unique_toolkit 1.17.3__py3-none-any.whl → 1.18.0__py3-none-any.whl

unique_toolkit/agentic/tools/a2a/evaluation/evaluator.py

@@ -3,6 +3,7 @@ from typing import override
 
 import unique_sdk
 from jinja2 import Template
+from pydantic import BaseModel
 from typing_extensions import TypedDict
 
 from unique_toolkit.agentic.evaluation.evaluation_manager import Evaluation
@@ -21,6 +22,7 @@ from unique_toolkit.agentic.tools.a2a.evaluation.config import (
     SubAgentEvaluationServiceConfig,
 )
 from unique_toolkit.agentic.tools.a2a.tool import SubAgentTool
+from unique_toolkit.agentic.tools.utils import failsafe
 from unique_toolkit.chat.schemas import (
     ChatMessageAssessmentLabel,
     ChatMessageAssessmentStatus,
@@ -38,7 +40,27 @@ class _SubAgentToolInfo(TypedDict):
     display_name: str
 
 
-NO_ASSESSMENTS_FOUND = "NO_ASSESSMENTS_FOUND"
+_NO_ASSESSMENTS_FOUND = "NO_ASSESSMENTS_FOUND"
+
+
+class _SingleAssessmentData(BaseModel):
+    name: str
+    explanation: str
+
+
+def _format_single_assessment_found(name: str, explanation: str) -> str:
+    return _SingleAssessmentData(name=name, explanation=explanation).model_dump_json()
+
+
+@failsafe(failure_return_value=False)
+def _is_single_assessment_found(value: str) -> bool:
+    _ = _SingleAssessmentData.model_validate_json(value)
+    return True
+
+
+def _parse_single_assessment_found(value: str) -> tuple[str, str]:
+    data = _SingleAssessmentData.model_validate_json(value)
+    return data.name, data.explanation
 
 
 class SubAgentEvaluationService(Evaluation):
@@ -99,15 +121,36 @@ class SubAgentEvaluationService(Evaluation):
 
                 sub_agents_display_data.append(data)
 
+        # No valid assessments found
         if len(sub_agents_display_data) == 0:
             logger.warning("No valid sub agent assessments found")
 
             return EvaluationMetricResult(
                 name=self.get_name(),
-                value=NO_ASSESSMENTS_FOUND,
+                # This is a trick to be able to indicate to `evaluation_metric_to_assessment`
+                # that no valid assessments were found
+                value=_NO_ASSESSMENTS_FOUND,
                 reason="No sub agents assessments found",
             )
 
+        # Only one valid assessment found, no need to perform summarization
+        if (
+            len(sub_agents_display_data) == 1
+            and len(sub_agents_display_data[0]["assessments"]) == 1
+        ):
+            assessment = sub_agents_display_data[0]["assessments"][0]
+            explanation = assessment["explanation"] or ""
+            name = sub_agents_display_data[0]["name"]
+            label = assessment["label"]
+
+            return EvaluationMetricResult(
+                name=self.get_name(),
+                value=label,
+                # This is a trick to be able to pass the display name to the UI in `evaluation_metric_to_assessment`
+                reason=_format_single_assessment_found(name, explanation),
+                is_positive=label == ChatMessageAssessmentLabel.GREEN,
+            )
+
         reason = await self._get_reason(sub_agents_display_data)
 
         return EvaluationMetricResult(
@@ -121,7 +164,7 @@ class SubAgentEvaluationService(Evaluation):
     async def evaluation_metric_to_assessment(
         self, evaluation_result: EvaluationMetricResult
     ) -> EvaluationAssessmentMessage:
-        if evaluation_result.value == NO_ASSESSMENTS_FOUND:
+        if evaluation_result.value == _NO_ASSESSMENTS_FOUND:
             return EvaluationAssessmentMessage(
                 status=ChatMessageAssessmentStatus.DONE,
                 explanation="No valid sub agents assessments found to consolidate.",
@@ -130,6 +173,16 @@ class SubAgentEvaluationService(Evaluation):
                 type=self.get_assessment_type(),
             )
 
