vectorvein 0.1.56__py3-none-any.whl → 0.1.59__py3-none-any.whl

vectorvein/chat_clients/anthropic_client.py

@@ -249,6 +249,8 @@ class AnthropicChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage:
         pass
@@ -265,6 +267,8 @@ class AnthropicChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> Generator[ChatCompletionDeltaMessage, None, None]:
         pass
@@ -281,6 +285,8 @@ class AnthropicChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage | Generator[ChatCompletionDeltaMessage, Any, None]:
         pass
@@ -296,6 +302,8 @@ class AnthropicChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ):
         if model is not None:
@@ -304,10 +312,6 @@ class AnthropicChatClient(BaseChatClient):
             self.stream = stream
         if temperature is not None:
             self.temperature = temperature
-        if isinstance(tools, OpenAINotGiven):
-            tools = NOT_GIVEN
-        if isinstance(tool_choice, OpenAINotGiven):
-            tool_choice = NOT_GIVEN
 
         if self.random_endpoint:
             self.random_endpoint = True
@@ -348,6 +352,8 @@ class AnthropicChatClient(BaseChatClient):
                         tool_choice=_tool_choice,
                         response_format=response_format,
                         stream_options=stream_options,
+                        top_p=top_p,
+                        skip_cutoff=skip_cutoff,
                         **kwargs,
                     )
                     for chunk in response:
@@ -373,12 +379,20 @@ class AnthropicChatClient(BaseChatClient):
                     tools=_tools,
                     tool_choice=_tool_choice,
                     response_format=response_format,
-                    stream_options=stream_options,
+                    top_p=top_p,
+                    skip_cutoff=skip_cutoff,
                     **kwargs,
                 )
 
         assert isinstance(self.raw_client, Anthropic | AnthropicVertex)
 
+        if isinstance(tools, OpenAINotGiven):
+            tools = NOT_GIVEN
+        if isinstance(tool_choice, OpenAINotGiven):
+            tool_choice = NOT_GIVEN
+        if isinstance(top_p, OpenAINotGiven) or top_p is None:
+            top_p = NOT_GIVEN
+
         raw_client = self.raw_client  # 调用完 self.raw_client 后，self.model_id 会被赋值
         self.model_setting = self.backend_settings.models[self.model]
         if self.model_id is None:
@@ -390,7 +404,7 @@ class AnthropicChatClient(BaseChatClient):
         else:
             system_prompt = ""
 
-        if self.context_length_control == ContextLengthControlType.Latest:
+        if not skip_cutoff and self.context_length_control == ContextLengthControlType.Latest:
             messages = cutoff_messages(
                 messages,
                 max_count=self.model_setting.context_length,
@@ -424,6 +438,7 @@ class AnthropicChatClient(BaseChatClient):
                 max_tokens=max_tokens,
                 tools=tools_params,
                 tool_choice=tool_choice_param,
+                top_p=top_p,
                 **kwargs,
             )
 
@@ -486,6 +501,7 @@ class AnthropicChatClient(BaseChatClient):
                 max_tokens=max_tokens,
                 tools=tools_params,
                 tool_choice=tool_choice_param,
+                top_p=top_p,
                 **kwargs,
             )
 
@@ -614,6 +630,8 @@ class AsyncAnthropicChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage:
         pass
@@ -630,6 +648,8 @@ class AsyncAnthropicChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> AsyncGenerator[ChatCompletionDeltaMessage, Any]:
         pass
@@ -646,6 +666,8 @@ class AsyncAnthropicChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage | AsyncGenerator[ChatCompletionDeltaMessage, Any]:
         pass
@@ -661,6 +683,8 @@ class AsyncAnthropicChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ):
         if model is not None:
@@ -669,10 +693,6 @@ class AsyncAnthropicChatClient(BaseAsyncChatClient):
             self.stream = stream
         if temperature is not None:
             self.temperature = temperature
-        if isinstance(tools, OpenAINotGiven):
-            tools = NOT_GIVEN
-        if isinstance(tool_choice, OpenAINotGiven):
-            tool_choice = NOT_GIVEN
 
