unique_toolkit 2026.32.4__tar.gz → 2026.34.0__tar.gz

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: unique-toolkit
-Version: 2026.32.4
+Version: 2026.34.0
 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>
@@ -16,7 +16,7 @@ Requires-Dist: sseclient>=0.0.27,<0.0.28
 Requires-Dist: openai>=1.105.0,<3
 Requires-Dist: platformdirs>=4.0.0,<5
 Requires-Dist: pillow>=12.1.1
-Requires-Dist: unique-sdk>=2026.32.4
+Requires-Dist: unique-sdk>=2026.34.0
 Requires-Dist: jambo>=0.1.2,<0.2
 Requires-Dist: jinja2>=3.1.6,<4
 Requires-Dist: docxtpl>=0.20.1,<0.21
@@ -42,10 +42,15 @@ Requires-Dist: langchain>=1.0.0,<2 ; extra == 'langchain'
 Requires-Dist: langchain-openai>=0.3.28 ; extra == 'langchain'
 Requires-Dist: langchain-core>=1.0.0,<2 ; extra == 'langchain'
 Requires-Dist: prometheus-client>=0.24.1,<1 ; extra == 'monitoring'
+Requires-Dist: opentelemetry-sdk>=1.27,<2 ; extra == 'otel'
+Requires-Dist: opentelemetry-exporter-otlp-proto-http>=1.27,<2 ; extra == 'otel'
+Requires-Dist: opentelemetry-instrumentation-fastapi>=0.65b0,<1 ; extra == 'otel'
+Requires-Dist: opentelemetry-instrumentation-requests>=0.65b0,<1 ; extra == 'otel'
 Requires-Python: >=3.12
 Provides-Extra: fastapi
 Provides-Extra: langchain
 Provides-Extra: monitoring
+Provides-Extra: otel
 Description-Content-Type: text/markdown
 
 # Unique Toolkit

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "unique_toolkit"
-version = "2026.32.4"
+version = "2026.34.0"
 description = ""
 readme = "README.md"
 requires-python = ">=3.12"
@@ -28,7 +28,7 @@ dependencies = [
     "openai>=1.105.0,<3",
     "platformdirs>=4.0.0,<5",
     "pillow>=12.1.1",
-    "unique-sdk>=2026.32.4",
+    "unique-sdk>=2026.34.0",
     "jambo>=0.1.2,<0.2",
     "jinja2>=3.1.6,<4",
     "docxtpl>=0.20.1,<0.21",
@@ -63,6 +63,12 @@ fastapi = [
 monitoring = [
     "prometheus_client>=0.24.1,<1",
 ]
+otel = [
+    "opentelemetry-sdk>=1.27,<2",
+    "opentelemetry-exporter-otlp-proto-http>=1.27,<2",
+    "opentelemetry-instrumentation-fastapi>=0.65b0,<1",
+    "opentelemetry-instrumentation-requests>=0.65b0,<1",
+]
 [build-system]
 requires = ["uv_build>=0.8.14,<0.9.0"]
 build-backend = "uv_build"
@@ -73,6 +79,10 @@ module-root = ""
 [dependency-groups]
 dev = [
     "types-cachetools>=5.0,<6",
+    "opentelemetry-sdk>=1.27,<2",
+    "opentelemetry-exporter-otlp-proto-http>=1.27,<2",
+    "opentelemetry-instrumentation-fastapi>=0.65b0,<1",
+    "opentelemetry-instrumentation-requests>=0.65b0,<1",
 ]
 
 [tool.uv]
@@ -119,7 +129,16 @@ known_first_party = ["unique_toolkit"]
 extend_exclude = ["docs", "tests", "unique_toolkit/.*/test", "unique_toolkit/.*/tests"]
 
 [tool.deptry.per_rule_ignores]
-DEP002 = ["langchain", "fastapi", "starlette", "uvicorn"]
+DEP002 = [
+    "langchain",
+    "fastapi",
+    "starlette",
+    "uvicorn",
+    "opentelemetry-sdk",
+    "opentelemetry-exporter-otlp-proto-http",
+    "opentelemetry-instrumentation-fastapi",
+    "opentelemetry-instrumentation-requests",
+]
 DEP004 = ["langchain_openai", "langchain_core"]
 
 # -----------------------------------------------------------------------------

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/unique_toolkit/agentic/debug_info_manager/debug_info_manager.py