+        if _is_single_assessment_found(evaluation_result.reason):
+            name, reason = _parse_single_assessment_found(evaluation_result.reason)
+            return EvaluationAssessmentMessage(
+                status=ChatMessageAssessmentStatus.DONE,
+                explanation=reason,
+                title=name,
+                label=evaluation_result.value,  # type: ignore
+                type=self.get_assessment_type(),
+            )
+
         return EvaluationAssessmentMessage(
             status=ChatMessageAssessmentStatus.DONE,
             explanation=evaluation_result.reason,
@@ -182,12 +235,6 @@ class SubAgentEvaluationService(Evaluation):
         )
 
     async def _get_reason(self, sub_agents_display_data: list[dict]) -> str:
-        if (
-            len(sub_agents_display_data) == 1
-            and len(sub_agents_display_data[0]["assessments"]) == 1
-        ):
-            return sub_agents_display_data[0]["assessments"][0]["explanation"] or ""
-
         messages = (
             MessagesBuilder()
             .system_message_append(self._config.summarization_system_message)

unique_toolkit/agentic/tools/a2a/postprocessing/_display.py

@@ -94,13 +94,17 @@ def _get_display_template(
         assistant_id_placeholder, "{%s}" % answer_placeholder, sep="\n\n"
     )  # Double line break is needed for markdown formatting
 
+    template = _add_line_break(template, before=True, after=False)
+
     if add_quote_border:
         template = _wrap_with_quote_border(template)
 
     match mode:
         case SubAgentResponseDisplayMode.DETAILS_OPEN:
             template = _wrap_with_details_tag(
-                template, "open", display_name_placeholder
+                template,
+                "open",
+                display_name_placeholder,
             )
         case SubAgentResponseDisplayMode.DETAILS_CLOSED:
             template = _wrap_with_details_tag(

unique_toolkit/agentic/tools/a2a/postprocessing/postprocessor.py

@@ -1,9 +1,12 @@
+import asyncio
 import logging
 import re
 from typing import TypedDict, override
 
 import unique_sdk
+from pydantic import BaseModel, Field
 
+from unique_toolkit._common.pydantic_helpers import get_configuration_dict
 from unique_toolkit.agentic.postprocessor.postprocessor_manager import Postprocessor
 from unique_toolkit.agentic.tools.a2a.postprocessing._display import (
     _build_sub_agent_answer_display,
@@ -37,12 +40,21 @@ class _SubAgentToolInfo(TypedDict):
     responses: dict[int, _SubAgentMessageInfo]
 
 
+class SubAgentResponsesPostprocessorConfig(BaseModel):
+    model_config = get_configuration_dict()
+
+    sleep_time_before_update: float = Field(
+        default=0.5, description="Time to sleep before updating the main agent message."
+    )
+
+
 class SubAgentResponsesPostprocessor(Postprocessor):
     def __init__(
         self,
         user_id: str,
         company_id: str,
         main_agent_chat_id: str,
+        config: SubAgentResponsesPostprocessorConfig | None = None,
     ) -> None:
         super().__init__(name=self.__class__.__name__)
 
@@ -52,6 +64,7 @@ class SubAgentResponsesPostprocessor(Postprocessor):
 
         self._assistant_id_to_tool_info: dict[str, _SubAgentToolInfo] = {}
         self._main_agent_message: SpaceMessage | None = None
+        self._config = config or SubAgentResponsesPostprocessorConfig()
 
     @override
     async def run(self, loop_response: LanguageModelStreamResponse) -> None:
@@ -60,6 +73,9 @@ class SubAgentResponsesPostprocessor(Postprocessor):
             company_id=self._company_id,
             chat_id=self._main_agent_chat_id,
         )
+        await asyncio.sleep(
+            self._config.sleep_time_before_update
+        )  # Frontend rendering issues
 