         if self.random_endpoint:
             self.random_endpoint = True
@@ -714,6 +734,8 @@ class AsyncAnthropicChatClient(BaseAsyncChatClient):
                         tool_choice=_tool_choice,
                         response_format=response_format,
                         stream_options=stream_options,
+                        top_p=top_p,
+                        skip_cutoff=skip_cutoff,
                         **kwargs,
                     )
                     async for chunk in response:
@@ -740,11 +762,20 @@ class AsyncAnthropicChatClient(BaseAsyncChatClient):
                     tools=_tools,
                     tool_choice=_tool_choice,
                     response_format=response_format,
+                    top_p=top_p,
+                    skip_cutoff=skip_cutoff,
                     **kwargs,
                 )
 
         assert isinstance(self.raw_client, AsyncAnthropic | AsyncAnthropicVertex)
 
+        if isinstance(tools, OpenAINotGiven):
+            tools = NOT_GIVEN
+        if isinstance(tool_choice, OpenAINotGiven):
+            tool_choice = NOT_GIVEN
+        if isinstance(top_p, OpenAINotGiven) or top_p is None:
+            top_p = NOT_GIVEN
+
         raw_client = self.raw_client  # 调用完 self.raw_client 后，self.model_id 会被赋值
         self.model_setting = self.backend_settings.models[self.model]
         if self.model_id is None:
@@ -756,7 +787,7 @@ class AsyncAnthropicChatClient(BaseAsyncChatClient):
         else:
             system_prompt = ""
 
-        if self.context_length_control == ContextLengthControlType.Latest:
+        if not skip_cutoff and self.context_length_control == ContextLengthControlType.Latest:
             messages = cutoff_messages(
                 messages,
                 max_count=self.model_setting.context_length,
@@ -790,6 +821,7 @@ class AsyncAnthropicChatClient(BaseAsyncChatClient):
                 max_tokens=max_tokens,
                 tools=tools_params,
                 tool_choice=tool_choice_param,
+                top_p=top_p,
                 **kwargs,
             )
 
@@ -852,6 +884,7 @@ class AsyncAnthropicChatClient(BaseAsyncChatClient):
                 max_tokens=max_tokens,
                 tools=tools_params,
                 tool_choice=tool_choice_param,
+                top_p=top_p,
                 **kwargs,
             )
 

vectorvein/chat_clients/base_client.py

@@ -73,6 +73,8 @@ class BaseChatClient(ABC):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage:
         pass
@@ -90,6 +92,8 @@ class BaseChatClient(ABC):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> Generator[ChatCompletionDeltaMessage, Any, None]:
         pass
@@ -107,6 +111,8 @@ class BaseChatClient(ABC):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage | Generator[ChatCompletionDeltaMessage, Any, None]:
         pass
@@ -123,6 +129,8 @@ class BaseChatClient(ABC):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage | Generator[ChatCompletionDeltaMessage, Any, None]:
         pass
@@ -137,6 +145,7 @@ class BaseChatClient(ABC):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
         **kwargs,
     ) -> Generator[ChatCompletionDeltaMessage, Any, None]:
         return self.create_completion(
@@ -206,6 +215,8 @@ class BaseAsyncChatClient(ABC):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage:
         pass
@@ -223,6 +234,8 @@ class BaseAsyncChatClient(ABC):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> AsyncGenerator[ChatCompletionDeltaMessage, None]:
         pass
@@ -240,6 +253,8 @@ class BaseAsyncChatClient(ABC):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage | AsyncGenerator[ChatCompletionDeltaMessage, None]:
         pass
@@ -256,6 +271,8 @@ class BaseAsyncChatClient(ABC):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage | AsyncGenerator[ChatCompletionDeltaMessage, None]:
         pass
@@ -270,6 +287,7 @@ class BaseAsyncChatClient(ABC):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
         **kwargs,
     ) -> AsyncGenerator[ChatCompletionDeltaMessage, None]:
         return await self.create_completion(

vectorvein/chat_clients/gemini_client.py

@@ -65,6 +65,8 @@ class GeminiChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage:
         pass
@@ -81,6 +83,8 @@ class GeminiChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> Generator[ChatCompletionDeltaMessage, None, None]:
         pass
@@ -97,6 +101,8 @@ class GeminiChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage | Generator[ChatCompletionDeltaMessage, Any, None]:
         pass
@@ -112,6 +118,8 @@ class GeminiChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ):
         if model is not None:
@@ -129,7 +137,7 @@ class GeminiChatClient(BaseChatClient):
         else:
             system_prompt = ""
 
