Fix AI review and comparison runtime

templates/base.html

@@ -10,7 +10,7 @@
     <link href="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0/dist/css/bootstrap.min.css" rel="stylesheet">
     <link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/bootstrap-icons@1.10.0/font/bootstrap-icons.css">
     <link rel="stylesheet" href="{{ url_for('static', filename='css/app.css') }}">
-    <link rel="stylesheet" href="{{ url_for('static', filename='css/chatbot.css') }}">
+    <link rel="stylesheet" href="{{ url_for('static', filename='css/chatbot.css', v='20260725.1') }}">
     <link rel="stylesheet" href="{{ url_for('static', filename='css/page_search.css') }}">
     {% block extra_css %}{% endblock %}
 </head>
@@ -184,7 +184,7 @@ <h3><i class="bi bi-robot me-2"></i>Statistical assistant</h3>
     </button>
 
     <script src="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0/dist/js/bootstrap.bundle.min.js"></script>
-    <script src="{{ url_for('static', filename='js/chatbot.js') }}"></script>
+    <script src="{{ url_for('static', filename='js/chatbot.js', v='20260725.1') }}"></script>
     <script>
     document.addEventListener('DOMContentLoaded', function () {
         document.querySelectorAll('[data-bs-toggle="tooltip"]').forEach(function (element) {

templates/blank.html

@@ -15,7 +15,7 @@
     <link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/bootstrap-icons@1.10.0/font/bootstrap-icons.css">
 
     <!-- Chatbot CSS -->
-    <link rel="stylesheet" href="{{ url_for('static', filename='css/chatbot.css') }}">
+    <link rel="stylesheet" href="{{ url_for('static', filename='css/chatbot.css', v='20260725.1') }}">
 
     <!-- Page Search CSS -->
     <link rel="stylesheet" href="{{ url_for('static', filename='css/page_search.css') }}">
@@ -92,11 +92,11 @@ <h3><i class="bi bi-robot me-2"></i>AI Assistant</h3>
     <script src="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0/dist/js/bootstrap.bundle.min.js"></script>
 
     <!-- Chatbot JS -->
-    <script src="{{ url_for('static', filename='js/chatbot.js') }}"></script>
+    <script src="{{ url_for('static', filename='js/chatbot.js', v='20260725.1') }}"></script>
 
     <!-- Page Search JS -->
     <script src="{{ url_for('static', filename='js/page_search.js') }}"></script>
 
     {% block scripts %}{% endblock %}
 </body>
-</html> 
+</html>

tests/test_integrations.py

@@ -76,6 +76,47 @@ def test_openai_supports_strict_structured_outputs(monkeypatch):
     }
 
 
+def test_openai_allows_a_larger_feature_specific_output_budget(monkeypatch):
+    monkeypatch.setenv("AI_ENHANCEMENT_ENABLED", "true")
+    monkeypatch.setenv("OPENAI_API_KEY", "sk-test")
+    monkeypatch.setenv("AI_MAX_OUTPUT_TOKENS", "400")
+    response = Mock(output_text='{"answer":"complete"}', status="completed")
+    client = Mock()
+    client.responses.create.return_value = response
+
+    with patch("utils.ai_service.OpenAI", return_value=client):
+        result = call_openai_api(
+            "Return the complete review.",
+            max_output_tokens=1_500,
+        )
+
+    assert result == '{"answer":"complete"}'
+    assert (
+        client.responses.create.call_args.kwargs["max_output_tokens"]
+        == 1_500
+    )
+
+
+def test_openai_rejects_an_incomplete_provider_response(monkeypatch):
+    monkeypatch.setenv("AI_ENHANCEMENT_ENABLED", "true")
+    monkeypatch.setenv("OPENAI_API_KEY", "sk-test")
+    response = Mock(
+        output_text='{"answer":"cut off',
+        status="incomplete",
+    )
+    client = Mock()
+    client.responses.create.return_value = response
+
+    with patch("utils.ai_service.OpenAI", return_value=client):
+        with pytest.raises(
+            OpenAIServiceError,
+            match="exceeded its output limit",
+        ) as error:
+            call_openai_api("Return the complete review.")
+
+    assert error.value.status_code == 502
+
+
 def test_recommendation_ai_is_limited_to_verified_candidates():
     model_database = {
         "Linear Regression": {
@@ -128,6 +169,7 @@ def test_recommendation_ai_is_limited_to_verified_candidates():
         }
     ]
     schema = generate.call_args.kwargs["response_schema"]
+    assert generate.call_args.kwargs["max_output_tokens"] == 1_500
     assert schema["properties"]["recommended_model"]["enum"] == [
         "Linear Regression",
         "Random Forest",

tests/test_main_routes.py

@@ -11,6 +11,8 @@ def test_home_page(self, client):
         response = client.get('/')
         assert response.status_code == 200
         assert b'statistical' in response.data.lower() or b'model' in response.data.lower()
+        assert b'chatbot.js?v=20260725.1' in response.data
+        assert b'chatbot.css?v=20260725.1' in response.data
     def test_analysis_form_page(self, client):
         """Test that the analysis form page loads."""
         response = client.get('/analysis-form')

utils/ai_service.py

@@ -55,11 +55,15 @@ def _timeout_seconds() -> float:
         return 45.0
 
 
-def _max_output_tokens() -> int:
-    raw_limit = os.environ.get("AI_MAX_OUTPUT_TOKENS", "400")
+def _max_output_tokens(override: Optional[int] = None) -> int:
+    raw_limit = (
+        override
+        if override is not None
+        else os.environ.get("AI_MAX_OUTPUT_TOKENS", "400")
+    )
     try:
         return min(max(int(raw_limit), 100), 1_500)
-    except ValueError:
+    except (TypeError, ValueError):
         return 400
 
 
@@ -75,6 +79,7 @@ def call_openai_api(
     safety_identifier: Optional[str] = None,
     response_schema: Optional[dict[str, Any]] = None,
     schema_name: str = "structured_response",
+    max_output_tokens: Optional[int] = None,
 ) -> str:
     """Generate text through OpenAI's Responses API."""
     if not is_ai_enabled():
@@ -100,7 +105,7 @@ def call_openai_api(
         "model": target_model,
         "instructions": system_prompt or DEFAULT_SYSTEM_PROMPT,
         "input": cleaned_prompt,
-        "max_output_tokens": _max_output_tokens(),
+        "max_output_tokens": _max_output_tokens(max_output_tokens),
         "reasoning": {"effort": _reasoning_effort()},
         "store": False,
     }
@@ -169,6 +174,11 @@ def call_openai_api(
 
     if response is None:
         raise OpenAIServiceError("The AI provider returned no response.", 502)
+    if getattr(response, "status", None) == "incomplete":
+        raise OpenAIServiceError(
+            "The AI provider response exceeded its output limit.",
+            502,
+        )
     content = response.output_text
     if not isinstance(content, str) or not content.strip():
         raise OpenAIServiceError("The AI provider returned an empty response.", 502)

utils/recommendation_ai.py

@@ -138,6 +138,7 @@ def review_recommendation(
         safety_identifier=safety_identifier,
         response_schema=_review_schema(verified_candidates),
         schema_name="model_recommendation_review",
+        max_output_tokens=1_500,
     )
 
     try: