unique_toolkit 0.5.39__py3-none-any.whl → 0.5.41__py3-none-any.whl

unique_toolkit/language_model/schemas.py

@@ -197,29 +197,28 @@ class LanguageModelTokenLimits(BaseModel):
     token_limit_input: Optional[int] = None
     token_limit_output: Optional[int] = None
 
+    fraction_input: float = Field(default=0.4, le=1, ge=0)
+
     @model_validator(mode="after")
-    def validate_model(self):
-        token_limit = self.token_limit
-        token_limit_input = self.token_limit_input
-        token_limit_output = self.token_limit_output
-
-        if (
-            token_limit is None
-            and token_limit_input is None
-            and token_limit_output is None
-        ):
-            raise ValueError(
-                "At least one of token_limit, token_limit_input or token_limit_output must be set"
-            )
-
-        if (
-            token_limit is None
-            and token_limit_input is not None
-            and token_limit_output is not None
-        ):
-            self.token_limit = token_limit_input + token_limit_output
-
-        return self
+    def check_required_fields(self):
+        # Best case input and output is determined
+        if self.token_limit_input and self.token_limit_output:
+            self.token_limit = self.token_limit_input + self.token_limit_output
+            self.fraction_input = self.token_limit_input / self.token_limit
+            return self
+
+        # Deal with case where only token_limit and optional fraction_input is given
+        if self.token_limit:
+            if not self.fraction_input:
+                self.fraction_input = 0.4
+
+            self.token_limit_input = self.fraction_input * self.token_limit
+            self.token_limit_output = (1 - self.fraction_input) * self.token_limit
+            return self
+
+        raise ValueError(
+            'Either "token_limit_input" and "token_limit_output" must be provided together, or "token_limit" must be provided.'
+        )
 
 
 class LanguageModelToolParameterProperty(BaseModel):

unique_toolkit/language_model/service.py

@@ -37,6 +37,7 @@ class LanguageModelService(BaseService):
         temperature: float = DEFAULT_COMPLETE_TEMPERATURE,
         timeout: int = DEFAULT_COMPLETE_TIMEOUT,
         tools: Optional[list[LanguageModelTool]] = None,
+        other_options: Optional[dict] = None,
     ):
         """
         Calls the completion endpoint synchronously without streaming the response.
@@ -47,25 +48,26 @@ class LanguageModelService(BaseService):
             temperature (float): The temperature value. Defaults to 0.
             timeout (int): The timeout value in milliseconds. Defaults to 240_000.
             tools (Optional[list[LanguageModelTool]]): The tools to use. Defaults to None.
+            other_options (Optional[dict]): The other options to use. Defaults to None.
 
         Returns:
             LanguageModelResponse: The LanguageModelResponse object.
         """
-        options = self._add_tools_to_options({}, tools)
-        options["temperature"] = temperature
-        messages = messages.model_dump(exclude_none=True)
-        model = (
-            model_name.name if isinstance(model_name, LanguageModelName) else model_name
+        options, model, messages_dict, _ = self.prepare_completion_params_util(
+            messages=messages,
+            model_name=model_name,
+            temperature=temperature,
+            tools=tools,
+            other_options=other_options,
         )
 
         try:
             response = unique_sdk.ChatCompletion.create(
                 company_id=self.event.company_id,
-                # TODO change or extend types in unique_sdk
                 model=model,
                 messages=cast(
                     list[unique_sdk.Integrated.ChatCompletionRequestMessage],
-                    messages,
+                    messages_dict,
                 ),
                 timeout=timeout,
                 options=options,  # type: ignore
@@ -84,6 +86,7 @@ class LanguageModelService(BaseService):
         temperature: float = DEFAULT_COMPLETE_TEMPERATURE,
         timeout: int = DEFAULT_COMPLETE_TIMEOUT,
         tools: Optional[list[LanguageModelTool]] = None,
+        other_options: Optional[dict] = None,
         logger: Optional[logging.Logger] = logging.getLogger(__name__),
     ) -> LanguageModelResponse:
         """
@@ -100,6 +103,7 @@ class LanguageModelService(BaseService):
             temperature (float): The temperature setting for the completion. Defaults to 0.
             timeout (int): The timeout value in milliseconds for the request. Defaults to 240_000.
             tools (Optional[list[LanguageModelTool]]): Optional list of tools to include in the request.
+            other_options (Optional[dict]): The other options to use. Defaults to None.
             logger (Optional[logging.Logger], optional): The logger used to log errors. Defaults to the logger for the current module.
 