@@ -24,6 +24,8 @@ class AnalyticsTool(TypedDict, total=False):
     is_sub_agent: bool
     is_mcp: bool
     loop_iteration: int
+    elicitation_approval: bool
+    elicitation_prompt_change: bool
 
 
 class AnalyticsSkill(TypedDict):
@@ -294,6 +296,10 @@ def _to_analytics_tool_entry(
         analytics_tool["is_mcp"] = True
     if "loop_iteration" in info:
         analytics_tool["loop_iteration"] = info["loop_iteration"]
+    if "elicitation_approval" in info:
+        analytics_tool["elicitation_approval"] = info["elicitation_approval"]
+    if "elicitation_prompt_change" in info:
+        analytics_tool["elicitation_prompt_change"] = info["elicitation_prompt_change"]
     return analytics_tool
 
 

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/unique_toolkit/agentic/evaluation/hallucination/hallucination_evaluation.py

@@ -36,6 +36,8 @@ class HallucinationEvaluation(Evaluation):
         self.config = config
         self._company_id = event.company_id
         self._user_id = event.user_id
+        self._chat_id = event.payload.chat_id
+        self._assistant_id = event.payload.assistant_id
         self._reference_manager = reference_manager
         self._user_message = event.payload.user_message.text
         super().__init__(EvaluationMetricName.HALLUCINATION)
@@ -75,6 +77,8 @@ class HallucinationEvaluation(Evaluation):
                 output_text=loop_response.message.text,
             ),
             config=self.config,
+            chat_id=self._chat_id,
+            assistant_id=self._assistant_id,
         )
 
         score_to_label = self.config.score_to_label

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/unique_toolkit/agentic/evaluation/hallucination/service.py

@@ -59,4 +59,6 @@ class HallucinationEvaluator:
             user_id=self.event.user_id,
             input=input,
             config=config,
+            chat_id=self.event.payload.chat_id,
+            assistant_id=self.event.payload.assistant_id,
         )

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/unique_toolkit/agentic/evaluation/hallucination/utils.py

@@ -37,6 +37,8 @@ async def check_hallucination(
     user_id: str,
     input: EvaluationMetricInput,
     config: EvaluationMetricConfig,
+    chat_id: str | None = None,
+    assistant_id: str | None = None,
 ) -> EvaluationMetricResult:
     """
     Analyzes the level of hallucination in the generated output by comparing it with the provided input
@@ -77,7 +79,12 @@ async def check_hallucination(
     try:
         msgs = _get_msgs(input, config)
         result = await LanguageModelService.complete_async_util(
-            company_id=company_id, user_id=user_id, messages=msgs, model_name=model_name
+            company_id=company_id,
+            user_id=user_id,
+            messages=msgs,
+            model_name=model_name,
+            chat_id=chat_id,
+            assistant_id=assistant_id,
         )
     except Exception as e:
         error_message = "Error occurred during hallucination metric analysis"

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/unique_toolkit/agentic/evaluation/tests/test_hallucination_external_context.py

@@ -36,6 +36,8 @@ def _make_event() -> SimpleNamespace:
         company_id="company_1",
         user_id="user_1",
         payload=SimpleNamespace(
+            chat_id="chat_1",
+            assistant_id="assistant_1",
             user_message=SimpleNamespace(text="what is ACME revenue?"),
         ),
     )
@@ -61,8 +63,18 @@ def captured_context(monkeypatch: pytest.MonkeyPatch) -> dict:
     """Patch check_hallucination to capture the context_texts it receives."""
     captured: dict = {}
 
-    async def _fake_check_hallucination(*, company_id, user_id, input, config):
+    async def _fake_check_hallucination(
+        *,
+        company_id,
+        user_id,
+        input,
+        config,
+        chat_id,
+        assistant_id,
+    ):
         captured["context_texts"] = list(input.context_texts or [])
+        captured["chat_id"] = chat_id
+        captured["assistant_id"] = assistant_id
         return EvaluationMetricResult(
             name=EvaluationMetricName.HALLUCINATION,
             value="LOW",
@@ -86,6 +98,8 @@ async def test_external_context_is_appended(captured_context: dict) -> None:
 
     # No chunks → context comes solely from the external (MCP) source.
     assert captured_context["context_texts"] == ["ACME revenue: 1M"]
+    assert captured_context["chat_id"] == "chat_1"
+    assert captured_context["assistant_id"] == "assistant_1"
     assert result.value == "LOW"
     assert result.is_positive is True
 

