unique_toolkit 0.7.11__py3-none-any.whl → 0.7.14__py3-none-any.whl

unique_toolkit/_common/validators.py

@@ -1,8 +1,57 @@
-from unique_toolkit.language_model import LanguageModel, LanguageModelName
+from typing import Annotated
 
+from pydantic import BeforeValidator, PlainSerializer
 
-def validate_and_init_language_model(value: LanguageModelName | LanguageModel | str):
-    if isinstance(value, LanguageModel):
-        return value
+from unique_toolkit.language_model import LanguageModelName
+from unique_toolkit.language_model.infos import (
+    LanguageModelInfo,
+    LanguageModelProvider,
+)
 
-    return LanguageModel(value)
+# TODO @klcd: Inform on deprecation of str as input
+LMI = Annotated[
+    LanguageModelInfo,
+    BeforeValidator(
+        lambda v: validate_and_init_language_model_info(v),
+        json_schema_input_type=str | LanguageModelName | LanguageModelInfo,
+    ),
+    PlainSerializer(
+        lambda v: serialize_lmi(v),
+        when_used="json",
+        return_type=str | LanguageModelInfo,
+    ),
+]
+
+
+def serialize_lmi(model: LanguageModelInfo) -> str | LanguageModelInfo:
+    if model.provider == LanguageModelProvider.CUSTOM:
+        return model
+
+    return model.name
+
+
+def validate_and_init_language_model_info(
+    v: str | LanguageModelName | LanguageModelInfo,
+) -> LanguageModelInfo:
+    """Validate and initialize a LanguageModelInfo object.
+
+    Args:
+        v: The input value to validate and initialize.
+
+    Returns:
+        LanguageModelInfo: The validated and initialized LanguageModelInfo object.
+
+    """
+    if isinstance(v, LanguageModelName):
+        return LanguageModelInfo.from_name(v)
+    if isinstance(v, str):
+        if v in [name.value for name in LanguageModelName]:
+            return LanguageModelInfo.from_name(LanguageModelName(v))
+
+        return LanguageModelInfo(
+            name=v,
+            version="custom",
+            provider=LanguageModelProvider.CUSTOM,
+        )
+
+    return v

unique_toolkit/evaluators/config.py

@@ -1,35 +1,26 @@
 from humps import camelize
-from pydantic import BaseModel, ConfigDict, field_validator
+from pydantic import BaseModel, ConfigDict
 
-from unique_toolkit._common.validators import validate_and_init_language_model
+from unique_toolkit._common.validators import LMI, LanguageModelInfo
 from unique_toolkit.evaluators.schemas import (
     EvaluationMetricName,
 )
 from unique_toolkit.language_model.infos import (
-    LanguageModel,
     LanguageModelName,
 )
 
-model_config = ConfigDict(
-    alias_generator=camelize,
-    populate_by_name=True,
-    arbitrary_types_allowed=True,
-    validate_default=True,
-    json_encoders={LanguageModel: lambda v: v.display_name},
-)
-
 
 class EvaluationMetricConfig(BaseModel):
-    model_config = model_config
+    model_config = ConfigDict(
+        alias_generator=camelize,
+        populate_by_name=True,
+        validate_default=True,
+    )
 
     enabled: bool = False
     name: EvaluationMetricName
-    language_model: LanguageModel = LanguageModel(
-        LanguageModelName.AZURE_GPT_35_TURBO_0125
+    language_model: LMI = LanguageModelInfo.from_name(
+        LanguageModelName.AZURE_GPT_35_TURBO_0125,
     )
     custom_prompts: dict[str, str] = {}
     score_to_emoji: dict[str, str] = {}
-
-    @field_validator("language_model", mode="before")
-    def validate_language_model(cls, value: LanguageModelName | LanguageModel):
-        return validate_and_init_language_model(value)

unique_toolkit/evaluators/context_relevancy/constants.py

@@ -7,7 +7,7 @@ from unique_toolkit.evaluators.schemas import (
     EvaluationMetricInputFieldName,
     EvaluationMetricName,
 )
-from unique_toolkit.language_model.infos import LanguageModel
+from unique_toolkit.language_model.infos import LanguageModelInfo
 from unique_toolkit.language_model.service import LanguageModelName
 
 SYSTEM_MSG_KEY = "systemPrompt"
@@ -23,7 +23,9 @@ context_relevancy_required_input_fields = [
 default_config = EvaluationMetricConfig(
     enabled=False,
     name=EvaluationMetricName.CONTEXT_RELEVANCY,
-    language_model=LanguageModel(LanguageModelName.AZURE_GPT_35_TURBO_0125),
+    language_model=LanguageModelInfo.from_name(
+        LanguageModelName.AZURE_GPT_35_TURBO_0125
+    ),
     score_to_emoji={"LOW": "🟢", "MEDIUM": "🟡", "HIGH": "🔴"},
     custom_prompts={
         SYSTEM_MSG_KEY: CONTEXT_RELEVANCY_METRIC_SYSTEM_MSG,

