unique_orchestrator 2026.30.0.dev0__tar.gz → 2026.30.0.dev2__tar.gz

{unique_orchestrator-2026.30.0.dev0 → unique_orchestrator-2026.30.0.dev2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: unique-orchestrator
-Version: 2026.30.0.dev0
+Version: 2026.30.0.dev2
 Summary: 
 Author: Andreas Hauri
 Author-email: Andreas Hauri <andreas.hauri@unique.ai>
@@ -9,12 +9,12 @@ Requires-Dist: pydantic>=2.8.2,<3
 Requires-Dist: pydantic-settings>=2.10.1,<3
 Requires-Dist: typing-extensions>=4.9.0,<5
 Requires-Dist: jinja2>=3.1.0,<4
-Requires-Dist: unique-toolkit>=2026.30.0.dev0,<2026.30.0rc0
+Requires-Dist: unique-toolkit>=2026.30.0.dev13,<2026.30.0rc0
 Requires-Dist: unique-stock-ticker>=2026.30.0.dev0,<2026.30.0rc0
 Requires-Dist: unique-follow-up-questions>=2026.30.0.dev0,<2026.30.0rc0
 Requires-Dist: unique-internal-search>=2026.30.0.dev0,<2026.30.0rc0
 Requires-Dist: unique-deep-research>=2026.30.0.dev0,<2026.30.0rc0
-Requires-Dist: unique-web-search>=2026.30.0.dev0,<2026.30.0rc0
+Requires-Dist: unique-web-search>=2026.30.0.dev2,<2026.30.0rc0
 Requires-Dist: unique-swot>=2026.30.0.dev0,<2026.30.0rc0
 Requires-Dist: unique-skill-tool>=2026.30.0.dev0,<2026.30.0rc0
 Requires-Dist: unique-user-memory>=2026.30.0.dev0,<2026.30.0rc0

{unique_orchestrator-2026.30.0.dev0 → unique_orchestrator-2026.30.0.dev2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "unique_orchestrator"
-version = "2026.30.0.dev0"
+version = "2026.30.0.dev2"
 description = ""
 readme = "README.md"
 license = { text = "Proprietary" }
@@ -13,12 +13,12 @@ dependencies = [
     "pydantic-settings>=2.10.1,<3",
     "typing-extensions>=4.9.0,<5",
     "jinja2>=3.1.0,<4",
-    "unique-toolkit>=2026.30.0.dev0,<2026.30.0rc0",
+    "unique-toolkit>=2026.30.0.dev13,<2026.30.0rc0",
     "unique-stock-ticker>=2026.30.0.dev0,<2026.30.0rc0",
     "unique-follow-up-questions>=2026.30.0.dev0,<2026.30.0rc0",
     "unique-internal-search>=2026.30.0.dev0,<2026.30.0rc0",
     "unique-deep-research>=2026.30.0.dev0,<2026.30.0rc0",
-    "unique-web-search>=2026.30.0.dev0,<2026.30.0rc0",
+    "unique-web-search>=2026.30.0.dev2,<2026.30.0rc0",
     "unique-swot>=2026.30.0.dev0,<2026.30.0rc0",
     "unique-skill-tool>=2026.30.0.dev0,<2026.30.0rc0",
     "unique-user-memory>=2026.30.0.dev0,<2026.30.0rc0",

unique_orchestrator-2026.30.0.dev2/unique_orchestrator/_builders/history_manager.py