     @override
     def apply_postprocessing_to_response(
@@ -125,7 +141,7 @@ class SubAgentResponsesPostprocessor(Postprocessor):
                 )
 
         loop_response.message.text = (
-            "\n\n".join(answers) + "<br>\n\n" + loop_response.message.text.strip()
+            "<br>\n\n".join(answers) + "<br>\n\n" + loop_response.message.text.strip()
         )
 
         return True

unique_toolkit/agentic/tools/a2a/tool/config.py

@@ -1,3 +1,5 @@
+from typing import Literal
+
 from pydantic import Field
 
 from unique_toolkit._common.pydantic_helpers import get_configuration_dict
@@ -27,6 +29,10 @@ class SubAgentToolConfig(BaseToolConfig):
         default=True,
         description="Whether this sub agent's references should be used in the main agent's response.",
     )
+    forced_tools: list[str] | None = Field(
+        default=None,
+        description="The list of tool names that will be forced to be called for this sub-agent.",
+    )
 
     tool_description_for_system_prompt: str = Field(
         default="",
@@ -61,3 +67,7 @@ class SubAgentToolConfig(BaseToolConfig):
         default=120.0,
         description="Maximum time in seconds to wait for the sub-agent response before timing out.",
     )
+    stop_condition: Literal["stoppedStreamingAt", "completedAt"] = Field(
+        default="completedAt",
+        description="The condition that will be used to stop the polling for the sub-agent response.",
+    )

unique_toolkit/agentic/tools/a2a/tool/service.py

@@ -282,8 +282,9 @@ class SubAgentTool(Tool[SubAgentToolConfig]):
                 text=tool_user_message,
                 chat_id=chat_id,
                 poll_interval=self.config.poll_interval,
+                tool_choices=self.config.forced_tools,
                 max_wait=self.config.max_wait,
-                stop_condition="completedAt",
+                stop_condition=self.config.stop_condition,
             )
         except TimeoutError as e:
             await self._notify_progress(

unique_toolkit/agentic/tools/utils/__init__.py

@@ -1 +1,19 @@
 """Utilities for tools."""
+
+from unique_toolkit.agentic.tools.utils.execution.execution import (
+    Result,
+    SafeTaskExecutor,
+    failsafe,
+    failsafe_async,
+    safe_execute,
+    safe_execute_async,
+)
+
+__all__ = [
+    "failsafe",
+    "failsafe_async",
+    "safe_execute",
+    "safe_execute_async",
+    "SafeTaskExecutor",
+    "Result",
+]

{unique_toolkit-1.17.3.dist-info → unique_toolkit-1.18.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: unique_toolkit
-Version: 1.17.3
+Version: 1.18.0
 Summary: 
 License: Proprietary
 Author: Cedric Klinkert
@@ -118,6 +118,11 @@ 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).
 
+## [1.18.0] - 2025-10-27
+- Temporary fix to rendering of sub agent responses.
+- Add config option `stop_condition` to `SubAgentToolConfig`
+- Add config option `tool_choices` to `SubAgentToolConfig`
+- Make sub agent evaluation use the name of the sub agent name if only one assessment is valid
 