unique_toolkit/evaluators/context_relevancy/utils.py

@@ -22,6 +22,7 @@ from unique_toolkit.evaluators.schemas import (
     EvaluationMetricName,
     EvaluationMetricResult,
 )
+from unique_toolkit.language_model import LanguageModelName
 from unique_toolkit.language_model.schemas import (
     LanguageModelMessages,
     LanguageModelSystemMessage,
@@ -34,12 +35,12 @@ logger = logging.getLogger(__name__)
 
 async def check_context_relevancy_async(
     company_id: str,
-    input: EvaluationMetricInput,
+    evaluation_metric_input: EvaluationMetricInput,
     config: EvaluationMetricConfig,
     logger: logging.Logger = logger,
 ) -> EvaluationMetricResult | None:
-    """
-    Analyzes the relevancy of the context provided for the given input and output.
+    """Analyzes the relevancy of the context provided for the given evaluation_metric_input and output.
+
     The analysis classifies the context relevancy level as:
     - low
     - medium
@@ -47,14 +48,14 @@ async def check_context_relevancy_async(
 
     This method performs the following steps:
     1. Logs the start of the analysis using the provided `logger`.
-    2. Validates the required fields in the `input` data.
+    2. Validates the required fields in the `evaluation_metric_input` data.
     3. Retrieves the messages using the `_get_msgs` method.
     4. Calls `LanguageModelService.complete_async_util` to get a completion result.
     5. Parses and returns the evaluation metric result based on the content of the completion result.
 
     Args:
         company_id (str): The company ID for the analysis.
-        input (EvaluationMetricInput): The input data used for evaluation, including the generated output and reference information.
+        evaluation_metric_input (EvaluationMetricInput): The evaluation_metric_input data used for evaluation, including the generated output and reference information.
         config (EvaluationMetricConfig): Configuration settings for the evaluation.
         logger (Optional[logging.Logger], optional): The logger used for logging information and errors. Defaults to the logger for the current module.
 
@@ -63,13 +64,23 @@ async def check_context_relevancy_async(
 
     Raises:
         EvaluatorException: If required fields are missing or an error occurs during the evaluation.
+
     """
-    model_name = config.language_model.name
-    logger.info(f"Analyzing context relevancy with {model_name}.")
+    model_group_name = (
+        config.language_model.name.value
+        if isinstance(config.language_model.name, LanguageModelName)
+        else config.language_model.name
+    )
+    logger.info(f"Analyzing context relevancy with {model_group_name}.")
 
-    input.validate_required_fields(context_relevancy_required_input_fields)
+    evaluation_metric_input.validate_required_fields(
+        context_relevancy_required_input_fields,
+    )
 
-    if input.context_texts and len(input.context_texts) == 0:
+    if (
+        evaluation_metric_input.context_texts
+        and len(evaluation_metric_input.context_texts) == 0
+    ):
         error_message = "No context texts provided."
         raise EvaluatorException(
             user_message=error_message,
@@ -77,11 +88,11 @@ async def check_context_relevancy_async(
         )
 
     try:
-        msgs = _get_msgs(input, config)
+        msgs = _get_msgs(evaluation_metric_input, config)
         result = await LanguageModelService.complete_async_util(
             company_id=company_id,
             messages=msgs,
-            model_name=model_name,
+            model_name=model_group_name,
         )
         result_content = result.choices[0].message.content
         if not result_content:
@@ -104,25 +115,28 @@ async def check_context_relevancy_async(
 
 
 def _get_msgs(
-    input: EvaluationMetricInput,
+    evaluation_metric_input: EvaluationMetricInput,
     config: EvaluationMetricConfig,
-):
-    """
-    Composes the messages for context relevancy analysis based on the provided input and configuration.
+) -> LanguageModelMessages:
+    """Composes the messages for context relevancy analysis.
+
+    The messages are based on the provided evaluation_metric_input and configuration.
 
     Args:
-        input (EvaluationMetricInput): The input data that includes context texts for the analysis.
+        evaluation_metric_input (EvaluationMetricInput): The evaluation_metric_input data that includes context texts for the analysis.
         config (EvaluationMetricConfig): The configuration settings for composing messages.
 
     Returns:
-        LanguageModelMessages: The composed messages as per the provided input and configuration.
+        LanguageModelMessages: The composed messages as per the provided evaluation_metric_input and configuration.
+
     """
     system_msg_content = _get_system_prompt(config)
     system_msg = LanguageModelSystemMessage(content=system_msg_content)
 
     user_msg_templ = Template(_get_user_prompt(config))
     user_msg_content = user_msg_templ.substitute(
-        input_text=input.input_text, contexts_text=input.get_joined_context_texts()
+        evaluation_metric_input_text=evaluation_metric_input.evaluation_metric_input_text,
+        contexts_text=evaluation_metric_input.get_joined_context_texts(),
     )
     user_msg = LanguageModelUserMessage(content=user_msg_content)
     return LanguageModelMessages([system_msg, user_msg])

unique_toolkit/evaluators/hallucination/constants.py

@@ -10,7 +10,7 @@ from unique_toolkit.evaluators.schemas import (
     EvaluationMetricName,
 )
 from unique_toolkit.language_model.infos import (
-    LanguageModel,
+    LanguageModelInfo,
     LanguageModelName,
 )
 