-        if self.context_length_control == ContextLengthControlType.Latest:
+        if not skip_cutoff and self.context_length_control == ContextLengthControlType.Latest:
             messages = cutoff_messages(
                 messages,
                 max_count=self.model_setting.context_length,
@@ -137,16 +145,19 @@ class GeminiChatClient(BaseChatClient):
                 model=self.model_setting.id,
             )
 
+        tools_params = {}
         if tools:
             tools_params = {"tools": [{"function_declarations": [tool["function"] for tool in tools]}]}
-        else:
-            tools_params = {}
 
         response_format_params = {}
         if response_format is not None:
             if response_format.get("type") == "json_object":
                 response_format_params = {"response_mime_type": "application/json"}
 
+        top_p_params = {}
+        if top_p:
+            top_p_params = {"top_p": top_p}
+
         if self.random_endpoint:
             self.random_endpoint = True
             endpoint_choice = random.choice(self.backend_settings.models[self.model].endpoints)
@@ -168,6 +179,7 @@ class GeminiChatClient(BaseChatClient):
             "generationConfig": {
                 "temperature": self.temperature,
                 "maxOutputTokens": max_tokens,
+                **top_p_params,
                 **response_format_params,
             },
             **tools_params,
@@ -189,7 +201,7 @@ class GeminiChatClient(BaseChatClient):
                 if self.http_client:
                     client = self.http_client
                 else:
-                    client = httpx.Client()
+                    client = httpx.Client(timeout=300)
                 with client.stream("POST", url, headers=headers, params=params, json=request_body) as response:
                     for chunk in response.iter_lines():
                         message = {"content": "", "tool_calls": []}
@@ -228,8 +240,10 @@ class GeminiChatClient(BaseChatClient):
             if self.http_client:
                 client = self.http_client
             else:
-                client = httpx.Client()
+                client = httpx.Client(timeout=300)
             response = client.post(url, json=request_body, headers=headers, params=params, timeout=None).json()
+            if "error" in response:
+                raise Exception(response["error"])
             result = {
                 "content": "",
                 "usage": {
@@ -302,6 +316,8 @@ class AsyncGeminiChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage:
         pass
@@ -318,6 +334,8 @@ class AsyncGeminiChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> AsyncGenerator[ChatCompletionDeltaMessage, Any]:
         pass
@@ -334,6 +352,8 @@ class AsyncGeminiChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage | AsyncGenerator[ChatCompletionDeltaMessage, Any]:
         pass
@@ -349,6 +369,8 @@ class AsyncGeminiChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ):
         if model is not None:
@@ -366,7 +388,7 @@ class AsyncGeminiChatClient(BaseAsyncChatClient):
         else:
             system_prompt = ""
 
-        if self.context_length_control == ContextLengthControlType.Latest:
+        if not skip_cutoff and self.context_length_control == ContextLengthControlType.Latest:
             messages = cutoff_messages(
                 messages,
                 max_count=self.model_setting.context_length,
@@ -374,16 +396,19 @@ class AsyncGeminiChatClient(BaseAsyncChatClient):
                 model=self.model_setting.id,
             )
 
+        tools_params = {}
         if tools:
             tools_params = {"tools": [{"function_declarations": [tool["function"] for tool in tools]}]}
-        else:
-            tools_params = {}
 
         response_format_params = {}
         if response_format is not None:
             if response_format.get("type") == "json_object":
                 response_format_params = {"response_mime_type": "application/json"}
 
