unique_toolkit 1.56.0__tar.gz → 1.57__tar.gz

{unique_toolkit-1.56.0 → unique_toolkit-1.57}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: unique-toolkit
-Version: 1.56.0
+Version: 1.57
 Summary: 
 Author: Cedric Klinkert, Martin Fadler, Sadique Sheik, Fabian Schläpfer, Pascal Hauri, Ejiroghene Oki-Peter, Rami Azouz, Ricardo Visini, Sanzio Monti
 Author-email: Cedric Klinkert <cedric.klinkert@unique.ch>, Martin Fadler <martin.fadler@unique.ch>, Sadique Sheik <sadique@unique.ch>, Fabian Schläpfer <fabian@unique.ch>, Pascal Hauri <pascal@unique.ch>, Ejiroghene Oki-Peter <oki-peter.ejiroghene@unique.ai>, Rami Azouz <rami.ext@unique.ch>, Ricardo Visini <ricardo.visini@unique.ch>, Sanzio Monti <sanzio.monti@unique.ai>

{unique_toolkit-1.56.0 → unique_toolkit-1.57}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "unique_toolkit"
-version = "1.56.0"
+version = "1.57"
 description = ""
 readme = "README.md"
 requires-python = ">=3.12"

{unique_toolkit-1.56.0 → unique_toolkit-1.57}/unique_toolkit/agentic/evaluation/evaluation_manager.py

@@ -1,5 +1,4 @@
 import asyncio
-import time
 from abc import ABC
 from logging import Logger
 
@@ -87,7 +86,6 @@ class EvaluationManager:
         self._chat_service = chat_service
         self._evaluations: dict[EvaluationMetricName, Evaluation] = {}
         self._evaluation_passed: bool = True
-        self._execution_times: dict[str, float] = {}
 
     def add_evaluation(self, evaluation: Evaluation):
         self._evaluations[evaluation.get_name()] = evaluation
@@ -101,8 +99,6 @@ class EvaluationManager:
         loop_response: LanguageModelStreamResponse,
         assistant_message_id: str,
     ) -> list[EvaluationMetricResult]:
-        self._execution_times = {}
-
         task_executor = SafeTaskExecutor(
             logger=self._logger,
         )
@@ -138,9 +134,6 @@ class EvaluationManager:
 
         return evaluation_results_unpacked
 
-    def get_execution_times(self) -> dict[str, float]:
-        return self._execution_times.copy()
-
     async def execute_evaluation_call(
         self,
         evaluation_name: EvaluationMetricName,
@@ -152,16 +145,13 @@ class EvaluationManager:
         evaluation_instance = self.get_evaluation_by_name(evaluation_name)
 
         if evaluation_instance:
-            start = time.perf_counter()
+            # Execute the evaluation
             await self._create_assistant_message(
                 evaluation_instance, assistant_message_id
             )
             evaluation_metric_result: EvaluationMetricResult = (
                 await evaluation_instance.run(loop_response)
             )
-            self._execution_times[str(evaluation_name)] = round(
-                time.perf_counter() - start, 3
-            )
             return evaluation_metric_result
 
         return EvaluationMetricResult(

{unique_toolkit-1.56.0 → unique_toolkit-1.57}/unique_toolkit/agentic/evaluation/tests/test_hallucination_utils.py

@@ -17,13 +17,13 @@ from unique_toolkit.agentic.evaluation.hallucination.utils import (
     context_text_from_stream_response,
 )
 from unique_toolkit.agentic.evaluation.schemas import EvaluationMetricInput
-from unique_toolkit.chat.schemas import ChatMessage
 from unique_toolkit.content import ContentReference
 from unique_toolkit.content.schemas import ContentChunk
 from unique_toolkit.language_model.schemas import (
     LanguageModelMessageRole,
     LanguageModelMessages,
     LanguageModelStreamResponse,
+    LanguageModelStreamResponseMessage,
 )
 