@@ -23,7 +23,7 @@ USER_MSG_DEFAULT_KEY = "userPromptDefault"
 hallucination_metric_default_config = EvaluationMetricConfig(
     enabled=False,
     name=EvaluationMetricName.HALLUCINATION,
-    language_model=LanguageModel(LanguageModelName.AZURE_GPT_4_0613),
+    language_model=LanguageModelInfo.from_name(LanguageModelName.AZURE_GPT_4_0613),
     score_to_emoji={"LOW": "🟢", "MEDIUM": "🟡", "HIGH": "🔴"},
     custom_prompts={
         SYSTEM_MSG_KEY: HALLUCINATION_METRIC_SYSTEM_MSG,

unique_toolkit/evaluators/hallucination/utils.py

@@ -20,6 +20,7 @@ from unique_toolkit.evaluators.schemas import (
     EvaluationMetricName,
     EvaluationMetricResult,
 )
+from unique_toolkit.language_model import LanguageModelName
 from unique_toolkit.language_model.schemas import (
     LanguageModelMessages,
     LanguageModelSystemMessage,
@@ -43,8 +44,9 @@ async def check_hallucination_async(
     config: EvaluationMetricConfig,
     logger: logging.Logger = logger,
 ) -> EvaluationMetricResult | None:
-    """
-    Analyzes the level of hallucination in the generated output by comparing it with the provided input
+    """Analyze the level of hallucination in the generated output.
+
+    by comparing it with the provided input
     and the contexts or history. The analysis classifies the hallucination level as:
     - low
     - medium
@@ -72,16 +74,23 @@ async def check_hallucination_async(
 
     Raises:
         EvaluatorException: If the context texts are empty, required fields are missing, or an error occurs during the evaluation.
+
     """
-    model_name = config.language_model.name
-    logger.info(f"Analyzing level of hallucination with {model_name}.")
+    model_group_name = (
+        config.language_model.name.value
+        if isinstance(config.language_model.name, LanguageModelName)
+        else config.language_model.name
+    )
+    logger.info(f"Analyzing level of hallucination with {model_group_name}.")
 
     input.validate_required_fields(hallucination_required_input_fields)
 
     try:
         msgs = _get_msgs(input, config, logger)
         result = await LanguageModelService.complete_async_util(
-            company_id=company_id, messages=msgs, model_name=model_name
+            company_id=company_id,
+            messages=msgs,
+            model_name=model_group_name,
         )
         result_content = result.choices[0].message.content
         if not result_content:
@@ -104,71 +113,72 @@ async def check_hallucination_async(
 
 
 def _get_msgs(
-    input: EvaluationMetricInput,
+    evaluation_metric_input: EvaluationMetricInput,
     config: EvaluationMetricConfig,
     logger: logging.Logger,
 ):
-    """
-    Composes the messages for hallucination analysis based on the provided input and configuration.
+    """Composes the messages for hallucination analysis based on the provided evaluation_metric_input and configuration.
 
     This method decides how to compose the messages based on the availability of context texts and history
-    message texts in the `input`
+    message texts in the `evaluation_metric_input`
 
     Args:
-        input (EvaluationMetricInput): The input data that includes context texts and history message texts
+        evaluation_metric_input (EvaluationMetricInput): The evaluation_metric_input data that includes context texts and history message texts
                                       for the analysis.
         config (EvaluationMetricConfig): The configuration settings for composing messages.
         logger (Optional[logging.Logger], optional): The logger used for logging debug information.
                                                      Defaults to the logger for the current module.
 
     Returns:
-        The composed messages as per the provided input and configuration. The exact type and structure
+        The composed messages as per the provided evaluation_metric_input and configuration. The exact type and structure
         depend on the implementation of the `compose_msgs` and `compose_msgs_default` methods.
 
     """
-    if input.context_texts or input.history_messages:
+    if (
+        evaluation_metric_input.context_texts
+        or evaluation_metric_input.history_messages
+    ):
         logger.debug("Using context / history for hallucination evaluation.")
-        return _compose_msgs(input, config)
-    else:
-        logger.debug("No contexts and history provided for hallucination evaluation.")
-        return _compose_msgs_default(input, config)
+        return _compose_msgs(evaluation_metric_input, config)
+    logger.debug("No contexts and history provided for hallucination evaluation.")
+    return _compose_msgs_default(evaluation_metric_input, config)
 
 
 def _compose_msgs(
-    input: EvaluationMetricInput,
+    evaluation_metric_input: EvaluationMetricInput,
     config: EvaluationMetricConfig,
 ):
-    """
-    Composes the hallucination analysis messages.
-    """
+    """Composes the hallucination analysis messages."""
     system_msg_content = _get_system_prompt_with_contexts(config)
     system_msg = LanguageModelSystemMessage(content=system_msg_content)
 