+        top_p_params = {}
+        if top_p:
+            top_p_params = {"top_p": top_p}
+
         if self.random_endpoint:
             self.random_endpoint = True
             endpoint_choice = random.choice(self.backend_settings.models[self.model].endpoints)
@@ -405,6 +430,7 @@ class AsyncGeminiChatClient(BaseAsyncChatClient):
             "generationConfig": {
                 "temperature": self.temperature,
                 "maxOutputTokens": max_tokens,
+                **top_p_params,
                 **response_format_params,
             },
             **tools_params,
@@ -426,7 +452,7 @@ class AsyncGeminiChatClient(BaseAsyncChatClient):
                 if self.http_client:
                     client = self.http_client
                 else:
-                    client = httpx.AsyncClient()
+                    client = httpx.AsyncClient(timeout=300)
                 async with client.stream("POST", url, headers=headers, params=params, json=request_body) as response:
                     async for chunk in response.aiter_lines():
                         message = {"content": "", "tool_calls": []}
@@ -465,10 +491,12 @@ class AsyncGeminiChatClient(BaseAsyncChatClient):
             if self.http_client:
                 client = self.http_client
             else:
-                client = httpx.AsyncClient()
+                client = httpx.AsyncClient(timeout=300)
             async with client:
                 response = await client.post(url, json=request_body, headers=headers, params=params, timeout=None)
                 response = response.json()
+                if "error" in response:
+                    raise Exception(response["error"])
                 result = {
                     "content": "",
                     "usage": {

vectorvein/chat_clients/minimax_client.py

@@ -91,6 +91,8 @@ class MiniMaxChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage:
         pass
@@ -107,6 +109,8 @@ class MiniMaxChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> Generator[ChatCompletionDeltaMessage, None, None]:
         pass
@@ -123,6 +127,8 @@ class MiniMaxChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage | Generator[ChatCompletionDeltaMessage, Any, None]:
         pass
@@ -138,6 +144,8 @@ class MiniMaxChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ):
         if model is not None:
@@ -162,7 +170,7 @@ class MiniMaxChatClient(BaseChatClient):
                 self.endpoint_id = endpoint_choice
             self.endpoint = settings.get_endpoint(self.endpoint_id)
 
-        if self.context_length_control == ContextLengthControlType.Latest:
+        if not skip_cutoff and self.context_length_control == ContextLengthControlType.Latest:
             messages = cutoff_messages(
                 messages,
                 max_count=self.model_setting.context_length,
@@ -190,6 +198,11 @@ class MiniMaxChatClient(BaseChatClient):
         else:
             tools_params = {}
 
+        if top_p:
+            top_p_params = {"top_p": top_p}
+        else:
+            top_p_params = {}
+
         if max_tokens is None:
             max_output_tokens = self.model_setting.max_output_tokens
             if max_output_tokens is not None:
@@ -218,6 +231,7 @@ class MiniMaxChatClient(BaseChatClient):
             "temperature": self.temperature,
             "stream": self.stream,
             "mask_sensitive_info": False,
+            **top_p_params,
             **tools_params,
             **kwargs,
         }
@@ -326,6 +340,8 @@ class AsyncMiniMaxChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage:
         pass
@@ -342,6 +358,8 @@ class AsyncMiniMaxChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> AsyncGenerator[ChatCompletionDeltaMessage, Any]:
         pass
@@ -358,6 +376,8 @@ class AsyncMiniMaxChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage | AsyncGenerator[ChatCompletionDeltaMessage, Any]:
         pass
@@ -373,6 +393,8 @@ class AsyncMiniMaxChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None = None,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ):
         if model is not None:
@@ -397,7 +419,7 @@ class AsyncMiniMaxChatClient(BaseAsyncChatClient):
                 self.endpoint_id = endpoint_choice
             self.endpoint = settings.get_endpoint(self.endpoint_id)
 
-        if self.context_length_control == ContextLengthControlType.Latest:
+        if not skip_cutoff and self.context_length_control == ContextLengthControlType.Latest:
             messages = cutoff_messages(
                 messages,
                 max_count=self.model_setting.context_length,
@@ -423,6 +445,11 @@ class AsyncMiniMaxChatClient(BaseAsyncChatClient):
         else:
             tools_params = {}
 