@@ -0,0 +1,97 @@
+from functools import partial
+from logging import Logger
+
+from unique_internal_search.uploaded_search.service import UploadedSearchTool
+from unique_toolkit.agentic.history_manager import (
+    history_manager as history_manager_module,
+)
+from unique_toolkit.agentic.history_manager.history_construction_with_contents import (
+    FileContentSerializer,
+)
+from unique_toolkit.agentic.history_manager.history_manager import (
+    HistoryManager,
+    HistoryManagerConfig,
+)
+from unique_toolkit.agentic.reference_manager.reference_manager import ReferenceManager
+from unique_toolkit.agentic.tools.openai_builtin.base import OpenAIBuiltInToolName
+from unique_toolkit.agentic.tools.openai_builtin.code_interpreter.postprocessors.artifacts import (
+    load_code_execution_metadata,
+)
+from unique_toolkit.agentic.tools.tool_manager import (
+    ResponsesApiToolManager,
+    ToolManager,
+)
+from unique_toolkit.app.schemas import ChatEvent
+from unique_toolkit.content import Content
+
+from unique_orchestrator.config import UniqueAIConfig
+
+
+def serialize_uploaded_file_for_history(
+    content: Content,
+    *,
+    uploaded_search_available: bool,
+    code_interpreter_available: bool,
+) -> str | None:
+    """Serialize user-uploaded file metadata for model history."""
+    if load_code_execution_metadata(content) is not None:
+        return None
+
+    lines = [f"User uploaded file: {content.key} ({content.id})"]
+    if (
+        uploaded_search_available
+        and content.is_ingested(default_if_unknown=True)
+        and not content.is_expired()
+    ):
+        lines.append("- Searchable using UploadedSearchTool")
+    if code_interpreter_available:
+        lines.append("- Available for processing in the code execution container")
+    return "\n".join(lines)
+
+
+def _get_file_content_serializer(
+    config: UniqueAIConfig,
+    tool_manager: ToolManager | ResponsesApiToolManager,
+) -> FileContentSerializer | None:
+    if not config.agent.input_token_distribution.serialize_uploaded_files_in_user_message:
+        return None
+
+    return partial(
+        serialize_uploaded_file_for_history,
+        uploaded_search_available=(
+            tool_manager.get_tool_by_name(UploadedSearchTool.name) is not None
+        ),
+        code_interpreter_available=(
+            tool_manager.get_tool_by_name(OpenAIBuiltInToolName.CODE_INTERPRETER)
+            is not None
+        ),
+    )
+
+
+def build_history_manager(
+    *,
+    event: ChatEvent,
+    logger: Logger,
+    config: UniqueAIConfig,
+    reference_manager: ReferenceManager,
+    tool_manager: ToolManager | ResponsesApiToolManager,
+) -> HistoryManager:
+    """Build a history manager after the active tools are known."""
+    history_manager_config = HistoryManagerConfig(
+        experimental_features=history_manager_module.ExperimentalFeatures(),
+        percent_of_max_tokens_for_history=config.agent.input_token_distribution.percent_for_history,
+        language_model=config.space.language_model,
+        uploaded_content_config=config.agent.services.uploaded_content_config,
+        enable_tool_call_persistence=config.agent.input_token_distribution.enable_tool_call_persistence,
+    )
+    return HistoryManager(
+        logger,
+        event,
+        history_manager_config,
+        config.space.language_model,
+        reference_manager,
+        file_content_serializer=_get_file_content_serializer(
+            config=config,
+            tool_manager=tool_manager,
+        ),
+    )

{unique_orchestrator-2026.30.0.dev0 → unique_orchestrator-2026.30.0.dev2}/unique_orchestrator/_builders/open_file_setup.py

@@ -8,30 +8,16 @@ from __future__ import annotations
 
 from datetime import datetime, timezone
 from logging import Logger
-from typing import TYPE_CHECKING
 
-from unique_toolkit.agentic.history_manager import (
-    history_manager as history_manager_module,
-)
-from unique_toolkit.agentic.history_manager.history_manager import (
-    HistoryManager,
-    HistoryManagerConfig,
-)
 from unique_toolkit.agentic.tools.experimental.open_file_tool import OpenFileTool
 from unique_toolkit.agentic.tools.tool_manager import (
     ResponsesApiToolManager,
 )
 from unique_toolkit.app.schemas import ChatEvent
 from unique_toolkit.content import Content
-from unique_toolkit.language_model.infos import LanguageModelInfo
 
 from unique_orchestrator.config import UniqueAIConfig
 
-if TYPE_CHECKING:
-    from unique_toolkit.agentic.reference_manager.reference_manager import (
-        ReferenceManager,
-    )
-
 
 def handle_uploaded_file_tool_choices(
     config: UniqueAIConfig,
@@ -76,39 +62,15 @@ def handle_uploaded_file_tool_choices(
 def configure_file_payload(
     config: UniqueAIConfig,
     event: ChatEvent,
-    logger: Logger,
-    history_manager: HistoryManager,
-    reference_manager: ReferenceManager,
-    language_model: LanguageModelInfo,
     tool_manager: ResponsesApiToolManager,
-) -> tuple[HistoryManager, list[str]]:
+) -> list[str]:
     """Configure file-in-payload handling for the Responses API.
 
-    When sending uploaded files as file parts, disables the UploadedContentConfig
-    mechanism so the HistoryManager doesn't also inject uploaded content as text
-    (which would duplicate what the input_file parts already provide).
+    Registers the OpenFileTool when send_files_in_payload is enabled, backed by
+    a shared registry for agent-requested KB file IDs.
 