         Returns:
@@ -108,19 +112,21 @@ class LanguageModelService(BaseService):
         Raises:
             Exception: If an error occurs during the request, an exception is raised and logged.
         """
-        options = cls._add_tools_to_options({}, tools)
-        options["temperature"] = temperature
-        messages = messages.model_dump(exclude_none=True, exclude={"tool_calls"})
-        model = (
-            model_name.name if isinstance(model_name, LanguageModelName) else model_name
+        options, model, messages_dict, _ = cls.prepare_completion_params_util(
+            messages=messages,
+            model_name=model_name,
+            temperature=temperature,
+            tools=tools,
+            other_options=other_options,
         )
+
         try:
             response = await unique_sdk.ChatCompletion.create_async(
                 company_id=company_id,
                 model=model,
                 messages=cast(
                     list[unique_sdk.Integrated.ChatCompletionRequestMessage],
-                    messages,
+                    messages_dict,
                 ),
                 timeout=timeout,
                 options=options,  # type: ignore
@@ -137,6 +143,7 @@ class LanguageModelService(BaseService):
         temperature: float = DEFAULT_COMPLETE_TEMPERATURE,
         timeout: int = DEFAULT_COMPLETE_TIMEOUT,
         tools: Optional[list[LanguageModelTool]] = None,
+        other_options: Optional[dict] = None,
     ) -> LanguageModelResponse:
         """
         Calls the completion endpoint asynchronously without streaming the response.
@@ -151,7 +158,7 @@ class LanguageModelService(BaseService):
             temperature (float): The temperature setting for the completion. Defaults to 0.0.
             timeout (int): The timeout value in milliseconds for the request. Defaults to 240,000.
             tools (Optional[list[LanguageModelTool]]): Optional list of tools to include in the request.
-
+            other_options (Optional[dict]): The other options to use. Defaults to None.
         Returns:
             LanguageModelResponse: The response object containing the completed result.
 
@@ -165,6 +172,7 @@ class LanguageModelService(BaseService):
             temperature=temperature,
             timeout=timeout,
             tools=tools,
+            other_options=other_options,
             logger=self.logger,
         )
 
@@ -178,6 +186,7 @@ class LanguageModelService(BaseService):
         timeout: int = DEFAULT_COMPLETE_TIMEOUT,
         tools: Optional[list[LanguageModelTool]] = None,
         start_text: Optional[str] = None,
+        other_options: Optional[dict] = None,
     ):
         """
         Streams a completion in the chat session synchronously.
@@ -191,16 +200,19 @@ class LanguageModelService(BaseService):
             timeout (int): The timeout value in milliseconds. Defaults to 240_000.
             tools (Optional[list[LanguageModelTool]]): The tools to use. Defaults to None.
             start_text (Optional[str]): The start text. Defaults to None.
-
+            other_options (Optional[dict]): The other options to use. Defaults to None.
         Returns:
             The LanguageModelStreamResponse object once the stream has finished.
         """
-        options = self._add_tools_to_options({}, tools)
-        options["temperature"] = temperature
-        search_context = self._to_search_context(content_chunks)
-        messages = messages.model_dump(exclude_none=True, by_alias=True)
-        model = (
-            model_name.name if isinstance(model_name, LanguageModelName) else model_name
+        options, model, messages_dict, search_context = (
+            self.prepare_completion_params_util(
+                messages=messages,
+                model_name=model_name,
+                temperature=temperature,
+                tools=tools,
+                other_options=other_options,
+                content_chunks=content_chunks,
+            )
         )
 
