unique_toolkit 1.75.0__tar.gz → 1.75.1__tar.gz

{unique_toolkit-1.75.0 → unique_toolkit-1.75.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: unique-toolkit
-Version: 1.75.0
+Version: 1.75.1
 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-1.75.0 → unique_toolkit-1.75.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "unique_toolkit"
-version = "1.75.0"
+version = "1.75.1"
 description = ""
 readme = "README.md"
 requires-python = ">=3.12"

{unique_toolkit-1.75.0 → unique_toolkit-1.75.1}/unique_toolkit/agentic/loop_runner/middleware/planning/planning.py

@@ -7,6 +7,7 @@ from openai.types.responses import (
     Response,
     ResponseFormatTextJSONSchemaConfigParam,
     ResponseInputItemParam,
+    ResponseOutputMessage,
     ResponseTextConfigParam,
 )
 from pydantic import BaseModel, Field
@@ -39,6 +40,26 @@ from unique_toolkit.language_model.schemas import (
 _LOGGER = logging.getLogger(__name__)
 
 
+def _get_first_output_text(response: Response) -> str | None:
+    """
+    The Responses API can return multiple output items when doing structured output
+    (https://community.openai.com/t/multiple-outputs-from-responses-api/1291026),
+    We should tell the model not to do this in the schema, as well as extract only the first one
+    """
+    if len(response.output) > 1:
+        _LOGGER.warning(
+            "Model returned %i outputs, keeping only the first one",
+            len(response.output),
+        )
+
+    for item in response.output:
+        if isinstance(item, ResponseOutputMessage):
+            for content in item.content:
+                if content.type == "output_text" and content.text:
+                    return content.text
+    return None
+
+
 class PlanningConfig(BaseModel):
     model_config = get_configuration_dict()
 
@@ -128,7 +149,7 @@ class ResponsesPlanningMiddleware(ResponsesLoopIterationRunner):
         self,
         openai_messages: str | list[ResponseInputItemParam],
         model_name: str,
-    ) -> Response | None:
+    ) -> str | None:
         planning_schema = get_planning_schema(self._config.planning_schema_config)
 
         response = await self._openai_client.responses.create(
@@ -148,29 +169,30 @@ class ResponsesPlanningMiddleware(ResponsesLoopIterationRunner):
             ),
         )
 
-        if not response.output_text:
+        output_text = _get_first_output_text(response)
+        if not output_text:
             _LOGGER.info("Empty planning response")
             return None
 
-        return response
+        return output_text
 
     async def __call__(
         self, **kwargs: Unpack[_ResponsesLoopIterationRunnerKwargs]
     ) -> ResponsesLanguageModelStreamResponse:
         openai_messages = convert_messages_to_openai(kwargs["messages"])
 
-        response = await self._run_plan_step(
+        output_text = await self._run_plan_step(
             openai_messages=openai_messages,
             model_name=kwargs["model"].name,
         )
 
-        if response is None:
+        if output_text is None:
             _LOGGER.info(
                 "Error executing planning step, proceeding without planning step"
             )
             return await self._loop_runner(**kwargs)
 
-        assistant_message = LanguageModelAssistantMessage(content=response.output_text)
+        assistant_message = LanguageModelAssistantMessage(content=output_text)
 
         if self._history_manager is not None:
             self._history_manager.add_assistant_message(assistant_message)