+        if top_p:
+            top_p_params = {"top_p": top_p}
+        else:
+            top_p_params = {}
+
         if max_tokens is None:
             max_output_tokens = self.model_setting.max_output_tokens
             if max_output_tokens is not None:
@@ -451,6 +478,7 @@ class AsyncMiniMaxChatClient(BaseAsyncChatClient):
             "temperature": self.temperature,
             "stream": self.stream,
             "mask_sensitive_info": False,
+            **top_p_params,
             **tools_params,
             **kwargs,
         }

vectorvein/chat_clients/openai_compatible_client.py

@@ -24,6 +24,7 @@ from ..types.llm_parameters import (
     ToolParam,
     ToolChoice,
     OpenAINotGiven,
+    AnthropicNotGiven,
     Usage,
     ChatCompletionMessage,
     ChatCompletionDeltaMessage,
@@ -97,6 +98,8 @@ class OpenAICompatibleChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None | OpenAINotGiven = NOT_GIVEN,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage:
         pass
@@ -113,6 +116,8 @@ class OpenAICompatibleChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None | OpenAINotGiven = NOT_GIVEN,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> Generator[ChatCompletionDeltaMessage, None, None]:
         pass
@@ -129,6 +134,8 @@ class OpenAICompatibleChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None | OpenAINotGiven = NOT_GIVEN,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage | Generator[ChatCompletionDeltaMessage, Any, None]:
         pass
@@ -144,6 +151,8 @@ class OpenAICompatibleChatClient(BaseChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None | OpenAINotGiven = NOT_GIVEN,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ):
         if model is not None:
@@ -152,13 +161,15 @@ class OpenAICompatibleChatClient(BaseChatClient):
             self.stream = stream
         if temperature is not None:
             self.temperature = temperature
+        if isinstance(top_p, AnthropicNotGiven):
+            top_p = NOT_GIVEN
 
         raw_client = self.raw_client  # 调用完 self.raw_client 后，self.model_id 会被赋值
         self.model_setting = self.backend_settings.models[self.model]
         if self.model_id is None:
             self.model_id = self.model_setting.id
 
-        if self.context_length_control == ContextLengthControlType.Latest:
+        if not skip_cutoff and self.context_length_control == ContextLengthControlType.Latest:
             messages = cutoff_messages(
                 messages,
                 max_count=self.model_setting.context_length,
@@ -206,6 +217,7 @@ class OpenAICompatibleChatClient(BaseChatClient):
                 stream=True,
                 temperature=self.temperature,
                 max_tokens=max_tokens,
+                top_p=top_p,
                 **_stream_options_params,
                 **self.response_format,
                 **tools_params,
@@ -260,6 +272,7 @@ class OpenAICompatibleChatClient(BaseChatClient):
                 stream=False,
                 temperature=self.temperature,
                 max_tokens=max_tokens,
+                top_p=top_p,
                 **self.response_format,
                 **tools_params,
                 **kwargs,
@@ -351,6 +364,8 @@ class AsyncOpenAICompatibleChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None | OpenAINotGiven = NOT_GIVEN,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage:
         pass
@@ -367,6 +382,8 @@ class AsyncOpenAICompatibleChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None | OpenAINotGiven = NOT_GIVEN,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> AsyncGenerator[ChatCompletionDeltaMessage, Any]:
         pass
@@ -383,6 +400,8 @@ class AsyncOpenAICompatibleChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None | OpenAINotGiven = NOT_GIVEN,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ) -> ChatCompletionMessage | AsyncGenerator[ChatCompletionDeltaMessage, Any]:
         pass
@@ -398,6 +417,8 @@ class AsyncOpenAICompatibleChatClient(BaseAsyncChatClient):
         tool_choice: ToolChoice | NotGiven = NOT_GIVEN,
         response_format: dict | None = None,
         stream_options: ChatCompletionStreamOptionsParam | None | OpenAINotGiven = NOT_GIVEN,
+        top_p: float | NotGiven | None = NOT_GIVEN,
+        skip_cutoff: bool = False,
         **kwargs,
     ):
         if model is not None:
@@ -406,13 +427,15 @@ class AsyncOpenAICompatibleChatClient(BaseAsyncChatClient):
             self.stream = stream
         if temperature is not None:
             self.temperature = temperature
+        if isinstance(top_p, AnthropicNotGiven):
+            top_p = NOT_GIVEN
 