     user_msg_templ = Template(_get_user_prompt_with_contexts(config))
     user_msg_content = user_msg_templ.substitute(
-        input_text=input.input_text,
-        contexts_text=input.get_joined_context_texts(tag_name="reference"),
-        history_messages_text=input.get_joined_history_texts(tag_name="conversation"),
-        output_text=input.output_text,
+        evaluation_metric_input_text=evaluation_metric_input.evaluation_metric_input_text,
+        contexts_text=evaluation_metric_input.get_joined_context_texts(
+            tag_name="reference",
+        ),
+        history_messages_text=evaluation_metric_input.get_joined_history_texts(
+            tag_name="conversation",
+        ),
+        output_text=evaluation_metric_input.output_text,
     )
     user_msg = LanguageModelUserMessage(content=user_msg_content)
     return LanguageModelMessages([system_msg, user_msg])
 
 
 def _compose_msgs_default(
-    input: EvaluationMetricInput,
+    evaluation_metric_input: EvaluationMetricInput,
     config: EvaluationMetricConfig,
 ):
-    """
-    Composes the hallucination analysis prompt without messages.
-    """
+    """Composes the hallucination analysis prompt without messages."""
     system_msg_content = _get_system_prompt_default(config)
     system_msg = LanguageModelSystemMessage(content=system_msg_content)
 
     user_msg_templ = Template(_get_user_prompt_default(config))
     user_msg_content = user_msg_templ.substitute(
-        input_text=input.input_text,
-        output_text=input.output_text,
+        evaluation_metric_input_text=evaluation_metric_input.evaluation_metric_input_text,
+        output_text=evaluation_metric_input.output_text,
     )
     user_msg = LanguageModelUserMessage(content=user_msg_content)
     return LanguageModelMessages([system_msg, user_msg])

unique_toolkit/language_model/functions.py

@@ -1,5 +1,5 @@
 import logging
-from typing import Type, cast
+from typing import cast
 
 import unique_sdk
 from pydantic import BaseModel
@@ -29,11 +29,10 @@ def complete(
     timeout: int = DEFAULT_COMPLETE_TIMEOUT,
     tools: list[LanguageModelTool] | None = None,
     other_options: dict | None = None,
-    structured_output_model: Type[BaseModel] | None = None,
+    structured_output_model: type[BaseModel] | None = None,
     structured_output_enforce_schema: bool = False,
 ) -> LanguageModelResponse:
-    """
-    Calls the completion endpoint synchronously without streaming the response.
+    """Call the completion endpoint synchronously without streaming the response.
 
     Args:
         company_id (str): The company ID associated with the request.
@@ -46,6 +45,7 @@ def complete(
 
     Returns:
         LanguageModelResponse: The response object containing the completed result.
+
     """
     options, model, messages_dict, _ = _prepare_completion_params_util(
         messages=messages,
@@ -62,7 +62,7 @@ def complete(
             company_id=company_id,
             model=model,
             messages=cast(
-                list[unique_sdk.Integrated.ChatCompletionRequestMessage],
+                "list[unique_sdk.Integrated.ChatCompletionRequestMessage]",
                 messages_dict,
             ),
             timeout=timeout,
@@ -82,11 +82,10 @@ async def complete_async(
     timeout: int = DEFAULT_COMPLETE_TIMEOUT,
     tools: list[LanguageModelTool] | None = None,
     other_options: dict | None = None,
-    structured_output_model: Type[BaseModel] | None = None,
+    structured_output_model: type[BaseModel] | None = None,
     structured_output_enforce_schema: bool = False,
 ) -> LanguageModelResponse:
-    """
-    Calls the completion endpoint asynchronously without streaming the response.
+    """Call the completion endpoint asynchronously without streaming the response.
 
     This method sends a request to the completion endpoint using the provided messages, model name,
     temperature, timeout, and optional tools. It returns a `LanguageModelResponse` object containing
@@ -105,7 +104,9 @@ async def complete_async(
         LanguageModelResponse: The response object containing the completed result.
 
     Raises:
-        Exception: If an error occurs during the request, an exception is raised and logged.
+        Exception: If an error occurs during the request, an exception is raised
+        and logged.
+
     """
     options, model, messages_dict, _ = _prepare_completion_params_util(
         messages=messages,
@@ -122,7 +123,7 @@ async def complete_async(
             company_id=company_id,
             model=model,
             messages=cast(
-                list[unique_sdk.Integrated.ChatCompletionRequestMessage],
+                "list[unique_sdk.Integrated.ChatCompletionRequestMessage]",
                 messages_dict,
             ),
             timeout=timeout,
@@ -130,7 +131,7 @@ async def complete_async(
         )
         return LanguageModelResponse(**response)
     except Exception as e:
-        logger.error(f"Error completing: {e}")  # type: ignore
+        logger.exception(f"Error completing: {e}")
         raise e
 
 
@@ -163,14 +164,14 @@ def _to_search_context(chunks: list[ContentChunk]) -> dict | None:
             endPage=chunk.end_page,
             order=chunk.order,
             object=chunk.object,
-        )  # type: ignore
+        )
         for chunk in chunks
     ]
 
 
 def _add_response_format_to_options(
     options: dict,
-    structured_output_model: Type[BaseModel],
+    structured_output_model: type[BaseModel],
     structured_output_enforce_schema: bool = False,
 ) -> dict:
     options["responseFormat"] = {
@@ -191,11 +192,10 @@ def _prepare_completion_params_util(
     tools: list[LanguageModelTool] | None = None,
     other_options: dict | None = None,
     content_chunks: list[ContentChunk] | None = None,
-    structured_output_model: Type[BaseModel] | None = None,
+    structured_output_model: type[BaseModel] | None = None,
     structured_output_enforce_schema: bool = False,
 ) -> tuple[dict, str, dict, dict | None]:
-    """
-    Prepares common parameters for completion requests.
+    """Prepare common parameters for completion requests.
 
