Merge 4d27bd4a3b into 05d8f11085

tests/gateway/test_empty_response_handling.py

@@ -0,0 +1,47 @@
+"""Tests for empty-response fallback handling in GatewayRunner."""
+
+import sys
+import types
+
+import pytest
+
+
+@pytest.fixture(autouse=True)
+def _mock_dotenv(monkeypatch):
+    """gateway.run imports dotenv at module level; stub it for tests."""
+    fake = types.ModuleType("dotenv")
+    fake.load_dotenv = lambda *a, **kw: None
+    monkeypatch.setitem(sys.modules, "dotenv", fake)
+
+
+@pytest.fixture()
+def runner():
+    from gateway.run import GatewayRunner
+
+    return GatewayRunner.__new__(GatewayRunner)
+
+
+class TestEmptyResponseFallback:
+    def test_reasoning_only_message(self, runner):
+        message = runner._build_empty_response_message({
+            "empty_response_reasoning": "structured reasoning answer",
+        })
+        assert message == (
+            "⚠️ The model produced internal reasoning but no visible response after all retries. "
+            "Try again or rephrase your question."
+        )
+
+    def test_truly_empty_message(self, runner):
+        message = runner._build_empty_response_message({})
+        assert message == (
+            "⚠️ The model returned no content after all retries. "
+            "Try again or rephrase your question."
+        )
+
+    def test_response_fallback_detection_handles_new_and_legacy_forms(self, runner):
+        assert runner._is_empty_response_fallback(
+            {"response_is_empty_fallback": True},
+            "⚠️ The model returned no content after all retries. Try again or rephrase your question.",
+        )
+        assert runner._is_empty_response_fallback({}, "(empty)")
+        assert not runner._is_empty_response_fallback({}, "hello world")

tests/run_agent/test_run_agent.py

@@ -250,6 +250,15 @@ def _mock_response(
         resp.usage = None
     return resp
 
+EMPTY_REASONING_RESPONSE = (
+    "⚠️ The model produced internal reasoning but no visible response after all retries. "
+    "Try again or rephrase your question."
+)
+EMPTY_TRULY_EMPTY_RESPONSE = (
+    "⚠️ The model returned no content after all retries. "
+    "Try again or rephrase your question."
+)
+
 
 # ===================================================================
 # Group 1: Pure Functions
@@ -2256,7 +2265,9 @@ class TestRunConversation:
 
         mock_compress.assert_not_called()  # no compression triggered
         assert result["completed"] is True
-        assert result["final_response"] == "(empty)"
+        assert result["final_response"] == EMPTY_TRULY_EMPTY_RESPONSE
+        assert result["response_is_empty_fallback"] is True
+        assert result["empty_response_reasoning"] is None
         assert result["api_calls"] == 6  # 1 original + 2 prefill + 3 retries
 
     def test_reasoning_only_response_prefill_then_empty(self, agent):
@@ -2276,7 +2287,9 @@ class TestRunConversation:
         ):
             result = agent.run_conversation("answer me")
         assert result["completed"] is True
-        assert result["final_response"] == "(empty)"
+        assert result["final_response"] == EMPTY_REASONING_RESPONSE
+        assert result["response_is_empty_fallback"] is True
+        assert result["empty_response_reasoning"] == "structured reasoning answer"
         assert result["api_calls"] == 6  # 1 original + 2 prefill + 3 retries
 
     def test_reasoning_only_prefill_succeeds_on_continuation(self, agent):
@@ -2323,7 +2336,9 @@ class TestRunConversation:
         ):
             result = agent.run_conversation("answer me")
         assert result["completed"] is True
-        assert result["final_response"] == "(empty)"
+        assert result["final_response"] == EMPTY_TRULY_EMPTY_RESPONSE
+        assert result["response_is_empty_fallback"] is True
+        assert result["empty_response_reasoning"] is None
         assert result["api_calls"] == 4  # 1 original + 3 retries
 
     def test_truly_empty_response_succeeds_on_nudge(self, agent):
@@ -2419,7 +2434,8 @@ class TestRunConversation:
         ):
             result = agent.run_conversation("answer me")
         assert result["completed"] is True
-        assert result["final_response"] == "(empty)"
+        assert result["final_response"] == EMPTY_TRULY_EMPTY_RESPONSE
+        assert result["response_is_empty_fallback"] is True
 
     def test_empty_response_emits_status_for_gateway(self, agent):
         """_emit_status is called during empty retries so gateway users see feedback."""
@@ -2445,7 +2461,8 @@ class TestRunConversation:
         ):
             result = agent.run_conversation("answer me")
 
-        assert result["final_response"] == "(empty)"
+        assert result["final_response"] == EMPTY_TRULY_EMPTY_RESPONSE
+        assert result["response_is_empty_fallback"] is True
         # Should have emitted retry statuses (3 retries) + final failure
         retry_msgs = [m for m in status_messages if "retrying" in m.lower()]
         assert len(retry_msgs) == 3, f"Expected 3 retry status messages, got {len(retry_msgs)}: {status_messages}"