 
@@ -580,9 +580,8 @@ def test_context_text_from_stream_response__extracts_context__using_from_ids_mod
     """
     # Arrange
     response: LanguageModelStreamResponse = LanguageModelStreamResponse(
-        message=ChatMessage(
+        message=LanguageModelStreamResponseMessage(
             id="msg_1",
-            chat_id="",
             previous_message_id=None,
             role=LanguageModelMessageRole.ASSISTANT,
             text="Test",
@@ -613,9 +612,8 @@ def test_context_text_from_stream_response__extracts_context__using_from_order_m
     """
     # Arrange
     response: LanguageModelStreamResponse = LanguageModelStreamResponse(
-        message=ChatMessage(
+        message=LanguageModelStreamResponseMessage(
             id="msg_1",
-            chat_id="",
             previous_message_id=None,
             role=LanguageModelMessageRole.ASSISTANT,
             text="Test",
@@ -655,9 +653,8 @@ def test_context_text_from_stream_response__extracts_context__using_from_origina
         )
     ]
     response: LanguageModelStreamResponse = LanguageModelStreamResponse(
-        message=ChatMessage(
+        message=LanguageModelStreamResponseMessage(
             id="msg_1",
-            chat_id="",
             previous_message_id=None,
             role=LanguageModelMessageRole.ASSISTANT,
             text="Test",
@@ -693,9 +690,8 @@ def test_context_text_from_stream_response__falls_back_to_default__with_invalid_
     """
     # Arrange
     response: LanguageModelStreamResponse = LanguageModelStreamResponse(
-        message=ChatMessage(
+        message=LanguageModelStreamResponseMessage(
             id="msg_1",
-            chat_id="",
             previous_message_id=None,
             role=LanguageModelMessageRole.ASSISTANT,
             text="Test",
@@ -735,9 +731,8 @@ def test_context_text_from_stream_response__falls_back_to_default__on_extraction
         )
     ]
     response: LanguageModelStreamResponse = LanguageModelStreamResponse(
-        message=ChatMessage(
+        message=LanguageModelStreamResponseMessage(
             id="msg_1",
-            chat_id="",
             previous_message_id=None,
             role=LanguageModelMessageRole.ASSISTANT,
             text="Test",
@@ -768,9 +763,8 @@ def test_context_text_from_stream_response__returns_text_strings__not_chunk_obje
     """
     # Arrange
     response: LanguageModelStreamResponse = LanguageModelStreamResponse(
-        message=ChatMessage(
+        message=LanguageModelStreamResponseMessage(
             id="msg_1",
-            chat_id="",
             previous_message_id=None,
             role=LanguageModelMessageRole.ASSISTANT,
             text="Test",

{unique_toolkit-1.56.0 → unique_toolkit-1.57}/unique_toolkit/agentic/history_manager/history_construction_with_contents.py

@@ -173,12 +173,12 @@ def get_full_history_with_contents(
 
     builder = LanguageModelMessages([]).builder()
     for c in grouped_elements:
-        # LanguageModelUserMessage has not field original_text
-        text = c.original_text if c.original_text else c.content
+        # LanguageModelUserMessage has not field original content
+        text = c.original_content if c.original_content else c.content
         if text is None:
             if c.role == ChatRole.USER:
                 raise ValueError(
-                    "Content or original_text of LanguageModelMessages should exist.",
+                    "Content or original_content of LanguageModelMessages should exist.",
                 )
             text = ""
 

{unique_toolkit-1.56.0 → unique_toolkit-1.57}/unique_toolkit/agentic/postprocessor/postprocessor_manager.py

@@ -1,5 +1,4 @@
 import asyncio
-import time
 from abc import ABC
 from logging import Logger
 
@@ -85,7 +84,6 @@ class PostprocessorManager:
         self._logger = logger
         self._chat_service = chat_service
         self._postprocessors: list[Postprocessor | ResponsesApiPostprocessor] = []
-        self._execution_times: dict[str, float] = {}
 
         # Allow to add postprocessors that should be run before or after the others.
         self._first_postprocessor: Postprocessor | ResponsesApiPostprocessor | None = (
@@ -125,8 +123,6 @@ class PostprocessorManager:
         self,
         loop_response: LanguageModelStreamResponse,
     ) -> None:
-        self._execution_times = {}
-
         task_executor = SafeTaskExecutor(
             logger=self._logger,
         )
@@ -162,19 +158,12 @@ class PostprocessorManager:
                 references=loop_response.message.references,
             )
 
-    def get_execution_times(self) -> dict[str, float]:
-        return self._execution_times.copy()
-
     async def execute_postprocessors(
         self,
         loop_response: LanguageModelStreamResponse,
         postprocessor_instance: Postprocessor | ResponsesApiPostprocessor,
     ) -> None:
-        start = time.perf_counter()
         await postprocessor_instance.run(loop_response)  # type: ignore
-        self._execution_times[postprocessor_instance.name] = round(
-            time.perf_counter() - start, 3
-        )
 
     async def remove_from_text(
         self,

{unique_toolkit-1.56.0 → unique_toolkit-1.57}/unique_toolkit/agentic/tools/tool_manager.py