     Returns:
         tuple containing:
@@ -203,18 +203,22 @@ def _prepare_completion_params_util(
         - model (str): Resolved model name
         - messages_dict (dict): Processed messages
         - search_context (dict | None): Processed content chunks if provided
-    """
 
+    """
     options = _add_tools_to_options({}, tools)
     if structured_output_model:
         options = _add_response_format_to_options(
-            options, structured_output_model, structured_output_enforce_schema
+            options,
+            structured_output_model,
+            structured_output_enforce_schema,
         )
     options["temperature"] = temperature
     if other_options:
         options.update(other_options)
 
-    model = model_name.name if isinstance(model_name, LanguageModelName) else model_name
+    model = (
+        model_name.value if isinstance(model_name, LanguageModelName) else model_name
+    )
 
     # Different methods need different message dump parameters
     messages_dict = messages.model_dump(

unique_toolkit/language_model/infos.py

@@ -23,6 +23,15 @@ class LanguageModelName(StrEnum):
     AZURE_o3_MINI_2025_0131 = "AZURE_o3_MINI_2025_0131"
     AZURE_GPT_45_PREVIEW_2025_0227 = "AZURE_GPT_45_PREVIEW_2025_0227"
     AZURE_GPT_41_2025_0414 = "AZURE_GPT_41_2025_0414"
+    AZURE_o3_2025_0416 = "AZURE_o3_2025_0416"
+    AZURE_o4_MINI_2025_0416 = "AZURE_o4_MINI_2025_0416"
+    ANTHROPIC_CLAUDE_3_7_SONNET = "litellm:anthropic-claude-3-7-sonnet"
+    ANTHROPIC_CLAUDE_3_7_SONNET_THINKING = (
+        "litellm:anthropic-claude-3-7-sonnet-thinking"
+    )
+    GEMINI_2_0_FLASH = "litellm:gemini-2-0-flash"
+    GEMINI_2_5_FLASH_PREVIEW_0417 = "litellm:gemini-2-5-flash-preview-04-17"
+    GEMINI_2_5_PRO_EXP_0325 = "litellm:gemini-2-5-pro-exp-03-25"
 
 
 class EncoderName(StrEnum):
@@ -57,6 +66,7 @@ def get_encoder_name(model_name: LanguageModelName) -> EncoderName:
 class LanguageModelProvider(StrEnum):
     AZURE = "AZURE"
     CUSTOM = "CUSTOM"
+    LITELLM = "LITELLM"
 