         try:
@@ -211,11 +223,10 @@ class LanguageModelService(BaseService):
                 userMessageId=self.event.payload.user_message.id,
                 messages=cast(
                     list[unique_sdk.Integrated.ChatCompletionRequestMessage],
-                    messages,
+                    messages_dict,
                 ),
                 chatId=self.event.payload.chat_id,
                 searchContext=search_context,
-                # TODO change or extend types in unique_sdk
                 model=model,
                 timeout=timeout,
                 assistantId=self.event.payload.assistant_id,
@@ -238,6 +249,7 @@ class LanguageModelService(BaseService):
         timeout: int = DEFAULT_COMPLETE_TIMEOUT,
         tools: Optional[list[LanguageModelTool]] = None,
         start_text: Optional[str] = None,
+        other_options: Optional[dict] = None,
     ):
         """
         Streams a completion in the chat session asynchronously.
@@ -251,17 +263,19 @@ class LanguageModelService(BaseService):
             timeout (int): The timeout value in milliseconds. Defaults to 240_000.
             tools (Optional[list[LanguageModelTool]]): The tools to use. Defaults to None.
             start_text (Optional[str]): The start text. Defaults to None.
-
+            other_options (Optional[dict]): The other options to use. Defaults to None.
         Returns:
             The LanguageModelStreamResponse object once the stream has finished.
         """
-
-        options = self._add_tools_to_options({}, tools)
-        options["temperature"] = temperature
-        search_context = self._to_search_context(content_chunks)
-        messages = messages.model_dump(exclude_none=True, by_alias=True)
-        model = (
-            model_name.name if isinstance(model_name, LanguageModelName) else model_name
+        options, model, messages_dict, search_context = (
+            self.prepare_completion_params_util(
+                messages=messages,
+                model_name=model_name,
+                temperature=temperature,
+                tools=tools,
+                other_options=other_options,
+                content_chunks=content_chunks,
+            )
         )
 
         try:
@@ -272,7 +286,7 @@ class LanguageModelService(BaseService):
                 userMessageId=self.event.payload.user_message.id,
                 messages=cast(
                     list[unique_sdk.Integrated.ChatCompletionRequestMessage],
-                    messages,
+                    messages_dict,
                 ),
                 chatId=self.event.payload.chat_id,
                 searchContext=search_context,
@@ -280,7 +294,6 @@ class LanguageModelService(BaseService):
                 timeout=timeout,
                 assistantId=self.event.payload.assistant_id,
                 debugInfo=debug_info,
-                # TODO change or extend types in unique_sdk
                 options=options,  # type: ignore
                 startText=start_text,
             )
@@ -321,3 +334,47 @@ class LanguageModelService(BaseService):
                 for tool in tools
             ]
         return options
+
+    @classmethod
+    def prepare_completion_params_util(
+        cls,
+        messages: LanguageModelMessages,
+        model_name: LanguageModelName | str,
+        temperature: float,
+        tools: Optional[list[LanguageModelTool]] = None,
+        other_options: Optional[dict] = None,
+        content_chunks: Optional[list[ContentChunk]] = None,
+    ) -> tuple[dict, str, dict, Optional[dict]]:
+        """
+        Prepares common parameters for completion requests.
+
+        Returns:
+            tuple containing:
+            - options (dict): Combined options including tools and temperature
+            - model (str): Resolved model name
+            - messages_dict (dict): Processed messages
+            - search_context (Optional[dict]): Processed content chunks if provided
+        """
+
+        options = cls._add_tools_to_options({}, tools)
+        options["temperature"] = temperature
+        if other_options:
+            options.update(other_options)
+
+        model = (
+            model_name.name if isinstance(model_name, LanguageModelName) else model_name
+        )
+
+        # Different methods need different message dump parameters
+        messages_dict = messages.model_dump(
+            exclude_none=True,
+            by_alias=content_chunks is not None,  # Use by_alias for streaming methods
+        )
+
+        search_context = (
+            LanguageModelService._to_search_context(content_chunks)
+            if content_chunks is not None
+            else None
+        )
+
+        return options, model, messages_dict, search_context