@@ -1,5 +1,4 @@
 import asyncio
-import time
 from logging import Logger, getLogger
 from typing import Generic, Literal, TypeVar, overload
 
@@ -289,16 +288,10 @@ class _ToolManager(Generic[_ApiMode]):
             raise ValueError(f"Tool {tool_call.name} cannot be run")
 
         if tool_instance:
-            tool_start = time.perf_counter()
+            # Execute the tool
             tool_response: ToolCallResponse = await tool_instance.run(
                 tool_call=tool_call
             )
-            tool_execution_time = round(time.perf_counter() - tool_start, 3)
-
-            if tool_response.debug_info is None:
-                tool_response.debug_info = {}
-            tool_response.debug_info["execution_time_s"] = tool_execution_time
-
             evaluation_checks = tool_instance.evaluation_check_list()
             self._tool_evaluation_check_list.update(evaluation_checks)
 

{unique_toolkit-1.56.0 → unique_toolkit-1.57}/unique_toolkit/chat/schemas.py

@@ -66,67 +66,26 @@ class ToolCall(BaseModel):
         )
 
 
-class ChatMessageAssessmentStatus(StrEnum):
-    PENDING = "PENDING"
-    DONE = "DONE"
-    ERROR = "ERROR"
-
-
-class ChatMessageAssessmentLabel(StrEnum):
-    RED = "RED"
-    YELLOW = "YELLOW"
-    GREEN = "GREEN"
-
-
-class ChatMessageAssessmentType(StrEnum):
-    HALLUCINATION = "HALLUCINATION"
-    COMPLIANCE = "COMPLIANCE"
-
-
-class ChatMessageAssessment(BaseModel):
-    model_config = model_config
-
-    id: str
-    object: str
-    message_id: str
-    status: ChatMessageAssessmentStatus
-    type: ChatMessageAssessmentType
-    title: str | None = None
-    explanation: str | None = None
-    label: ChatMessageAssessmentLabel | None = None
-    is_visible: bool
-    created_at: datetime | None = None
-    updated_at: datetime | None = None
-
-
 class ChatMessage(BaseModel):
-    # This model maps to PublicMessageDto from the backend public API:
-    # next/services/node-chat/src/public-api/2023-12-06/dtos/message/public-message.dto.ts
-    # Fields previous_message_id, tool_calls, tool_call_id are internal extensions not in the DTO.
+    # TODO: The below seems not to be True anymore @irina-unique. To be checked in separate PR
+    # This model should strictly meets https://github.com/Unique-AG/monorepo/blob/master/node/apps/node-chat/src/public-api/2023-12-06/dtos/message/public-message.dto.ts
     model_config = model_config
 
     id: str | None = None
     chat_id: str
     object: str | None = None
-    # alias="text" applies only to construction (model_validate / __init__).
-    # model_dump() uses the field name "content", not "text". Use model_dump(by_alias=True)
-    # to get "text" as the key, or access via the .text property.
     content: str | None = Field(default=None, alias="text")
-    original_text: str | None = None
+    original_content: str | None = Field(default=None, alias="originalText")
     role: ChatMessageRole
-    previous_message_id: str | None = None
     gpt_request: list[dict] | None = None
     tool_calls: list[ToolCall] | None = None
     tool_call_id: str | None = None
     debug_info: dict | None = {}
     created_at: datetime | None = None
     completed_at: datetime | None = None
-    started_streaming_at: datetime | None = None
-    stopped_streaming_at: datetime | None = None
     user_aborted_at: datetime | None = None
     updated_at: datetime | None = None
-    references: list[ContentReference] | None = []
-    assessment: list[ChatMessageAssessment] | None = None
+    references: list[ContentReference] | None = None
 
     # TODO make sdk return role consistently in lowercase
     # Currently needed as sdk returns role in uppercase
@@ -134,14 +93,6 @@ class ChatMessage(BaseModel):
     def set_role(cls, value: str):
         return value.lower()
 
