vertexai_images_allowed = extra_allowed_fields | set(GoogleVertexAIImagesInput.model_fields)
+    google_genai_images_allowed = extra_allowed_fields | set(GoogleGenAIImagesInput.model_fields)
 
     google_required = {"model", "contents"}
     google_vertex_required = {"model", "prompt"}
+    google_genai_images_required = {"model", "prompt"}
     openai_resp_required = {"input", "model"}
     openai_chat_required = {"messages", "model"}
     openai_img_required = {"prompt"}
@@ -90,6 +93,9 @@ def validate_fields_by_provider_and_invoke_type(
         case ((ProviderType.OPENAI | ProviderType.AZURE), InvokeType.IMAGE_EDIT_GENERATION):
             allowed = openai_images_edit_allowed
             required = openai_edit_required
+        case (ProviderType.GOOGLE, InvokeType.IMAGE_GENERATION_GENAI):
+            allowed = google_genai_images_allowed
+            required = google_genai_images_required
         case _:
             raise ValueError(f"Unsupported provider/invoke_type: {instance.provider} + {instance.invoke_type}")