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/unique_toolkit/agentic/evaluation/tests/test_hallucination_utils.py

@@ -1,6 +1,7 @@
 """Tests for hallucination evaluation utils."""
 
 from typing import List, Optional
+from unittest.mock import AsyncMock
 
 import pytest
 
@@ -14,6 +15,7 @@ from unique_toolkit.agentic.evaluation.hallucination.utils import (
     _from_order_source_selection_mode,
     _from_original_response_source_selection_mode,
     _get_msgs,
+    check_hallucination,
     context_text_from_stream_response,
 )
 from unique_toolkit.agentic.evaluation.schemas import EvaluationMetricInput
@@ -25,8 +27,11 @@ from unique_toolkit.language_model.infos import (
     ModelCapabilities,
 )
 from unique_toolkit.language_model.schemas import (
+    LanguageModelAssistantMessage,
+    LanguageModelCompletionChoice,
     LanguageModelMessageRole,
     LanguageModelMessages,
+    LanguageModelResponse,
     LanguageModelStreamResponse,
 )
 
@@ -101,6 +106,49 @@ def evaluation_input() -> EvaluationMetricInput:
     )
 
 
+@pytest.mark.ai
+@pytest.mark.asyncio
+async def test_check_hallucination__forwards_attribution_ids(
+    monkeypatch: pytest.MonkeyPatch,
+    hallucination_config: HallucinationConfig,
+    evaluation_input: EvaluationMetricInput,
+) -> None:
+    """Purpose: Verify hallucination completion receives chat attribution IDs.
+    Why this matters: Secondary evaluator usage must link to the originating chat.
+    Setup summary: Mock the completion utility, run the check, and inspect its call.
+    """
+    complete_async_util = AsyncMock(
+        return_value=LanguageModelResponse(
+            choices=[
+                LanguageModelCompletionChoice(
+                    index=0,
+                    message=LanguageModelAssistantMessage(
+                        content='{"value": "low", "reason": "grounded"}'
+                    ),
+                    finish_reason="stop",
+                )
+            ]
+        )
+    )
+    monkeypatch.setattr(
+        "unique_toolkit.agentic.evaluation.hallucination.utils."
+        "LanguageModelService.complete_async_util",
+        complete_async_util,
+    )
+
+    await check_hallucination(
+        company_id="company_1",
+        user_id="user_1",
+        input=evaluation_input.model_copy(update={"history_messages": []}),
+        config=hallucination_config,
+        chat_id="chat_1",
+        assistant_id="assistant_1",
+    )
+
+    assert complete_async_util.await_args.kwargs["chat_id"] == "chat_1"
+    assert complete_async_util.await_args.kwargs["assistant_id"] == "assistant_1"
+
+
 @pytest.mark.ai
 def test_default_source_selection_mode__selects_chunks__by_source_id_match(
     sample_references: List[ContentReference],

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/unique_toolkit/agentic/feature_flags/__init__.py

@@ -1,6 +1,7 @@
 from unique_toolkit.agentic.feature_flags.feature_flags import (
+    FeatureFlagNames,
     FeatureFlags,
     feature_flags,
 )
 
-__all__ = ["FeatureFlags", "feature_flags"]
+__all__ = ["FeatureFlagNames", "FeatureFlags", "feature_flags"]

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/unique_toolkit/agentic/feature_flags/feature_flags.py

@@ -1,11 +1,41 @@
+from enum import StrEnum
 from typing import Any
 
 from pydantic import Field, field_validator
 from pydantic_settings import BaseSettings, SettingsConfigDict
+from typing_extensions import deprecated
 