-    @property
-    def text(self) -> str | None:
-        return self.content
-
-    @text.setter
-    def text(self, value: str | None) -> None:
-        self.content = value
-
     # Ensure tool_call_ids is required if role is 'tool'
     @model_validator(mode="after")
     def check_tool_call_ids_for_tool_role(self):
@@ -182,6 +133,39 @@ class ChatMessage(BaseModel):
                 raise NotImplementedError
 
 
+class ChatMessageAssessmentStatus(StrEnum):
+    PENDING = "PENDING"
+    DONE = "DONE"
+    ERROR = "ERROR"
+
+
+class ChatMessageAssessmentLabel(StrEnum):
+    RED = "RED"
+    YELLOW = "YELLOW"
+    GREEN = "GREEN"
+
+
+class ChatMessageAssessmentType(StrEnum):
+    HALLUCINATION = "HALLUCINATION"
+    COMPLIANCE = "COMPLIANCE"
+
+
+class ChatMessageAssessment(BaseModel):
+    model_config = model_config
+
+    id: str
+    object: str
+    message_id: str
+    status: ChatMessageAssessmentStatus
+    type: ChatMessageAssessmentType
+    title: str | None = None
+    explanation: str | None = None
+    label: ChatMessageAssessmentLabel | None = None
+    is_visible: bool
+    created_at: datetime | None = None
+    updated_at: datetime | None = None
+
+
 class MessageLogStatus(StrEnum):
     RUNNING = "RUNNING"
     COMPLETED = "COMPLETED"

{unique_toolkit-1.56.0 → unique_toolkit-1.57}/unique_toolkit/content/schemas.py

@@ -117,7 +117,7 @@ class ContentReference(BaseModel):
     url: str
     original_index: list[int] = Field(
         default=[],
-        description="List of indices in the ChatMessage original_text this reference refers to. This is usually the id in the functionCallResponse. List type due to implementation in node-chat",
+        description="List of indices in the ChatMessage original_content this reference refers to. This is usually the id in the functionCallResponse. List type due to implementation in node-chat",
     )
 
     @classmethod

{unique_toolkit-1.56.0 → unique_toolkit-1.57}/unique_toolkit/language_model/__init__.py

@@ -20,7 +20,6 @@ from .schemas import (
     LanguageModelStreamResponseMessage,
     LanguageModelSystemMessage,
     LanguageModelTokenLimits,
-    LanguageModelTokenUsage,
     LanguageModelTool,
     LanguageModelToolDescription,
     LanguageModelToolMessage,
@@ -58,7 +57,6 @@ __all__ = [
     "LanguageModelStreamResponseMessage",
     "LanguageModelSystemMessage",
     "LanguageModelTokenLimits",
-    "LanguageModelTokenUsage",
     "LanguageModelTool",
     "LanguageModelToolDescription",
     "LanguageModelToolMessage",

{unique_toolkit-1.56.0 → unique_toolkit-1.57}/unique_toolkit/language_model/functions.py

@@ -30,6 +30,7 @@ from unique_toolkit.language_model.schemas import (
     LanguageModelMessages,
     LanguageModelResponse,
     LanguageModelStreamResponse,
+    LanguageModelStreamResponseMessage,
     LanguageModelTool,
     LanguageModelToolDescription,
 )
@@ -527,13 +528,12 @@ def _create_language_model_stream_response_with_references(
         search_context=content_chunks,
     )
 
-    stream_response_message = ChatMessage(
+    stream_response_message = LanguageModelStreamResponseMessage(
         id="stream_unknown",
         previous_message_id=None,
         role=LanguageModelMessageRole.ASSISTANT,
         text=message.content or "",
         original_text=content,
-        chat_id="stream_unknown",
         references=[
             ContentReference(**u.model_dump()) for u in message.references or []
         ],
@@ -657,9 +657,8 @@ async def stream_complete_with_references_openai(
             ]
 
         return LanguageModelStreamResponse(
-            message=ChatMessage(
+            message=LanguageModelStreamResponseMessage(
                 id=message_id or uuid4().hex,
-                chat_id="",
                 previous_message_id=None,
                 role=LanguageModelMessageRole.ASSISTANT,
                 text=text_,

{unique_toolkit-1.56.0 → unique_toolkit-1.57}/unique_toolkit/language_model/schemas.py

@@ -1,5 +1,6 @@
 import json
 import math
+from enum import StrEnum
 from typing import Any, Literal, Self, TypeVar
 from uuid import uuid4
 
@@ -39,8 +40,7 @@ from pydantic import (
 )
 from typing_extensions import deprecated, overload
 
