unique_toolkit 0.8.23__py3-none-any.whl → 0.8.24__py3-none-any.whl

unique_toolkit/evals/evaluation_manager.py

@@ -78,11 +78,9 @@ class EvaluationManager:
         self,
         logger: Logger,
         chat_service: ChatService,
-        assistant_message_id: str,
     ):
         self._logger = logger
         self._chat_service = chat_service
-        self._assistant_message_id = assistant_message_id
         self._evaluations: dict[EvaluationMetricName, Evaluation] = {}
         self._evaluation_passed: bool = True
 
@@ -96,6 +94,7 @@ class EvaluationManager:
         self,
         selected_evaluation_names: list[EvaluationMetricName],
         loop_response: LanguageModelStreamResponse,
+        assistant_message_id: str,
     ) -> list[EvaluationMetricResult]:
         task_executor = SafeTaskExecutor(
             logger=self._logger,
@@ -106,6 +105,7 @@ class EvaluationManager:
                 self.execute_evaluation_call,
                 loop_response=loop_response,
                 evaluation_name=evaluation_name,
+                assistant_message_id=assistant_message_id,
             )
             for evaluation_name in selected_evaluation_names
         ]
@@ -126,6 +126,7 @@ class EvaluationManager:
         self,
         evaluation_name: EvaluationMetricName,
         loop_response: LanguageModelStreamResponse,
+        assistant_message_id: str,
     ) -> EvaluationMetricResult:
         self._logger.info(f"Processing tool call: {evaluation_name}")
 
@@ -133,13 +134,15 @@ class EvaluationManager:
 
         if evaluation_instance:
             # Execute the evaluation
-            await self._create_assistant_message(evaluation_instance)
+            await self._create_assistant_message(
+                evaluation_instance, assistant_message_id
+            )
             evaluation_metric_result: EvaluationMetricResult = (
                 await evaluation_instance.run(loop_response)
             )
             # show results to the user
             await self._show_message_assessment(
-                evaluation_instance, evaluation_metric_result
+                evaluation_instance, evaluation_metric_result, assistant_message_id
             )
 
             return evaluation_metric_result
@@ -182,6 +185,7 @@ class EvaluationManager:
         self,
         evaluation_instance: Evaluation,
         evaluation_metric_result: EvaluationMetricResult,
+        assistant_message_id: str,
     ) -> None:
         evaluation_assessment_message = (
             await evaluation_instance.evaluation_metric_to_assessment(
@@ -189,7 +193,7 @@ class EvaluationManager:
             )
         )
         await self._chat_service.modify_message_assessment_async(
-            assistant_message_id=self._assistant_message_id,
+            assistant_message_id=assistant_message_id,
             status=evaluation_assessment_message.status,
             title=evaluation_assessment_message.title,
             explanation=evaluation_assessment_message.explanation,
@@ -197,9 +201,11 @@ class EvaluationManager:
             type=evaluation_assessment_message.type,
         )
 
-    async def _create_assistant_message(self, evaluation_instance: Evaluation):
+    async def _create_assistant_message(
+        self, evaluation_instance: Evaluation, assistant_message_id: str
+    ):
         await self._chat_service.create_message_assessment_async(
-            assistant_message_id=self._assistant_message_id,
+            assistant_message_id=assistant_message_id,
             status=ChatMessageAssessmentStatus.PENDING,
             type=evaluation_instance.get_assessment_type(),
         )

{unique_toolkit-0.8.23.dist-info → unique_toolkit-0.8.24.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: unique_toolkit
-Version: 0.8.23
+Version: 0.8.24
 Summary: 
 License: Proprietary
 Author: Martin Fadler
@@ -114,9 +114,13 @@ 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).
 
+## [0.8.24] - 2025-08-25
+- Optimized hallucination manager
+
 ## [0.8.23] - 2025-08-27
 - Add MCP manager that handles MCP related logic
 
+
 ## [0.8.22] - 2025-08-25
 - Add DeepSeek-R1, DeepSeek-V3.1, Qwen3-235B-A22B and Qwen3-235B-A22B-Thinking-2507 to supported model list
 

{unique_toolkit-0.8.23.dist-info → unique_toolkit-0.8.24.dist-info}/RECORD

@@ -40,7 +40,7 @@ unique_toolkit/evals/config.py,sha256=ywHIrJs5SFdKr1WXfrofWuFfzb0iPQw8iZDpq5oEug
 unique_toolkit/evals/context_relevancy/prompts.py,sha256=EdHFUOB581yVxcOL8482KUv_LzaRjuiem71EF8udYMc,1331
 unique_toolkit/evals/context_relevancy/schema.py,sha256=lm9x0jExOinUk9itqC8ZpgReC7yj1VDwEMppxlZGqpY,2923
 unique_toolkit/evals/context_relevancy/service.py,sha256=txTWIhV65QGFhxG1jCb5TTqZc_c7K9I8pi2HVDTAfm8,8384
-unique_toolkit/evals/evaluation_manager.py,sha256=luavQrMTrxEi7hRhpUSlGLYpEza7eQCg5Nk46ogbabw,7721
+unique_toolkit/evals/evaluation_manager.py,sha256=g-8qa_6_p53C9Okx8iNkuoIXYSJrf-6sQ-xku7bo9kI,7895
 unique_toolkit/evals/exception.py,sha256=7lcVbCyoN4Md1chNJDFxpUYyWbVrcr9dcc3TxWykJTc,115
 unique_toolkit/evals/hallucination/constants.py,sha256=FLcXl5XU07jCvS8YPX9l6UjTaqyQ8YvnSKpx4Z6wZ2Y,1997
 unique_toolkit/evals/hallucination/hallucination_evaluation.py,sha256=TuZ88jeVn0tVr9d0GhWyJSxKNA16nhvr2xRPo-yK8OM,3063
@@ -112,7 +112,7 @@ unique_toolkit/tools/utils/execution/execution.py,sha256=vjG2Y6awsGNtlvyQAGCTthQ
 unique_toolkit/tools/utils/source_handling/schema.py,sha256=vzAyf6ZWNexjMO0OrnB8y2glGkvAilmGGQXd6zcDaKw,870
 unique_toolkit/tools/utils/source_handling/source_formatting.py,sha256=C7uayNbdkNVJdEARA5CENnHtNY1SU6etlaqbgHNyxaQ,9152
 unique_toolkit/tools/utils/source_handling/tests/test_source_formatting.py,sha256=oM5ZxEgzROrnX1229KViCAFjRxl9wCTzWZoinYSHleM,6979
-unique_toolkit-0.8.23.dist-info/LICENSE,sha256=GlN8wHNdh53xwOPg44URnwag6TEolCjoq3YD_KrWgss,193
-unique_toolkit-0.8.23.dist-info/METADATA,sha256=e9_t3WfrtzAJirY80IWJtPgvk1shO1OIgR5wRlDw8O4,28527
-unique_toolkit-0.8.23.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
-unique_toolkit-0.8.23.dist-info/RECORD,,
+unique_toolkit-0.8.24.dist-info/LICENSE,sha256=GlN8wHNdh53xwOPg44URnwag6TEolCjoq3YD_KrWgss,193
+unique_toolkit-0.8.24.dist-info/METADATA,sha256=0vvS5x-HsApDZlOrB7vnLQPal3u7Ucqxv7hZa35ebOo,28588
+unique_toolkit-0.8.24.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
+unique_toolkit-0.8.24.dist-info/RECORD,,