{unique_toolkit-0.5.39.dist-info → unique_toolkit-0.5.41.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: unique_toolkit
-Version: 0.5.39
+Version: 0.5.41
 Summary: 
 License: Proprietary
 Author: Martin Fadler
@@ -100,7 +100,11 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/), 
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.5.41] - 2024-12-11
+- Update `LanguageModelTokenLimits` includes a fraction_input now to always have input/output token limits available.
 
+## [0.5.40] - 2024-12-11
+- Add `other_options` to `LanguageModelService.complete`, `LanguageModelService.complete_async`, `LanguageModelService.stream_complete` and `LanguageModelService.stream_complete_async`
 
 ## [0.5.39] - 2024-12-09
 - Add `contentIds` to `Search.create` and `Search.create_async`

{unique_toolkit-0.5.39.dist-info → unique_toolkit-0.5.41.dist-info}/RECORD

@@ -37,10 +37,10 @@ unique_toolkit/evaluators/output_parser.py,sha256=eI72qkzK1dZyUvnfP2SOAQCGBj_-Pw
 unique_toolkit/evaluators/schemas.py,sha256=Jaue6Uhx75X1CyHKWj8sT3RE1JZXTqoLtfLt2xQNCX8,2507
 unique_toolkit/language_model/__init__.py,sha256=YuhyczGPj6w9xX-sOVUhmozvzIFxcckHFEkeMBecr5s,1784
 unique_toolkit/language_model/infos.py,sha256=kQK6F3r8xTN7oT6b39J7rxW-Y4iPXjx_Fr9bCOVQdm0,12509
-unique_toolkit/language_model/schemas.py,sha256=wg_ly66UvLOVNdrq8xr-0uN0HPY41UIk5mVPxdG-VGs,6687
-unique_toolkit/language_model/service.py,sha256=s5X6EStyYumiYqlD9gkW4GANif18d9QZOMysmUSfv8M,13433
+unique_toolkit/language_model/schemas.py,sha256=-5yTJsEUMKSc6qZ1zdBU9Eewo9qQlpqlVssKHIwUY64,6925
+unique_toolkit/language_model/service.py,sha256=brNCPRA0XxgqHi2rI5i2lyFCkUiw4MNMe1VaR3UgWmY,15500
 unique_toolkit/language_model/utils.py,sha256=bPQ4l6_YO71w-zaIPanUUmtbXC1_hCvLK0tAFc3VCRc,1902
-unique_toolkit-0.5.39.dist-info/LICENSE,sha256=GlN8wHNdh53xwOPg44URnwag6TEolCjoq3YD_KrWgss,193
-unique_toolkit-0.5.39.dist-info/METADATA,sha256=30jmpa5iHACeXTpmNB0XyY6GSanJvFAi10S8K0qg4Ew,14411
-unique_toolkit-0.5.39.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
-unique_toolkit-0.5.39.dist-info/RECORD,,
+unique_toolkit-0.5.41.dist-info/LICENSE,sha256=GlN8wHNdh53xwOPg44URnwag6TEolCjoq3YD_KrWgss,193
+unique_toolkit-0.5.41.dist-info/METADATA,sha256=im-lBhwsW2ZgUZoA6mfyogye1c6ZIFXqgaw_Ilf2yLs,14764
+unique_toolkit-0.5.41.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
+unique_toolkit-0.5.41.dist-info/RECORD,,