 from unique_toolkit._common.config_checker import register_config
 
 
+class FeatureFlagNames(StrEnum):
+    """Flag name constants used by unique_toolkit's own agentic features.
+
+    Values match both the env-var fallback names (``FeatureFlags`` below) and
+    the configuration-backend registry keys used by
+    ``unique_toolkit.experimental.resources.feature_flags.FeatureFlagClient``.
+    """
+
+    enable_selected_uploaded_files_un_18215 = (
+        "FEATURE_FLAG_ENABLE_SELECTED_UPLOADED_FILES_UN_18215"
+    )
+    enable_mcp_metadata_fallback_un_19145 = (
+        "FEATURE_FLAG_ENABLE_MCP_METADATA_FALLBACK_UN_19145"
+    )
+    enable_html_rendering_un_15131 = "FEATURE_FLAG_ENABLE_HTML_RENDERING_UN_15131"
+    enable_code_execution_fence_un_17972 = (
+        "FEATURE_FLAG_ENABLE_CODE_EXECUTION_FENCE_UN_17972"
+    )
+    enable_html_with_fence_un_17927 = "FEATURE_FLAG_ENABLE_HTML_WITH_FENCE_UN_17927"
+    enable_web_search_argument_screening_un_18741 = (
+        "FEATURE_FLAG_ENABLE_WEB_SEARCH_ARGUMENT_SCREENING_UN_18741"
+    )
+    enable_new_answers_ui_un_14411 = "FEATURE_FLAG_ENABLE_NEW_ANSWERS_UI_UN_14411"
+
+
+@deprecated(
+    "Use unique_toolkit.experimental.resources.feature_flags.is_flag_enabled instead"
+)
 class FeatureFlag:
     """A feature flag that can be enabled globally or for specific company IDs.
 
@@ -45,6 +75,9 @@ class FeatureFlag:
         return f"FeatureFlag({self.value})"
 
 
+@deprecated(
+    "Use unique_toolkit.experimental.resources.feature_flags.is_flag_enabled instead"
+)
 @register_config()
 class FeatureFlags(BaseSettings):
     """Feature flags loaded from environment variables.
@@ -145,5 +178,7 @@ class FeatureFlags(BaseSettings):
         return FeatureFlag(False)
 
 
-# Initialize once at module load - import this where needed
+# Deprecated, unused elsewhere in this toolkit — kept only for external
+# backward compatibility. Use unique_toolkit.experimental.resources.feature_flags
+# .is_flag_enabled instead.
 feature_flags = FeatureFlags()

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/unique_toolkit/agentic/history_manager/history_manager.py

@@ -348,8 +348,6 @@ class HistoryManager:
             int(m)
             for m in re.findall(r"\[source(\d+)\]", assistant_text, re.IGNORECASE)
         }
-        if not cited:
-            return records
 
         return [
             record.model_copy(
@@ -374,8 +372,9 @@ def _strip_uncited_sources_from_content(content: str, cited: set[int]) -> str:
 
     The content is expected to be a JSON array of dicts containing at least
     a "source_number" key.  Items whose "source_number" is not in *cited*
-    are removed.  If the content is not valid JSON or not in the expected
-    format it is returned unchanged.
+    are removed. If no sources are cited, the response is replaced with a short
+    message. If the content is not valid JSON or not in the expected format it
+    is returned unchanged.
     """
     try:
         data = json.loads(content)
@@ -390,5 +389,10 @@ def _strip_uncited_sources_from_content(content: str, cited: set[int]) -> str:
     ):
         return content
 
+    if not cited:
+        return (
+            "The search did not produce any relevant information that could be cited."
+        )
+
     filtered = [item for item in data if item.get("source_number") in cited]
     return serialize_tool_content_json(filtered)

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/unique_toolkit/agentic/history_manager/loop_token_reducer.py

@@ -109,7 +109,19 @@ class LoopTokenReducer:
         self._max_db_source_number: int = -1
         self._db_source_map: dict[int, ContentChunk] = {}
         self._enable_tool_call_persistence = enable_tool_call_persistence
-        self._selected_content_ids = get_selected_uploaded_content_ids(event)
+        self._event = event
+        # Lazily resolved on first async use since the flag check is async;
+        # `_resolved` distinguishes "not yet checked" from a resolved `None` (flag off).
+        self._selected_content_ids: set[str] | None = None
+        self._selected_content_ids_resolved = False
+
+    async def _get_selected_content_ids(self) -> set[str] | None:
+        if not self._selected_content_ids_resolved:
+            self._selected_content_ids = await get_selected_uploaded_content_ids(
+                self._event
+            )
+            self._selected_content_ids_resolved = True
+        return self._selected_content_ids
 
