unique-toolkit 2026.40.0.dev0__tar.gz → 2026.40.0.dev2__tar.gz

{unique_toolkit-2026.40.0.dev0 → unique_toolkit-2026.40.0.dev2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: unique-toolkit
-Version: 2026.40.0.dev0
+Version: 2026.40.0.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-2026.40.0.dev0 → unique_toolkit-2026.40.0.dev2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "unique_toolkit"
-version = "2026.40.0.dev0"
+version = "2026.40.0.dev2"
 description = ""
 readme = "README.md"
 requires-python = ">=3.12"

{unique_toolkit-2026.40.0.dev0 → unique_toolkit-2026.40.0.dev2}/unique_toolkit/agentic/evaluation/hallucination/utils.py

@@ -203,7 +203,8 @@ def _compose_msgs(
     )
     system_msg = LanguageModelSystemMessage(content=system_msg_content)
 
-    # Render user message
+    # Render user message. Pass history text only when history exists so the
+    # template can omit an empty conversation section.
     user_msg_content = render_template(
         user_template,
         input_text=input.input_text,
@@ -211,7 +212,7 @@ def _compose_msgs(
         if has_context
         else None,
         history_messages_text=input.get_joined_history_texts(tag_name="conversation")
-        if has_context
+        if has_context and input.history_messages
         else None,
         output_text=input.output_text,
     )

{unique_toolkit-2026.40.0.dev0 → unique_toolkit-2026.40.0.dev2}/unique_toolkit/agentic/evaluation/tests/test_hallucination_utils.py

@@ -1224,3 +1224,67 @@ def test_get_msgs__ignores_context_images__when_model_lacks_vision() -> None:
     # Assert
     assert result.root[0].content == no_context_result.root[0].content
     assert isinstance(result.root[1].content, str)
+
+
+@pytest.mark.unit
+def test_compose_msgs__omits_conversation_section__when_history_is_empty():
+    """A single-turn evaluation must not be shown an empty conversation.
+
+    The template guards the section with `{% if history_messages_text %}`,
+    and the getter returns a truthy placeholder when there is no history, so
+    the guard used to never fire: every caller evaluating one turn shipped a
+    `Conversation:` section whose body said there was no conversation, while
+    the system prompt announced one.
+    """
+    msgs = _compose_msgs(
+        EvaluationMetricInput(
+            input_text="What is the policy rate?",
+            context_texts=["The policy rate is 0.25%."],
+            history_messages=[],
+            output_text="0.25%.",
+        ),
+        HallucinationConfig(),
+        has_context=True,
+    )
+
+    user_content = msgs.root[-1].content
+    assert isinstance(user_content, str)
+    assert "Conversation:" not in user_content
+    assert "<No conversation texts provided>" not in user_content
+    # The references the model does have must still be there.
+    assert "The policy rate is 0.25%." in user_content
+
+
+@pytest.mark.unit
+def test_compose_msgs__keeps_conversation_section__when_history_is_present():
+    msgs = _compose_msgs(
+        EvaluationMetricInput(
+            input_text="And in March?",
+            context_texts=["The March rate was 0.50%."],
+            history_messages=[
+                ChatMessage(
+                    id="msg_1",
+                    chat_id="chat_1",
+                    previous_message_id=None,
+                    role=LanguageModelMessageRole.USER,
+                    text="What is the policy rate?",
+                ),
+                ChatMessage(
+                    id="msg_2",
+                    chat_id="chat_1",
+                    previous_message_id="msg_1",
+                    role=LanguageModelMessageRole.ASSISTANT,
+                    text="0.25%.",
+                ),
+            ],
+            output_text="0.50%.",
+        ),
+        HallucinationConfig(),
+        has_context=True,
+    )
+
+    user_content = msgs.root[-1].content
+    assert isinstance(user_content, str)
+    assert "Conversation:" in user_content
+    assert "What is the policy rate?" in user_content
+    assert "<No conversation texts provided>" not in user_content

{unique_toolkit-2026.40.0.dev0 → unique_toolkit-2026.40.0.dev2}/unique_toolkit/agentic/tools/openai_builtin/code_interpreter/postprocessors/code_display.py

@@ -37,7 +37,7 @@ class ShowExecutedCodePostprocessorConfig(BaseModel):
     model_config = get_configuration_dict()
     enable: bool = Field(
         default=True,
-        description="Enable display of executed code before the assistant message",
+        description="Show the executed source code to the user",
     )
     remove_from_history: SkipJsonSchema[bool] = (
         Field(  # At the moment, it's not possible to keep executed code in the history