 
 class ModelCapabilities(StrEnum):
@@ -290,6 +300,44 @@ class LanguageModelInfo(BaseModel):
                     info_cutoff_at=date(2023, 10, 1),
                     published_at=date(2025, 1, 31),
                 )
+            case LanguageModelName.AZURE_o3_2025_0416:
+                return cls(
+                    name=model_name,
+                    capabilities=[
+                        ModelCapabilities.STRUCTURED_OUTPUT,
+                        ModelCapabilities.FUNCTION_CALLING,
+                        ModelCapabilities.STREAMING,
+                        ModelCapabilities.REASONING,
+                        ModelCapabilities.VISION,
+                    ],
+                    provider=LanguageModelProvider.AZURE,
+                    version="2025-04-16",
+                    encoder_name=EncoderName.O200K_BASE,
+                    token_limits=LanguageModelTokenLimits(
+                        token_limit_input=200_000, token_limit_output=100_000
+                    ),
+                    info_cutoff_at=date(2024, 5, 31),
+                    published_at=date(2025, 4, 16),
+                )
+            case LanguageModelName.AZURE_o4_MINI_2025_0416:
+                return cls(
+                    name=model_name,
+                    capabilities=[
+                        ModelCapabilities.STRUCTURED_OUTPUT,
+                        ModelCapabilities.FUNCTION_CALLING,
+                        ModelCapabilities.STREAMING,
+                        ModelCapabilities.REASONING,
+                        ModelCapabilities.VISION,
+                    ],
+                    provider=LanguageModelProvider.AZURE,
+                    version="2025-04-16",
+                    encoder_name=EncoderName.O200K_BASE,
+                    token_limits=LanguageModelTokenLimits(
+                        token_limit_input=200_000, token_limit_output=100_000
+                    ),
+                    info_cutoff_at=date(2024, 5, 31),
+                    published_at=date(2025, 4, 16),
+                )
             case LanguageModelName.AZURE_GPT_45_PREVIEW_2025_0227:
                 return cls(
                     name=model_name,
@@ -326,6 +374,98 @@ class LanguageModelInfo(BaseModel):
                     info_cutoff_at=date(2024, 5, 31),
                     published_at=date(2025, 4, 14),
                 )
+            case LanguageModelName.ANTHROPIC_CLAUDE_3_7_SONNET:
+                return cls(
+                    name=model_name,
+                    capabilities=[
+                        ModelCapabilities.FUNCTION_CALLING,
+                        ModelCapabilities.STREAMING,
+                        ModelCapabilities.VISION,
+                    ],
+                    provider=LanguageModelProvider.LITELLM,
+                    version="claude-3-7-sonnet",
+                    encoder_name=EncoderName.O200K_BASE,  # TODO: Update encoder with litellm
+                    token_limits=LanguageModelTokenLimits(
+                        token_limit_input=200_000, token_limit_output=128_000
+                    ),
+                    info_cutoff_at=date(2024, 10, 31),
+                    published_at=date(2025, 2, 24),
+                )
+            case LanguageModelName.ANTHROPIC_CLAUDE_3_7_SONNET_THINKING:
+                return cls(
+                    name=model_name,
+                    capabilities=[
+                        ModelCapabilities.FUNCTION_CALLING,
+                        ModelCapabilities.STREAMING,
+                        ModelCapabilities.VISION,
+                        ModelCapabilities.REASONING,
+                    ],
+                    provider=LanguageModelProvider.LITELLM,
+                    version="claude-3-7-sonnet-thinking",
+                    encoder_name=EncoderName.O200K_BASE,  # TODO: Update encoder with litellm
+                    token_limits=LanguageModelTokenLimits(
+                        token_limit_input=200_000, token_limit_output=128_000
+                    ),
+                    info_cutoff_at=date(2024, 10, 31),
+                    published_at=date(2025, 2, 24),
+                )
+            case LanguageModelName.GEMINI_2_0_FLASH:
+                return cls(
+                    name=model_name,
+                    capabilities=[
+                        ModelCapabilities.FUNCTION_CALLING,
+                        ModelCapabilities.STREAMING,
+                        ModelCapabilities.VISION,
+                        ModelCapabilities.STRUCTURED_OUTPUT,
+                        ModelCapabilities.REASONING,
+                    ],
+                    provider=LanguageModelProvider.LITELLM,
+                    version="gemini-2-0-flash",
+                    encoder_name=EncoderName.O200K_BASE,  # TODO: Update encoder with litellm
+                    token_limits=LanguageModelTokenLimits(
+                        token_limit_input=1_048_576, token_limit_output=8_192
+                    ),
+                    info_cutoff_at=date(2024, 8, 1),
+                    published_at=date(2025, 2, 1),
+                )
+            case LanguageModelName.GEMINI_2_5_FLASH_PREVIEW_0417:
+                return cls(
+                    name=model_name,
+                    capabilities=[
+                        ModelCapabilities.FUNCTION_CALLING,
+                        ModelCapabilities.STREAMING,
+                        ModelCapabilities.VISION,
+                        ModelCapabilities.STRUCTURED_OUTPUT,
+                        ModelCapabilities.REASONING,
+                    ],
+                    provider=LanguageModelProvider.LITELLM,
+                    version="gemini-2-5-flash-preview-04-17",
+                    encoder_name=EncoderName.O200K_BASE,  # TODO:Replace with LLM tokenizer
+                    token_limits=LanguageModelTokenLimits(
+                        token_limit_input=1_048_576, token_limit_output=65_536
+                    ),
+                    info_cutoff_at=date(2025, 1, day=1),
+                    published_at=date(2025, 4, 1),
+                )
+            case LanguageModelName.GEMINI_2_5_PRO_EXP_0325:
+                return cls(
+                    name=model_name,
+                    capabilities=[
+                        ModelCapabilities.FUNCTION_CALLING,
+                        ModelCapabilities.STREAMING,
+                        ModelCapabilities.VISION,
+                        ModelCapabilities.STRUCTURED_OUTPUT,
+                        ModelCapabilities.REASONING,
+                    ],
+                    provider=LanguageModelProvider.LITELLM,
+                    version="gemini-2-5-pro-exp-0325",
+                    encoder_name=EncoderName.O200K_BASE,  # TODO: Update encoder with litellm
+                    token_limits=LanguageModelTokenLimits(
+                        token_limit_input=1_048_576, token_limit_output=65_536
+                    ),
+                    info_cutoff_at=date(2025, 1, day=1),
+                    published_at=date(2025, 3, 1),
+                )
             case _:
                 if isinstance(model_name, LanguageModelName):
                     raise ValueError(
@@ -352,8 +492,10 @@ class LanguageModelInfo(BaseModel):
 
 @deprecated(
     """
-Use `LanguageModelInfo` instead of `LanguageModel`
-"""
+Use `LanguageModelInfo` instead of `LanguageModel`.
+
+`LanguageModel` will be deprecated on 31.12.2025
+""",
 )
 class LanguageModel:
     _info: ClassVar[LanguageModelInfo]
@@ -363,8 +505,8 @@ class LanguageModel:
 
     @property
     def info(self) -> LanguageModelInfo:
-        """
-        Returns all infos about the model:
+        """Return all infos about the model.
+
         - name
         - version
         - provider

{unique_toolkit-0.7.11.dist-info → unique_toolkit-0.7.14.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: unique_toolkit
-Version: 0.7.11
+Version: 0.7.14
 Summary: 
 License: Proprietary
 Author: Martin Fadler
@@ -111,6 +111,17 @@ 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).
 