     @property
     def max_db_source_number(self) -> int:
@@ -354,6 +366,7 @@ class LoopTokenReducer:
         Returns:
             list[LanguageModelMessage]: The history
         """
+        selected_content_ids = await self._get_selected_content_ids()
         if self._enable_tool_call_persistence:
             (
                 full_history,
@@ -365,7 +378,7 @@ class LoopTokenReducer:
                 chat_service=self._chat_service,
                 content_service=self._content_service,
                 file_content_serializer=self._file_content_serializer,
-                selected_content_ids=self._selected_content_ids,
+                selected_content_ids=selected_content_ids,
             )
             self._max_db_source_number = max_src
             self._db_source_map = src_map
@@ -376,7 +389,7 @@ class LoopTokenReducer:
                 chat_service=self._chat_service,
                 content_service=self._content_service,
                 file_content_serializer=self._file_content_serializer,
-                selected_content_ids=self._selected_content_ids,
+                selected_content_ids=selected_content_ids,
             )
 
         if remove_from_text is not None:

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/unique_toolkit/agentic/history_manager/utils.py

@@ -3,10 +3,11 @@ import logging
 from copy import deepcopy
 from typing import Any
 
-from unique_toolkit.agentic.feature_flags import feature_flags
+from unique_toolkit.agentic.feature_flags import FeatureFlagNames
 from unique_toolkit.agentic.tools.schemas import Source
 from unique_toolkit.app.schemas import ChatEvent
 from unique_toolkit.content.schemas import ContentChunk
+from unique_toolkit.experimental.resources.feature_flags import is_flag_enabled
 from unique_toolkit.language_model.schemas import (
     LanguageModelAssistantMessage,
     LanguageModelMessage,
@@ -16,15 +17,16 @@ from unique_toolkit.language_model.schemas import (
 logger = logging.getLogger(__name__)
 
 
-def get_selected_uploaded_content_ids(event: ChatEvent) -> set[str] | None:
+async def get_selected_uploaded_content_ids(event: ChatEvent) -> set[str] | None:
     """Derive the set of content IDs the user explicitly selected.
 
     Returns ``None`` when the feature flag is disabled (meaning *all*
     uploaded images should be included), or a ``set[str]`` of IDs when
     only a subset should be attached.
     """
-    if not feature_flags.enable_selected_uploaded_files_un_18215.is_enabled(
-        event.company_id
+    if not await is_flag_enabled(
+        FeatureFlagNames.enable_selected_uploaded_files_un_18215,
+        company_id=event.company_id,
     ):
         return None
 

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/unique_toolkit/agentic/tools/a2a/tool/service.py

@@ -22,7 +22,7 @@ from unique_toolkit._common.referencing import (
 )
 from unique_toolkit._common.utils.jinja.render import render_template
 from unique_toolkit.agentic.evaluation.schemas import EvaluationMetricName
-from unique_toolkit.agentic.feature_flags import feature_flags
+from unique_toolkit.agentic.feature_flags import FeatureFlagNames
 from unique_toolkit.agentic.tools.a2a.response_watcher import SubAgentResponseWatcher
 from unique_toolkit.agentic.tools.a2a.tool._memory import (
     get_sub_agent_short_term_memory_manager,
@@ -53,6 +53,7 @@ from unique_toolkit.chat.schemas import (
 )
 from unique_toolkit.chat.service import ChatService
 from unique_toolkit.content import ContentChunk, ContentReference
+from unique_toolkit.experimental.resources.feature_flags import is_flag_enabled
 from unique_toolkit.language_model import (
     LanguageModelFunction,
     LanguageModelService,
@@ -431,11 +432,9 @@ class SubAgentTool(Tool[SubAgentToolConfig]):
         message: str,
         state: ProgressState,
     ) -> None:
-        if (
-            self._tool_progress_reporter is not None
-            and not feature_flags.enable_new_answers_ui_un_14411.is_enabled(
-                self._company_id
-            )
+        if self._tool_progress_reporter is not None and not await is_flag_enabled(
+            FeatureFlagNames.enable_new_answers_ui_un_14411,
+            company_id=self._company_id or "",
         ):
             await self._tool_progress_reporter.notify_from_tool_call(
                 tool_call=tool_call,

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/unique_toolkit/agentic/tools/mcp/tool_wrapper.py

