unique-toolkit 2026.36.0.dev1__tar.gz → 2026.36.0.dev2__tar.gz

{unique_toolkit-2026.36.0.dev1 → unique_toolkit-2026.36.0.dev2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: unique-toolkit
-Version: 2026.36.0.dev1
+Version: 2026.36.0.dev2
 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-2026.36.0.dev1 → unique_toolkit-2026.36.0.dev2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "unique_toolkit"
-version = "2026.36.0.dev1"
+version = "2026.36.0.dev2"
 description = ""
 readme = "README.md"
 requires-python = ">=3.12"

{unique_toolkit-2026.36.0.dev1 → unique_toolkit-2026.36.0.dev2}/unique_toolkit/language_model/functions.py

@@ -46,12 +46,15 @@ logger = logging.getLogger(f"toolkit.language_model.{__name__}")
 def _completion_headers(
     chat_id: str | None,
     assistant_id: str | None,
+    assistant_message_id: str | None = None,
 ) -> dict[str, str] | None:
     headers: dict[str, str] = {}
     if chat_id:
         headers["x-chat-id"] = chat_id
     if assistant_id:
         headers["x-assistant-id"] = assistant_id
+    if assistant_message_id:
+        headers["x-assistant-message-id"] = assistant_message_id
     return headers or None
 
 
@@ -68,6 +71,7 @@ def complete(
     user_id: str | None = None,
     chat_id: str | None = None,
     assistant_id: str | None = None,
+    assistant_message_id: str | None = None,
 ) -> LanguageModelResponse:
     """Call the completion endpoint synchronously without streaming the response.
 
@@ -103,7 +107,7 @@ def complete(
         )
 
     try:
-        headers = _completion_headers(chat_id, assistant_id)
+        headers = _completion_headers(chat_id, assistant_id, assistant_message_id)
         if headers is None:
             response = unique_sdk.ChatCompletion.create(
                 company_id=company_id,
@@ -142,6 +146,7 @@ async def complete_async(
     structured_output_enforce_schema: bool = False,
     chat_id: str | None = None,
     assistant_id: str | None = None,
+    assistant_message_id: str | None = None,
 ) -> LanguageModelResponse:
     """Call the completion endpoint asynchronously without streaming the response.
 
@@ -186,7 +191,7 @@ async def complete_async(
     )
 
     try:
-        headers = _completion_headers(chat_id, assistant_id)
+        headers = _completion_headers(chat_id, assistant_id, assistant_message_id)
         if headers is None:
             response = await unique_sdk.ChatCompletion.create_async(
                 company_id=company_id,
@@ -502,6 +507,7 @@ def complete_with_references(
     other_options: dict[str, Any] | None = None,
     chat_id: str | None = None,
     assistant_id: str | None = None,
+    assistant_message_id: str | None = None,
 ) -> LanguageModelStreamResponse:
     # Use toolkit language model functions for chat completion
 
@@ -516,6 +522,7 @@ def complete_with_references(
         other_options=other_options,
         chat_id=chat_id,
         assistant_id=assistant_id,
+        assistant_message_id=assistant_message_id,
     )
 
     return _create_language_model_stream_response_with_references(
@@ -539,6 +546,7 @@ async def complete_with_references_async(
     other_options: dict[str, Any] | None = None,
     chat_id: str | None = None,
     assistant_id: str | None = None,
+    assistant_message_id: str | None = None,
 ) -> LanguageModelStreamResponse:
     # Use toolkit language model functions for chat completion
     response = await complete_async(
@@ -552,6 +560,7 @@ async def complete_with_references_async(
         other_options=other_options,
         chat_id=chat_id,
         assistant_id=assistant_id,
+        assistant_message_id=assistant_message_id,
     )
 
     return _create_language_model_stream_response_with_references(

{unique_toolkit-2026.36.0.dev1 → unique_toolkit-2026.36.0.dev2}/unique_toolkit/language_model/service.py

@@ -5,7 +5,11 @@ from pydantic import BaseModel
 from typing_extensions import deprecated
 
 from unique_toolkit._common.validate_required_values import validate_required_values
-from unique_toolkit.app.schemas import AssistantWebhookEvent, BaseEvent
+from unique_toolkit.app.schemas import (
+    AssistantWebhookEvent,
+    BaseEvent,
+    ChatEventPayload,
+)
 from unique_toolkit.app.unique_settings import UniqueSettings
 from unique_toolkit.content.schemas import ContentChunk
 from unique_toolkit.language_model.constants import (
@@ -67,10 +71,13 @@ class LanguageModelService:
             self._event = event
             self._chat_id = None
             self._assistant_id = None
+            self._assistant_message_id = None
             if isinstance(event, AssistantWebhookEvent):
                 payload = event.payload
                 self._chat_id = payload.chat_id
                 self._assistant_id = payload.assistant_id
+                if isinstance(payload, ChatEventPayload):
+                    self._assistant_message_id = payload.assistant_message.id
             return
 
         [company_id, user_id] = validate_required_values([company_id, user_id])
@@ -79,6 +86,7 @@ class LanguageModelService:
         self._user_id: str = user_id
         self._chat_id: str | None = None
         self._assistant_id: str | None = None
+        self._assistant_message_id: str | None = None
 
     @classmethod
     def from_event(cls, event: BaseEvent[Any]):
@@ -89,6 +97,8 @@ class LanguageModelService:
         if isinstance(event, AssistantWebhookEvent):
             service._chat_id = event.payload.chat_id
             service._assistant_id = event.payload.assistant_id
+            if isinstance(event.payload, ChatEventPayload):
+                service._assistant_message_id = event.payload.assistant_message.id
         return service
 
     @classmethod
@@ -251,6 +261,7 @@ class LanguageModelService:
             structured_output_enforce_schema=structured_output_enforce_schema,
             chat_id=self._chat_id,
             assistant_id=self._assistant_id,
+            assistant_message_id=self._assistant_message_id,
         )
 
     async def complete_async(
@@ -281,6 +292,7 @@ class LanguageModelService:
             structured_output_enforce_schema=structured_output_enforce_schema,
             chat_id=self._chat_id,
             assistant_id=self._assistant_id,
+            assistant_message_id=self._assistant_message_id,
         )
 
     @classmethod
@@ -344,6 +356,7 @@ class LanguageModelService:
             start_text=start_text,
             chat_id=self._chat_id,
             assistant_id=self._assistant_id,
+            assistant_message_id=self._assistant_message_id,
         )
 
     async def complete_with_references_async(
@@ -371,4 +384,5 @@ class LanguageModelService:
             start_text=start_text,
             chat_id=self._chat_id,
             assistant_id=self._assistant_id,
+            assistant_message_id=self._assistant_message_id,
         )