-    Also registers the OpenFileTool when send_files_in_payload is enabled,
-    backed by a shared registry for agent-requested KB file IDs.
-
-    Returns the (possibly updated) history_manager and the agent file registry.
+    Returns the registry shared with the OpenFileTool.
     """
-    if config.agent.experimental.open_file_tool_config.send_uploaded_files_in_payload:
-        upload_free_config = HistoryManagerConfig(
-            experimental_features=history_manager_module.ExperimentalFeatures(),
-            percent_of_max_tokens_for_history=config.agent.input_token_distribution.percent_for_history,
-            language_model=language_model,
-            uploaded_content_config=None,
-            enable_tool_call_persistence=config.agent.input_token_distribution.enable_tool_call_persistence,
-        )
-        history_manager = HistoryManager(
-            logger,
-            event,
-            upload_free_config,
-            language_model,
-            reference_manager,
-        )
-
     agent_file_registry: list[str] = []
 
     if config.agent.experimental.open_file_tool_config.send_files_in_payload:
@@ -120,4 +82,4 @@ def configure_file_payload(
             )
         )
 
-    return history_manager, agent_file_registry
+    return agent_file_registry

{unique_orchestrator-2026.30.0.dev0 → unique_orchestrator-2026.30.0.dev2}/unique_orchestrator/config.py

@@ -348,6 +348,11 @@ class HistoryConfig(BaseToolConfig):
         description="Persist tool calls and reconstruct tool call history across turns.",
     )
 
+    serialize_uploaded_files_in_user_message: bool = Field(
+        default=False,
+        description="Include an indication of each uploaded file in the user message.",
+    )
+
     def max_history_tokens(self, max_input_token: int) -> int:
         return int(self.percent_for_history * max_input_token)
 
@@ -457,6 +462,10 @@ class ExperimentalConfig(BaseToolConfig):
         description="If set, the main agent will use the Responses API from OpenAI",
     )
 
+    use_experimental_python_streaming: bool = Field(
+        default=False,
+    )
+
 
 class UniqueAIAgentConfig(BaseToolConfig):
     max_loop_iterations: Annotated[

{unique_orchestrator-2026.30.0.dev0 → unique_orchestrator-2026.30.0.dev2}/unique_orchestrator/tests/test_unique_ai_plan_or_execute_include.py

@@ -9,7 +9,10 @@ from unittest.mock import AsyncMock, MagicMock
 import pytest
 from unique_toolkit.agentic.tools.tool_manager import ResponsesApiToolManager
 
-from unique_orchestrator.unique_ai_builder import ResponsesStreamingHandler
+from unique_orchestrator.unique_ai_builder import (
+    ResponsesStreamingHandler,
+    _ChatServiceResponsesStreaming,
+)
 
 
 def _make_handler(
@@ -17,15 +20,11 @@ def _make_handler(
 ) -> ResponsesStreamingHandler:
     tool_manager = MagicMock(spec=ResponsesApiToolManager)
     tool_manager.get_required_include_params.return_value = include_params
-    chat_service = MagicMock()
-    chat_service.complete_responses_with_references_async = AsyncMock(
-        return_value=MagicMock()
-    )
-    chat_service.complete_responses_with_references = MagicMock(
-        return_value=MagicMock()
-    )
+    inner = MagicMock()
+    inner.complete_with_references_async = AsyncMock(return_value=MagicMock())
+    inner.complete_with_references = MagicMock(return_value=MagicMock())
     return ResponsesStreamingHandler(
-        chat_service=chat_service,
+        inner=inner,
         tool_manager=tool_manager,
     )
 
@@ -45,9 +44,7 @@ async def test_responses_streaming_handler__injects_include__when_tool_manager_r
 
     await handler.complete_with_references_async(messages=[], model_name="model")
 
-    call_kwargs = (
-        handler._chat_service.complete_responses_with_references_async.await_args.kwargs
-    )
+    call_kwargs = handler._inner.complete_with_references_async.await_args.kwargs
     assert call_kwargs["include"] == ["code_interpreter_call.outputs"]
     handler._tool_manager.get_required_include_params.assert_called_once()
 
@@ -66,9 +63,7 @@ async def test_responses_streaming_handler__omits_include__when_tool_manager_ret
 
     await handler.complete_with_references_async(messages=[], model_name="model")
 