+## [0.7.14] - 2025-05-08
+- Fix bug not selecting the correct llm
+- Add LMI type for flexible init of LanguageModelInfo
+- Replace LanguageModel with LanguageModelInfo in hallucination check 
+
+## [0.7.13] - 2025-05-07
+- Adding litellm models `litellm:anthropic-claude-3-7-sonnet`, `litellm:anthropic-claude-3-7-sonnet-thinking`, `litellm:gemini-2-0-flash`, `gemini-2-5-flash-preview-04-17` , `litellm:gemini-2-5-pro-exp-03-25`
+
+## [0.7.12] - 2025-05-02
+- add `AZURE_o3_2025_0416` and `AZURE_o4_MINI_2025_0416` as part of the models
+
 ## [0.7.11] - 2025-04-28
 - Removing `STRUCTURED_OUTPUT` capability from `AZURE_GPT_35_TURBO_0125`, `AZURE_GPT_4_TURBO_2024_0409` and `AZURE_GPT_4o_2024_0513`
 

{unique_toolkit-0.7.11.dist-info → unique_toolkit-0.7.14.dist-info}/RECORD

@@ -3,7 +3,7 @@ unique_toolkit/_common/_base_service.py,sha256=S8H0rAebx7GsOldA7xInLp3aQJt9yEPDQ
 unique_toolkit/_common/_time_utils.py,sha256=ztmTovTvr-3w71Ns2VwXC65OKUUh-sQlzbHdKTQWm-w,135
 unique_toolkit/_common/exception.py,sha256=caQIE1btsQnpKCHqL2cgWUSbHup06enQu_Pt7uGUTTE,727
 unique_toolkit/_common/validate_required_values.py,sha256=Y_M1ub9gIKP9qZ45F6Zq3ZHtuIqhmOjl8Z2Vd3avg8w,588
-unique_toolkit/_common/validators.py,sha256=w5lzvRxl0sBTvv0CXLF9UwtJyKmmS2lez0KXaqapgBE,258
+unique_toolkit/_common/validators.py,sha256=l7-hWyRTZ3aF_e73oTQFZdz93s06VhNWVpkERbg2a64,1569
 unique_toolkit/app/__init__.py,sha256=jgwWfu27U911kZE1yRq920ZULGLAQGycD3222YxUvsY,1182
 unique_toolkit/app/init_logging.py,sha256=Sh26SRxOj8i8dzobKhYha2lLrkrMTHfB1V4jR3h23gQ,678
 unique_toolkit/app/init_sdk.py,sha256=Nv4Now4pMfM0AgRhbtatLpm_39rKxn0WmRLwmPhRl-8,1285
@@ -31,24 +31,24 @@ unique_toolkit/embedding/schemas.py,sha256=1GvKCaSk4jixzVQ2PKq8yDqwGEVY_hWclYtoA
 unique_toolkit/embedding/service.py,sha256=ptwNNe2ji7FGqAb5VayedrB9T5b1T00XABwYtgvlGO8,4076
 unique_toolkit/embedding/utils.py,sha256=v86lo__bCJbxZBQ3OcLu5SuwT6NbFfWlcq8iyk6BuzQ,279
 unique_toolkit/evaluators/__init__.py,sha256=3Rfpnowm7MUXHWmeU4UV4s_3Hk-sw3V20oBwQCYlejQ,50
-unique_toolkit/evaluators/config.py,sha256=iYiBi7M6u5MG9nVgpxl9dKfoS4j72stA6Hl-MQHmYp8,1056
+unique_toolkit/evaluators/config.py,sha256=_DIXToJ-hGNpDAdWa7Q6GMjAsxiC_DquLF-SS5s9rTE,717
 unique_toolkit/evaluators/constants.py,sha256=1oI93jsh0R_TjX_8OenliiiywVe3vTooSnaMqtq6R18,27
-unique_toolkit/evaluators/context_relevancy/constants.py,sha256=9mAGc23e5XjTYOBfeuZVbaqOyYrvRoXYjfUnsBOVShU,1126
+unique_toolkit/evaluators/context_relevancy/constants.py,sha256=QG2x32LzV42kAkeWTPuLvOX9NlTSxJlsAgDyxomUBmY,1158
 unique_toolkit/evaluators/context_relevancy/prompts.py,sha256=gTlWP7fDuxhrXhCYNCqXMbCey_DalZMdi5l-a6RHgk0,713
 unique_toolkit/evaluators/context_relevancy/service.py,sha256=9hzdMuF4A4T97-3X3zcXgrDISLn1bleZ6tTL1bHa9dQ,1722
-unique_toolkit/evaluators/context_relevancy/utils.py,sha256=E9ljdRNbwYlx04fQDLvgF4SwxvlTJT0vE328PlUF6KA,5191
+unique_toolkit/evaluators/context_relevancy/utils.py,sha256=qwTkKah6S2hkEGOHxVdQ6RvV6OcjKj4eyd09TcJZlho,5813