         raw_client = self.raw_client  # 调用完 self.raw_client 后，self.model_id 会被赋值
         self.model_setting = self.backend_settings.models[self.model]
         if self.model_id is None:
             self.model_id = self.model_setting.id
 
-        if self.context_length_control == ContextLengthControlType.Latest:
+        if not skip_cutoff and self.context_length_control == ContextLengthControlType.Latest:
             messages = cutoff_messages(
                 messages,
                 max_count=self.model_setting.context_length,
@@ -460,6 +483,7 @@ class AsyncOpenAICompatibleChatClient(BaseAsyncChatClient):
                 stream=self.stream,
                 temperature=self.temperature,
                 max_tokens=max_tokens,
+                top_p=top_p,
                 **_stream_options_params,
                 **self.response_format,
                 **tools_params,
@@ -515,6 +539,7 @@ class AsyncOpenAICompatibleChatClient(BaseAsyncChatClient):
                 stream=self.stream,
                 temperature=self.temperature,
                 max_tokens=max_tokens,
+                top_p=top_p,
                 **self.response_format,
                 **tools_params,
                 **kwargs,

vectorvein/chat_clients/utils.py

@@ -201,6 +201,10 @@ def get_token_counts(text: str | dict, model: str = "", use_token_server_first:
         result = response.json()
         return result["data"]["total_tokens"]
     elif model.startswith("gemini"):
+        # TODO: gemini-exp-1114 暂时不支持，使用 gemini-1.5-flash 代替
+        if model == "gemini-exp-1114":
+            model = "gemini-1.5-flash"
+
         model_setting = settings.gemini.models[model]
         if len(model_setting.endpoints) == 0:
             return len(get_gpt_35_encoding().encode(text))
@@ -208,13 +212,14 @@ def get_token_counts(text: str | dict, model: str = "", use_token_server_first:
         if isinstance(endpoint_id, dict):
             endpoint_id = endpoint_id["endpoint_id"]
         endpoint = settings.get_endpoint(endpoint_id)
+
         base_url = f"{endpoint.api_base}/models/{model_setting.id}:countTokens"
         params = {"key": endpoint.api_key}
         request_body = {
             "contents": {
                 "role": "USER",
                 "parts": [
-                    {"text": "TEXT"},
+                    {"text": text},
                 ],
             },
         }

vectorvein/types/defaults.py

@@ -609,6 +609,13 @@ MINIMAX_MODELS: Final[Dict[str, Dict[str, Any]]] = {
         "function_call_available": True,
         "response_format_available": True,
     },
+    "abab7-preview": {
+        "id": "abab7-preview",
+        "context_length": 245760,
+        "max_output_tokens": 245760,
+        "function_call_available": True,
+        "response_format_available": True,
+    },
 }
 
 # Gemini models
@@ -616,7 +623,8 @@ GEMINI_DEFAULT_MODEL: Final[str] = "gemini-1.5-pro"
 GEMINI_MODELS: Final[Dict[str, Dict[str, Any]]] = {
     "gemini-1.5-pro": {
         "id": "gemini-1.5-pro",
-        "context_length": 1048576,
+        "context_length": 2097152,
+        "max_output_tokens": 8192,
         "function_call_available": True,
         "response_format_available": True,
         "native_multimodal": True,
@@ -624,6 +632,22 @@ GEMINI_MODELS: Final[Dict[str, Dict[str, Any]]] = {
     "gemini-1.5-flash": {
         "id": "gemini-1.5-flash",
         "context_length": 1048576,
+        "max_output_tokens": 8192,
+        "function_call_available": True,
+        "response_format_available": True,
+        "native_multimodal": True,
+    },
+    "gemini-exp-1114": {
+        "id": "gemini-exp-1114",
+        "context_length": 32767,
+        "function_call_available": True,
+        "response_format_available": True,
+        "native_multimodal": True,
+    },
+    "gemini-1.5-flash-8b": {
+        "id": "gemini-1.5-flash-8b",
+        "context_length": 1048576,
+        "max_output_tokens": 8192,
         "function_call_available": True,
         "response_format_available": True,
         "native_multimodal": True,