-    call_kwargs = (
-        handler._chat_service.complete_responses_with_references_async.await_args.kwargs
-    )
+    call_kwargs = handler._inner.complete_with_references_async.await_args.kwargs
     assert "include" not in call_kwargs
 
 
@@ -81,9 +76,7 @@ def test_responses_streaming_handler__injects_include__sync_path() -> None:
 
     handler.complete_with_references(messages=[], model_name="model")
 
-    call_kwargs = (
-        handler._chat_service.complete_responses_with_references.call_args.kwargs
-    )
+    call_kwargs = handler._inner.complete_with_references.call_args.kwargs
     assert call_kwargs["include"] == ["code_interpreter_call.outputs"]
 
 
@@ -94,7 +87,34 @@ def test_responses_streaming_handler__omits_include__sync_path_when_empty() -> N
 
     handler.complete_with_references(messages=[], model_name="model")
 
-    call_kwargs = (
-        handler._chat_service.complete_responses_with_references.call_args.kwargs
-    )
+    call_kwargs = handler._inner.complete_with_references.call_args.kwargs
     assert "include" not in call_kwargs
+
+
+@pytest.mark.ai
+@pytest.mark.asyncio
+async def test_chat_service_responses_streaming__forwards_to_responses_methods() -> (
+    None
+):
+    """
+    Purpose: The legacy adapter must route the protocol's
+    `complete_with_references[_async]` to ChatService's Responses-specific
+    methods, not the Chat Completions ones.
+    Why this matters: ChatService exposes both; routing to the wrong method
+    would silently stream through the Chat Completions API.
+    """
+    chat_service = MagicMock()
+    chat_service.complete_responses_with_references = MagicMock(
+        return_value=MagicMock()
+    )
+    chat_service.complete_responses_with_references_async = AsyncMock(
+        return_value=MagicMock()
+    )
+    adapter = _ChatServiceResponsesStreaming(chat_service)
+
+    adapter.complete_with_references(messages=[], model_name="model")
+    await adapter.complete_with_references_async(messages=[], model_name="model")
+
+    chat_service.complete_responses_with_references.assert_called_once()
+    chat_service.complete_responses_with_references_async.assert_awaited_once()
+    chat_service.complete_with_references.assert_not_called()

{unique_orchestrator-2026.30.0.dev0 → unique_orchestrator-2026.30.0.dev2}/unique_orchestrator/unique_ai_builder.py

@@ -20,13 +20,7 @@ from unique_toolkit.agentic.evaluation.evaluation_manager import EvaluationManag
 from unique_toolkit.agentic.evaluation.hallucination.hallucination_evaluation import (
     HallucinationEvaluation,
 )
-from unique_toolkit.agentic.history_manager import (
-    history_manager as history_manager_module,
-)
-from unique_toolkit.agentic.history_manager.history_manager import (
-    HistoryManager,
-    HistoryManagerConfig,
-)
+from unique_toolkit.agentic.history_manager.history_manager import HistoryManager
 from unique_toolkit.agentic.message_log_manager.service import MessageStepLogger
 from unique_toolkit.agentic.postprocessor.postprocessor_manager import (
     PostprocessorManager,
@@ -65,9 +59,13 @@ from unique_toolkit.agentic.tools.tool_manager import (
 )
 from unique_toolkit.agentic.tools.tool_progress_reporter import ToolProgressReporter
 from unique_toolkit.app.schemas import ChatEvent, McpServer
+from unique_toolkit.app.unique_settings import UniqueSettings
 from unique_toolkit.chat.service import ChatService
 from unique_toolkit.content import Content
 from unique_toolkit.content.service import ContentService
+from unique_toolkit.experimental.integrations.openai.streaming.event_routing import (
+    ResponsesCompleteWithReferences,
+)
 from unique_toolkit.language_model.infos import LanguageModelInfo, ModelCapabilities
 from unique_toolkit.protocols.support import ResponsesSupportCompleteWithReferences
 from unique_user_memory.user_memory import load_user_memory
@@ -77,6 +75,7 @@ from unique_orchestrator._builders import (
     build_loop_iteration_runner,
     build_responses_loop_iteration_runner,
 )
+from unique_orchestrator._builders.history_manager import build_history_manager
 from unique_orchestrator._builders.open_file_setup import (
     configure_file_payload,
     handle_uploaded_file_tool_choices,
@@ -94,19 +93,45 @@ from unique_orchestrator.unique_ai import UniqueAI
 from unique_orchestrator.utils import filter_uploaded_documents_by_selection
 