 ## [1.17.3] - 2025-10-27
 - Update Hallucination check citation regex parsing pattern

{unique_toolkit-1.17.3.dist-info → unique_toolkit-1.18.0.dist-info}/RECORD

@@ -61,14 +61,14 @@ unique_toolkit/agentic/tools/a2a/config.py,sha256=6diTTSiS2prY294LfYozB-db2wmJ6j
 unique_toolkit/agentic/tools/a2a/evaluation/__init__.py,sha256=_cR8uBwLbG7lyXoRskTpItzacgs4n23e2LeqClrytuc,354
 unique_toolkit/agentic/tools/a2a/evaluation/_utils.py,sha256=GtcPAMWkwGwJ--hBxn35ow9jN0VKYx8h2qMUXR8DCho,1877
 unique_toolkit/agentic/tools/a2a/evaluation/config.py,sha256=Ra5rzJArS3r8C7RTmzKB8cLEYh4w-u3pe_XLgul3GOA,2355
-unique_toolkit/agentic/tools/a2a/evaluation/evaluator.py,sha256=K4GkVOQwAUofjMF1-ofIGV3XPY1vOnOA8aw6CducRc0,7248
+unique_toolkit/agentic/tools/a2a/evaluation/evaluator.py,sha256=-XDqkC2fLDFMkl8KZIefb1fl9dvlwug-uzcWmhil0vk,9144
 unique_toolkit/agentic/tools/a2a/evaluation/summarization_user_message.j2,sha256=acP1YqD_sCy6DT0V2EIfhQTmaUKeqpeWNJ7RGgceo8I,271
 unique_toolkit/agentic/tools/a2a/manager.py,sha256=FkO9jY7o8Td0t-HBkkatmxwhJGSJXmYkFYKFhPdbpMo,1674
 unique_toolkit/agentic/tools/a2a/postprocessing/__init__.py,sha256=R90CSecxJrKH7TbwiMYPyTwsXUUmouL8fbEUlL4ee9Q,362
-unique_toolkit/agentic/tools/a2a/postprocessing/_display.py,sha256=z6U62NaL_i02mlE2PiaNKmhaB71xjEb-lxM7kgv3Nuk,5085
+unique_toolkit/agentic/tools/a2a/postprocessing/_display.py,sha256=rdWk-6M6V27v7G3dqaaDj1vXcgsFvDrHswFuSKQfd2I,5186
 unique_toolkit/agentic/tools/a2a/postprocessing/_utils.py,sha256=JsWwylR2Ao_L0wk1UlhqeN2fTxPnrbhoi1klYHVBnLk,750
 unique_toolkit/agentic/tools/a2a/postprocessing/config.py,sha256=hqo3vQZX63yXoLT7wN13vf0rC7qKiozKKfdkicYCQno,1061
-unique_toolkit/agentic/tools/a2a/postprocessing/postprocessor.py,sha256=xVclK3X6f-IU9W1jK_OO2yMRpPjVMmeLeQSTTlZQOKM,9703
+unique_toolkit/agentic/tools/a2a/postprocessing/postprocessor.py,sha256=lEjJ1u-CTpoyDT7OFaBBOklONUCUU-CGsosHJRYUfkI,10335
 unique_toolkit/agentic/tools/a2a/postprocessing/test/test_consolidate_references.py,sha256=0hlTbtqClyMqHNnZRRxETgjVhYrtRu3xUnZx0JSE-uo,28030
 unique_toolkit/agentic/tools/a2a/postprocessing/test/test_display.py,sha256=Hn2GOsVhpCkpFV9Asvy_IyiYvCd88rxDzi0E_zt2kWc,37553
 unique_toolkit/agentic/tools/a2a/postprocessing/test/test_postprocessor_reference_functions.py,sha256=GxSkkY-Xgd61Bk8sIRfEtkT9hqL1VgPLWrq-6XoB0rA,11360
@@ -76,8 +76,8 @@ unique_toolkit/agentic/tools/a2a/prompts.py,sha256=0ILHL_RAcT04gFm2d470j4Gho7PoJ
 unique_toolkit/agentic/tools/a2a/tool/__init__.py,sha256=JIJKZBTLTA39OWhxoUd6uairxmqINur1Ex6iXDk9ef8,197
 unique_toolkit/agentic/tools/a2a/tool/_memory.py,sha256=w8bxjokrqHQZgApd55b5rHXF-DpgJwaKTg4CvLBLamc,1034
 unique_toolkit/agentic/tools/a2a/tool/_schema.py,sha256=wMwyunViTnxaURvenkATEvyfXn5LvLaP0HxbYqdZGls,158