{vectorvein-0.1.56.dist-info → vectorvein-0.1.59.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: vectorvein
-Version: 0.1.56
+Version: 0.1.59
 Summary: VectorVein python SDK
 Author-Email: Anderson <andersonby@163.com>
 License: MIT

{vectorvein-0.1.56.dist-info → vectorvein-0.1.59.dist-info}/RECORD

@@ -1,24 +1,24 @@
-vectorvein-0.1.56.dist-info/METADATA,sha256=3YTghMyf-ZK9VzwhaqGwf_xuhKmVI5IA7SKI1O73Sww,633
-vectorvein-0.1.56.dist-info/WHEEL,sha256=thaaA2w1JzcGC48WYufAs8nrYZjJm8LqNfnXFOFyCC4,90
-vectorvein-0.1.56.dist-info/entry_points.txt,sha256=6OYgBcLyFCUgeqLgnvMyOJxPCWzgy7se4rLPKtNonMs,34
+vectorvein-0.1.59.dist-info/METADATA,sha256=GCU-0oftBD1QtnzFXvknbCMq1mdXwBiduQT_B6y95eI,633
+vectorvein-0.1.59.dist-info/WHEEL,sha256=thaaA2w1JzcGC48WYufAs8nrYZjJm8LqNfnXFOFyCC4,90
+vectorvein-0.1.59.dist-info/entry_points.txt,sha256=6OYgBcLyFCUgeqLgnvMyOJxPCWzgy7se4rLPKtNonMs,34
 vectorvein/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 vectorvein/chat_clients/__init__.py,sha256=Oev7Lv1DIEWCMD-2Pm7e2cwzX7JFQTnIK-j6o4iUuyQ,17725
-vectorvein/chat_clients/anthropic_client.py,sha256=G68JGM98E0pYyi8Tjvo4VQtnug9ncugFbb4d0DrPVQo,36122
+vectorvein/chat_clients/anthropic_client.py,sha256=phDFgXPV-eNla7ZFPdcZx1fPOWlAFWc3C_mP9VcNkO0,37439
 vectorvein/chat_clients/baichuan_client.py,sha256=CVMvpgjdrZGv0BWnTOBD-f2ufZ3wq3496wqukumsAr4,526
-vectorvein/chat_clients/base_client.py,sha256=0Uj0e-JR0a68sRS_WfUMVd91Av7lzJh6-DukjutlaD0,9497
+vectorvein/chat_clients/base_client.py,sha256=N1Swm6b9Gos7zLSH-qCSxgnDRCHPmuWZcw_H9zVnGJs,10297
 vectorvein/chat_clients/deepseek_client.py,sha256=3qWu01NlJAP2N-Ff62d5-CZXZitlizE1fzb20LNetig,526
-vectorvein/chat_clients/gemini_client.py,sha256=DlQI5p8qG5erThMb17ojB2gofzTkwVy3veFC8_Cbpc4,19902
+vectorvein/chat_clients/gemini_client.py,sha256=e7xZdZm0-W2iXy3S-J5b1bO9YqhGxcv0Y5HPYcQnDds,21098
 vectorvein/chat_clients/groq_client.py,sha256=Uow4pgdmFi93ZQSoOol2-0PhhqkW-S0XuSldvppz5U4,498
 vectorvein/chat_clients/local_client.py,sha256=55nOsxzqUf79q3Y14MKROA71zxhsT7p7FsDZ89rts2M,422
-vectorvein/chat_clients/minimax_client.py,sha256=3HetFZbmgvEzWFaH_Gbj99lsh9DmIOhfpVdbV4HxrSI,19116
+vectorvein/chat_clients/minimax_client.py,sha256=uNYz3ccJOhTAgzkDNvWfM4_LaBaOpHjrne1YNieba28,20122
 vectorvein/chat_clients/mistral_client.py,sha256=1aKSylzBDaLYcFnaBIL4-sXSzWmXfBeON9Q0rq-ziWw,534