 
+def _configure_follow_up_questions(
+    *,
+    config: UniqueAIConfig,
+    event: ChatEvent,
+    logger: Logger,
+    history_manager: HistoryManager,
+    postprocessor_manager: PostprocessorManager,
+) -> None:
+    follow_up_config = config.agent.services.follow_up_questions_config
+    if follow_up_config and follow_up_config.number_of_questions > 0:
+        postprocessor_manager.set_last_postprocessor(
+            FollowUpPostprocessor(
+                logger=logger,
+                config=follow_up_config,
+                event=event,
+                historyManager=history_manager,
+                llm_service=LanguageModelService.from_event(event),
+            )
+        )
+
+
 class ResponsesStreamingHandler(ResponsesSupportCompleteWithReferences):
     """Streaming handler for Responses API runs.
 
     Injects `include` params from the tool manager on every call so the
     orchestrator loop stays generic (no isinstance checks in UniqueAI).
+
+    Delegates the actual streaming to an injected
+    `ResponsesSupportCompleteWithReferences` (the event-routing
+    `ResponsesCompleteWithReferences` orchestrator), so this class only owns
+    the `include` injection.
     """
 
     def __init__(
         self,
-        chat_service: ChatService,
+        inner: ResponsesSupportCompleteWithReferences,
         tool_manager: ResponsesApiToolManager,
     ) -> None:
-        self._chat_service: ChatService = chat_service
+        self._inner: ResponsesSupportCompleteWithReferences = inner
         self._tool_manager: ResponsesApiToolManager = tool_manager
 
     @override
@@ -114,13 +139,26 @@ class ResponsesStreamingHandler(ResponsesSupportCompleteWithReferences):
         include = self._tool_manager.get_required_include_params()
         if include:
             kwargs["include"] = include
-        return self._chat_service.complete_responses_with_references(*args, **kwargs)
+        return self._inner.complete_with_references(*args, **kwargs)
 
     @override
     async def complete_with_references_async(self, *args: Any, **kwargs: Any) -> Any:
         include = self._tool_manager.get_required_include_params()
         if include:
             kwargs["include"] = include
+        return await self._inner.complete_with_references_async(*args, **kwargs)
+
+
+class _ChatServiceResponsesStreaming(ResponsesSupportCompleteWithReferences):
+    def __init__(self, chat_service: ChatService) -> None:
+        self._chat_service: ChatService = chat_service
+
+    @override
+    def complete_with_references(self, *args: Any, **kwargs: Any) -> Any:
+        return self._chat_service.complete_responses_with_references(*args, **kwargs)
+
+    @override
+    async def complete_with_references_async(self, *args: Any, **kwargs: Any) -> Any:
         return await self._chat_service.complete_responses_with_references_async(
             *args, **kwargs
         )
@@ -168,7 +206,6 @@ class _CommonComponents(NamedTuple):
     uploaded_images: list[Content]
     thinking_manager: ThinkingManager
     reference_manager: ReferenceManager
-    history_manager: HistoryManager
     evaluation_manager: EvaluationManager
     postprocessor_manager: PostprocessorManager
     message_step_logger: MessageStepLogger
@@ -290,21 +327,6 @@ async def _build_common(
 
     reference_manager = ReferenceManager()
 
-    history_manager_config = HistoryManagerConfig(
-        experimental_features=history_manager_module.ExperimentalFeatures(),
-        percent_of_max_tokens_for_history=config.agent.input_token_distribution.percent_for_history,
-        language_model=config.space.language_model,
-        uploaded_content_config=config.agent.services.uploaded_content_config,
-        enable_tool_call_persistence=config.agent.input_token_distribution.enable_tool_call_persistence,
-    )
-    history_manager = HistoryManager(
-        logger,
-        event,
-        history_manager_config,
-        config.space.language_model,
-        reference_manager,
-    )
-
     evaluation_manager = EvaluationManager(logger=logger, chat_service=chat_service)
     if config.agent.services.evaluation_config:
         evaluation_manager.add_evaluation(
@@ -348,21 +370,6 @@ async def _build_common(
             )
         )
 