-unique_toolkit/agentic/tools/a2a/tool/config.py,sha256=07W8cZFc32yRvIeUFEQXi8R1cSGqlXL4pZb7UlB3Rvs,2490
-unique_toolkit/agentic/tools/a2a/tool/service.py,sha256=6Eru-jqlt13lIBVC26LhH0Oxb3qVdDQIUWRoChf8ysw,10525
+unique_toolkit/agentic/tools/a2a/tool/config.py,sha256=NcrS0hzIeYvjv1oMobAAPlZrbTPPWhcPhi0jUHLFBTI,2903
+unique_toolkit/agentic/tools/a2a/tool/service.py,sha256=Rutos0nmY4PhvO_ETlHr5mRdQFa2UuzSJWC2rmiAJng,10593
 unique_toolkit/agentic/tools/agent_chunks_hanlder.py,sha256=x32Dp1Z8cVW5i-XzXbaMwX2KHPcNGmqEU-FB4AV9ZGo,1909
 unique_toolkit/agentic/tools/config.py,sha256=QD83iy2xAJFyoPggYyLWq-MaSGSq00yS9iI31My1NBc,5387
 unique_toolkit/agentic/tools/factory.py,sha256=A1Aliwx037UAk9ADiDsg0zjCWWnvzV_PxwJNoPTvW6c,1434
@@ -97,7 +97,7 @@ unique_toolkit/agentic/tools/test/test_tool_progress_reporter.py,sha256=dod5QPqg
 unique_toolkit/agentic/tools/tool.py,sha256=m56VLxiHuKU2_J5foZp00xhm5lTxWEW7zRLGbIE9ssU,6744
 unique_toolkit/agentic/tools/tool_manager.py,sha256=DtxJobe_7QKFe6CjnMhCP-mnKO6MjnZeDXsO3jBoC9w,16283
 unique_toolkit/agentic/tools/tool_progress_reporter.py,sha256=ixud9VoHey1vlU1t86cW0-WTvyTwMxNSWBon8I11SUk,7955
-unique_toolkit/agentic/tools/utils/__init__.py,sha256=iD1YYzf9LcJFv95Z8BqCAFSewNBabybZRZyvPKGfvro,27
+unique_toolkit/agentic/tools/utils/__init__.py,sha256=s75sjY5nrJchjLGs3MwSIqhDW08fFXIaX7eRQjFIA4s,346
 unique_toolkit/agentic/tools/utils/execution/__init__.py,sha256=OHiKpqBnfhBiEQagKVWJsZlHv8smPp5OI4dFIexzibw,37
 unique_toolkit/agentic/tools/utils/execution/execution.py,sha256=vjG2Y6awsGNtlvyQAGCTthQ5thWHYnn-vzZXaYLb3QE,7922
 unique_toolkit/agentic/tools/utils/source_handling/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -165,7 +165,7 @@ unique_toolkit/short_term_memory/service.py,sha256=5PeVBu1ZCAfyDb2HLVvlmqSbyzBBu
 unique_toolkit/smart_rules/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 unique_toolkit/smart_rules/compile.py,sha256=Ozhh70qCn2yOzRWr9d8WmJeTo7AQurwd3tStgBMPFLA,1246
 unique_toolkit/test_utilities/events.py,sha256=_mwV2bs5iLjxS1ynDCjaIq-gjjKhXYCK-iy3dRfvO3g,6410
-unique_toolkit-1.17.3.dist-info/LICENSE,sha256=GlN8wHNdh53xwOPg44URnwag6TEolCjoq3YD_KrWgss,193
-unique_toolkit-1.17.3.dist-info/METADATA,sha256=WpWG0ZZEMzpUBMiEOmvdBoGE-q0mfGKX_2RkyQXxwW4,38235
-unique_toolkit-1.17.3.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
-unique_toolkit-1.17.3.dist-info/RECORD,,
+unique_toolkit-1.18.0.dist-info/LICENSE,sha256=GlN8wHNdh53xwOPg44URnwag6TEolCjoq3YD_KrWgss,193
+unique_toolkit-1.18.0.dist-info/METADATA,sha256=6XAUmwya1a8ILqkVKETFxYEaCF4p4Q3GfhRmgWO6tcI,38528
+unique_toolkit-1.18.0.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
+unique_toolkit-1.18.0.dist-info/RECORD,,