 unique_toolkit/evaluators/exception.py,sha256=7lcVbCyoN4Md1chNJDFxpUYyWbVrcr9dcc3TxWykJTc,115
-unique_toolkit/evaluators/hallucination/constants.py,sha256=DEycXlxY9h01D0iF3aU5LIdPrDJ-5OkF0VdXDLn_tSs,1440
+unique_toolkit/evaluators/hallucination/constants.py,sha256=KDhmSlRBnUkfEAFQLaD80rKtj6p-ZJ3L98hqNmNL7xI,1458
 unique_toolkit/evaluators/hallucination/prompts.py,sha256=9yCpO_WGLDvYfPWKL1VuRA-jt0P_-A-qvLUOmuv-Nks,3320
 unique_toolkit/evaluators/hallucination/service.py,sha256=k8qro5Lw4Ak58m4HYp3G4HPLIaexeFySIIVvW6fAdeA,2408
-unique_toolkit/evaluators/hallucination/utils.py,sha256=4KTJH8low_fBzOcuVlcHB2FRrtIiN8TR6uuU8EGwjJM,7668
+unique_toolkit/evaluators/hallucination/utils.py,sha256=gO2AOzDQwVTev2_5vDKgJ9A6A9e0himJyAta_wglVG8,8326
 unique_toolkit/evaluators/output_parser.py,sha256=eI72qkzK1dZyUvnfP2SOAQCGBj_-PwX5wy_aLPMsJMY,883
 unique_toolkit/evaluators/schemas.py,sha256=Jaue6Uhx75X1CyHKWj8sT3RE1JZXTqoLtfLt2xQNCX8,2507
 unique_toolkit/language_model/__init__.py,sha256=jWko_vQj48wjnpTtlkg8iNdef0SMI3FN2kGywXRTMzg,1880
 unique_toolkit/language_model/builder.py,sha256=aIAXWWUoB5G-HONJiAt3MdRGd4jdP8nA-HYX2D2WlSI,3048
 unique_toolkit/language_model/constants.py,sha256=B-topqW0r83dkC_25DeQfnPk3n53qzIHUCBS7YJ0-1U,119
-unique_toolkit/language_model/functions.py,sha256=I5jHhHsKoq7GwEQyTrM8LXB2n_6dvMAk7UklenjuHSY,7945
-unique_toolkit/language_model/infos.py,sha256=CwdyoHhq645DC2-Y2jwkHS3m5umjmpmb1EQJ70zOdjg,18967
+unique_toolkit/language_model/functions.py,sha256=0oSkG4xpbxeaVTJide6g-zunBrsBRuvp7UQlKVbjpSk,7949
+unique_toolkit/language_model/infos.py,sha256=qPf4Xlanet8jf0apZ6-qxS_6zmDd6p9D40it2TqmF3w,25910
 unique_toolkit/language_model/prompt.py,sha256=JSawaLjQg3VR-E2fK8engFyJnNdk21zaO8pPIodzN4Q,3991
 unique_toolkit/language_model/schemas.py,sha256=rrwzUgKANFOrdehCULW8Hh03uRW3tsE5dXpWqxmClfg,8618
 unique_toolkit/language_model/service.py,sha256=FUf-HTKNslrMAh8qFMco_ZpP-N0t_iAFWK3juldoUe8,8343
@@ -58,7 +58,7 @@ unique_toolkit/short_term_memory/constants.py,sha256=698CL6-wjup2MvU19RxSmQk3gX7
 unique_toolkit/short_term_memory/functions.py,sha256=3WiK-xatY5nh4Dr5zlDUye1k3E6kr41RiscwtTplw5k,4484
 unique_toolkit/short_term_memory/schemas.py,sha256=OhfcXyF6ACdwIXW45sKzjtZX_gkcJs8FEZXcgQTNenw,1406
 unique_toolkit/short_term_memory/service.py,sha256=vEKFxP1SScPrFniso492fVthWR1sosdFibhiNF3zRvI,8081
-unique_toolkit-0.7.11.dist-info/LICENSE,sha256=GlN8wHNdh53xwOPg44URnwag6TEolCjoq3YD_KrWgss,193
-unique_toolkit-0.7.11.dist-info/METADATA,sha256=OLYyH8X2yEgAkauCTWjsCJvIiI6JjmJnElRSeqS-Pn8,21641
-unique_toolkit-0.7.11.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
-unique_toolkit-0.7.11.dist-info/RECORD,,
+unique_toolkit-0.7.14.dist-info/LICENSE,sha256=GlN8wHNdh53xwOPg44URnwag6TEolCjoq3YD_KrWgss,193
+unique_toolkit-0.7.14.dist-info/METADATA,sha256=PE6_LSa3whHD6727vPI4Wu2JEuHpPchisWC1RcuFNMA,22172
+unique_toolkit-0.7.14.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
+unique_toolkit-0.7.14.dist-info/RECORD,,