-    if (
-        config.agent.services.follow_up_questions_config
-        and config.agent.services.follow_up_questions_config.number_of_questions > 0
-    ):
-        # Should run last to make sure the follow up questions are displayed last.
-        postprocessor_manager.set_last_postprocessor(
-            FollowUpPostprocessor(
-                logger=logger,
-                config=config.agent.services.follow_up_questions_config,
-                event=event,
-                historyManager=history_manager,
-                llm_service=LanguageModelService.from_event(event),
-            )
-        )
-
     user_memory_text = ""
     if config.space.allow_user_memory:
         user_memory_config = config.agent.services.user_memory_config
@@ -401,7 +408,6 @@ async def _build_common(
         uploaded_images=uploaded_images,
         thinking_manager=thinking_manager,
         reference_manager=reference_manager,
-        history_manager=history_manager,
         evaluation_manager=evaluation_manager,
         tool_progress_reporter=tool_progress_reporter,
         mcp_manager=mcp_manager,
@@ -543,18 +549,26 @@ async def _build_responses(
 
     agent_file_registry: list[str] = []
     if config.agent.experimental.open_file_tool_config.enabled:
-        history_manager, agent_file_registry = configure_file_payload(
+        agent_file_registry = configure_file_payload(
             config,
             event,
-            logger,
-            common_components.history_manager,
-            common_components.reference_manager,
-            config.space.language_model,
             tool_manager,
         )
-        common_components = common_components._replace(
-            history_manager=history_manager,
-        )
+
+    history_manager = build_history_manager(
+        event=event,
+        logger=logger,
+        config=config,
+        reference_manager=common_components.reference_manager,
+        tool_manager=tool_manager,
+    )
+    _configure_follow_up_questions(
+        config=config,
+        event=event,
+        logger=logger,
+        history_manager=history_manager,
+        postprocessor_manager=postprocessor_manager,
+    )
 
     await configure_skill_tool(
         config=config,
@@ -568,12 +582,26 @@ async def _build_responses(
 
     loop_iteration_runner = build_responses_loop_iteration_runner(
         config=config,
-        history_manager=common_components.history_manager,
+        history_manager=history_manager,
         openai_client=client,
     )
 
+    if config.agent.experimental.use_experimental_python_streaming:
+        inner: ResponsesSupportCompleteWithReferences = ResponsesCompleteWithReferences(
+            UniqueSettings.from_chat_event(event),
+            additional_headers={
+                "x-model": config.space.language_model.name,
+                "x-user-id": event.user_id,
+                "x-company-id": event.company_id,
+                "x-assistant-id": event.payload.assistant_id,
+                "x-chat-id": event.payload.chat_id,
+            },
+        )
+    else:
+        inner = _ChatServiceResponsesStreaming(common_components.chat_service)
+
     streaming_handler = ResponsesStreamingHandler(
-        chat_service=common_components.chat_service,
+        inner=inner,
         tool_manager=tool_manager,
     )
 
@@ -601,7 +629,7 @@ async def _build_responses(
         tool_manager=tool_manager,
         thinking_manager=common_components.thinking_manager,
         streaming_handler=streaming_handler,
-        history_manager=common_components.history_manager,
+        history_manager=history_manager,
         reference_manager=common_components.reference_manager,
         evaluation_manager=common_components.evaluation_manager,
         postprocessor_manager=postprocessor_manager,
@@ -641,6 +669,21 @@ async def _build_completions(
     if force_uploaded_search:
         tool_manager.add_forced_tool(UploadedSearchTool.name)
 
+    history_manager = build_history_manager(
+        event=event,
+        logger=logger,
+        config=config,
+        reference_manager=common_components.reference_manager,
+        tool_manager=tool_manager,
+    )
+    _configure_follow_up_questions(
+        config=config,
+        event=event,
+        logger=logger,
+        history_manager=history_manager,
+        postprocessor_manager=common_components.postprocessor_manager,
+    )
+
     await configure_skill_tool(
         config=config,
         logger=logger,
@@ -669,7 +712,7 @@ async def _build_completions(
 
     loop_iteration_runner = build_loop_iteration_runner(
         config=config,
-        history_manager=common_components.history_manager,
+        history_manager=history_manager,
         llm_service=common_components.llm_service,
         chat_service=common_components.chat_service,
     )
@@ -683,7 +726,7 @@ async def _build_completions(
         uploaded_documents=common_components.uploaded_documents,
         tool_manager=tool_manager,
         thinking_manager=common_components.thinking_manager,
-        history_manager=common_components.history_manager,
+        history_manager=history_manager,
         reference_manager=common_components.reference_manager,
         streaming_handler=common_components.chat_service,
         evaluation_manager=common_components.evaluation_manager,