@@ -9,7 +9,7 @@ from typing import Any, Dict
 import unique_sdk
 
 from unique_toolkit.agentic.evaluation.schemas import EvaluationMetricName
-from unique_toolkit.agentic.feature_flags import feature_flags
+from unique_toolkit.agentic.feature_flags import FeatureFlagNames
 from unique_toolkit.agentic.tools.mcp.models import MCPToolConfig
 from unique_toolkit.agentic.tools.schemas import ToolCallResponse
 from unique_toolkit.agentic.tools.tool import Tool
@@ -21,6 +21,7 @@ from unique_toolkit.app.schemas import ChatEvent, McpServer, McpTool
 from unique_toolkit.chat.schemas import MessageLog, MessageLogStatus
 from unique_toolkit.chat.service import ChatService
 from unique_toolkit.content.functions import upload_content_from_bytes
+from unique_toolkit.experimental.resources.feature_flags import is_flag_enabled
 from unique_toolkit.language_model import LanguageModelService
 from unique_toolkit.language_model.schemas import (
     LanguageModelFunction,
@@ -133,11 +134,9 @@ class MCPToolWrapper(Tool[MCPToolConfig]):
         )
 
         # Notify progress if reporter is available
-        if (
-            self._tool_progress_reporter
-            and not feature_flags.enable_new_answers_ui_un_14411.is_enabled(
-                self._event.company_id
-            )
+        if self._tool_progress_reporter and not await is_flag_enabled(
+            FeatureFlagNames.enable_new_answers_ui_un_14411,
+            company_id=self._event.company_id,
         ):
             await self._tool_progress_reporter.notify_from_tool_call(
                 tool_call=tool_call,
@@ -172,11 +171,9 @@ class MCPToolWrapper(Tool[MCPToolConfig]):
             )
 
             # Notify completion
-            if (
-                self._tool_progress_reporter
-                and not feature_flags.enable_new_answers_ui_un_14411.is_enabled(
-                    self._event.company_id
-                )
+            if self._tool_progress_reporter and not await is_flag_enabled(
+                FeatureFlagNames.enable_new_answers_ui_un_14411,
+                company_id=self._event.company_id,
             ):
                 await self._tool_progress_reporter.notify_from_tool_call(
                     tool_call=tool_call,
@@ -198,11 +195,9 @@ class MCPToolWrapper(Tool[MCPToolConfig]):
             self.logger.error(f"Error executing MCP tool {self.name}: {e}")
 
             # Notify failure
-            if (
-                self._tool_progress_reporter
-                and not feature_flags.enable_new_answers_ui_un_14411.is_enabled(
-                    self._event.company_id
-                )
+            if self._tool_progress_reporter and not await is_flag_enabled(
+                FeatureFlagNames.enable_new_answers_ui_un_14411,
+                company_id=self._event.company_id,
             ):
                 await self._tool_progress_reporter.notify_from_tool_call(
                     tool_call=tool_call,

{unique_toolkit-2026.32.4 → unique_toolkit-2026.34.0}/unique_toolkit/agentic/tools/openai_builtin/code_interpreter/postprocessors/code_display.py

@@ -6,11 +6,15 @@ from typing import override
 from pydantic import BaseModel, Field
 from pydantic.json_schema import SkipJsonSchema
 
-from unique_toolkit.agentic.feature_flags.feature_flags import feature_flags
+from unique_toolkit.agentic.feature_flags import FeatureFlagNames
 from unique_toolkit.agentic.postprocessor.postprocessor_manager import (
     ResponsesApiPostprocessor,
 )
 from unique_toolkit.agentic.tools.config import get_configuration_dict
+from unique_toolkit.experimental.resources.feature_flags import (
+    COMPANY_ID_PLACEHOLDER,
+    is_flag_enabled,
+)
 from unique_toolkit.language_model.schemas import ResponsesLanguageModelStreamResponse
 
 _TEMPLATE = """
@@ -56,15 +60,15 @@ class ShowExecutedCodePostprocessor(ResponsesApiPostprocessor):
         super().__init__(self.__class__.__name__)
         self._config = config
         self._company_id = company_id
-        self._is_enabled = (
-            self._config.enable
-            and not feature_flags.enable_code_execution_fence_un_17972.is_enabled(
-                company_id
-            )
-        )
+        # Resolved in run() (before apply_postprocessing_to_response) since flag evaluation is async.
+        self._is_enabled = False
 
     @override
     async def run(self, loop_response: ResponsesLanguageModelStreamResponse) -> None:
+        self._is_enabled = self._config.enable and not await is_flag_enabled(
+            FeatureFlagNames.enable_code_execution_fence_un_17972,
+            company_id=self._company_id or COMPANY_ID_PLACEHOLDER,
+        )
         if self._is_enabled:
             await asyncio.sleep(self._config.sleep_time_before_display)