-from unique_toolkit.chat.schemas import ChatMessage
-from unique_toolkit.chat.schemas import ChatMessageRole as LanguageModelMessageRole
+from unique_toolkit.content.schemas import ContentReference
 from unique_toolkit.language_model._responses_api_utils import (
     convert_user_message_content_to_responses_api,
 )
@@ -54,14 +54,33 @@ model_config = ConfigDict(
 )
 
 
-# Backward compatibility alias — use ChatMessage directly.
-# LanguageModelStreamResponseMessage was a toolkit-only subclass with stricter types
-# (required id/content, non-null references). The backend DTOs (PublicMessageDto,
-# PublicStreamResultDto) use a single message type, so this distinction was unnecessary.
-# Note: id and content are now optional (str | None) instead of required (str). This is a
-# deliberate contract relaxation — the backend always returns non-null values in stream
-# responses, but callers that relied on the stricter types should add explicit null checks.
-LanguageModelStreamResponseMessage = ChatMessage
+# Equivalent to
+# from openai.types.chat.chat_completion_role import ChatCompletionRole
+class LanguageModelMessageRole(StrEnum):
+    USER = "user"
+    SYSTEM = "system"
+    ASSISTANT = "assistant"
+    TOOL = "tool"
+
+
+# This is tailored to the unique backend
+class LanguageModelStreamResponseMessage(BaseModel):
+    model_config = model_config
+
+    id: str
+    previous_message_id: (
+        str | None
+    )  # Stream response can return a null previous_message_id if an assisstant message is manually added
+    role: LanguageModelMessageRole
+    text: str
+    original_text: str | None = None
+    references: list[ContentReference] = []
+
+    # TODO make sdk return role in lowercase
+    # Currently needed as sdk returns role in uppercase
+    @field_validator("role", mode="before")
+    def set_role(cls, value: str):
+        return value.lower()
 
 
 class LanguageModelFunction(BaseModel):
@@ -144,21 +163,12 @@ class LanguageModelFunction(BaseModel):
             )
 
 
-class LanguageModelTokenUsage(BaseModel):
-    model_config = model_config
-
-    completion_tokens: int | None = None
-    prompt_tokens: int | None = None
-    total_tokens: int | None = None
-
-
 class LanguageModelStreamResponse(BaseModel):
     model_config = model_config
 
-    message: ChatMessage
+    message: LanguageModelStreamResponseMessage
     tool_calls: list[LanguageModelFunction] | None = None
     stopped_by_user: bool = False
-    usage: LanguageModelTokenUsage | None = None
 
     def is_empty(self) -> bool:
         """

{unique_toolkit-1.56.0 → unique_toolkit-1.57}/unique_toolkit/services/chat_service.py

@@ -140,50 +140,6 @@ class ChatService(ChatServiceDeprecated):
 
         return self._elicitation_service
 
-    def get_debug_info(self) -> dict[str, Any]:
-        """Retrieves the debug information from the current user message.
-
-        Returns:
-            dict[str, Any]: The debug information stored on the user message,
-                or an empty dict if retrieval fails or no debug info exists.
-
-        """
-        try:
-            raw_msg = unique_sdk.Message.retrieve(
-                user_id=self._user_id,
-                company_id=self._company_id,
-                id=self._user_message_id,
-                chatId=self._chat_id,
-            )
-            return getattr(raw_msg, "debugInfo", {}) or {}
-        except Exception:
-            logging.getLogger(__name__).warning(
-                "Failed to retrieve debug info from user message", exc_info=True
-            )
-            return {}
-
-    async def get_debug_info_async(self) -> dict[str, Any]:
-        """Retrieves the debug information from the current user message asynchronously.
-
-        Returns:
-            dict[str, Any]: The debug information stored on the user message,
-                or an empty dict if retrieval fails or no debug info exists.
-
-        """
-        try:
-            raw_msg = await unique_sdk.Message.retrieve_async(
-                user_id=self._user_id,
-                company_id=self._company_id,
-                id=self._user_message_id,
-                chatId=self._chat_id,
-            )
-            return getattr(raw_msg, "debugInfo", {}) or {}
-        except Exception:
-            logging.getLogger(__name__).warning(
-                "Failed to retrieve debug info from user message", exc_info=True
-            )
-            return {}
-
     async def update_debug_info_async(self, debug_info: dict):
         """Updates the debug information for the chat session.