 vectorvein/chat_clients/moonshot_client.py,sha256=gbu-6nGxx8uM_U2WlI4Wus881rFRotzHtMSoYOcruGU,526
 vectorvein/chat_clients/openai_client.py,sha256=Nz6tV45pWcsOupxjnsRsGTicbQNJWIZyxuJoJ5DGMpg,527
-vectorvein/chat_clients/openai_compatible_client.py,sha256=D2VmhpDVct4w2y58s87An7x22koOdkxSOKR2-so9lJI,22044
+vectorvein/chat_clients/openai_compatible_client.py,sha256=hbSggW5itvTRk3Qb3rejNTt3vd0r6R95irLTjeSzm2g,23102
 vectorvein/chat_clients/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 vectorvein/chat_clients/qwen_client.py,sha256=-ryh-m9PgsO0fc4ulcCmPTy1155J8YUy15uPoJQOHA0,513
 vectorvein/chat_clients/stepfun_client.py,sha256=zsD2W5ahmR4DD9cqQTXmJr3txrGuvxbRWhFlRdwNijI,519
-vectorvein/chat_clients/utils.py,sha256=UqDV4ljwZRbndZdkyHkIBis1ciRt5gjPFOJ0vnR2Uas,26669
+vectorvein/chat_clients/utils.py,sha256=6rjXUv39XMIEKKmA-vG8HzOHntCVeFvlrirZcduKCIA,26828
 vectorvein/chat_clients/xai_client.py,sha256=eLFJJrNRJ-ni3DpshODcr3S1EJQLbhVwxyO1E54LaqM,491
 vectorvein/chat_clients/yi_client.py,sha256=RNf4CRuPJfixrwLZ3-DEc3t25QDe1mvZeb9sku2f8Bc,484
 vectorvein/chat_clients/zhipuai_client.py,sha256=Ys5DSeLCuedaDXr3PfG1EW2zKXopt-awO2IylWSwY0s,519
@@ -26,11 +26,11 @@ vectorvein/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 vectorvein/server/token_server.py,sha256=36F9PKSNOX8ZtYBXY_l-76GQTpUSmQ2Y8EMy1H7wtdQ,1353
 vectorvein/settings/__init__.py,sha256=g01y74x0k2JEAqNpRGG0PDs0NTULjOAZV6HRhydPX1c,3874
 vectorvein/settings/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-vectorvein/types/defaults.py,sha256=Kf7kdY2-dbcWR_oQkqoHfBLtr2kg-M_Va3-37TlZU-o,22941
+vectorvein/types/defaults.py,sha256=DC0fJ2MtXYNXiWkDdnpGYXuwCoSOcuB6PseI_y-VDo0,23730
 vectorvein/types/enums.py,sha256=7KTJSVtQueImmbr1fSwv3rQVtc0RyMWXJmoE2tDOaso,1667
 vectorvein/types/exception.py,sha256=gnW4GnJ76jND6UGnodk9xmqkcbeS7Cz2rvncA2HpD5E,69
 vectorvein/types/llm_parameters.py,sha256=5o-C_yXxxQWZy_e8OWowB2107GTS-Eawx4Mvb1q55Co,5256
 vectorvein/types/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 vectorvein/utilities/media_processing.py,sha256=CTRq-lGlFkFgP_FSRhNwF_qUgmOrXPf2_1Ok9HY42_g,5887
 vectorvein/utilities/retry.py,sha256=6KFS9R2HdhqM3_9jkjD4F36ZSpEx2YNFGOVlpOsUetM,2208
-vectorvein-0.1.56.dist-info/RECORD,,
+vectorvein-0.1.59.dist-info/RECORD,,