Address CI review: incomplete status handling, pricing, help text

P1: Treat status='incomplete' as a hard error even when content exists,
since truncated reviews may silently suppress findings. Print
incomplete_details and suggest remediation (reduce diff, lower context).
P2: Fix gpt-5.4-pro pricing from (15,60) to (30,180) per official rates.
P3: Fix argparse description to say "Responses API" not "Chat Completions".

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: igerber

.claude/scripts/openai_review.py

@@ -854,7 +854,7 @@ def apply_token_budget(
 # MAINTENANCE: Update when OpenAI changes pricing.
 PRICING = {
     "gpt-5.4": (2.50, 15.00),
-    "gpt-5.4-pro": (15.00, 60.00),
+    "gpt-5.4-pro": (30.00, 180.00),
     "gpt-4.1": (2.00, 8.00),
     "gpt-4.1-mini": (0.40, 1.60),
     "o3": (2.00, 8.00),
@@ -1209,6 +1209,24 @@ def call_openai(
 
     content = _extract_response_text(result)
 
+    # Treat truncated responses as errors — partial reviews may suppress findings.
+    status = result.get("status")
+    if content.strip() and status == "incomplete":
+        detail = result.get("incomplete_details") or ""
+        print(
+            "Error: Review was truncated (status='incomplete'). "
+            "Output may be missing findings.",
+            file=sys.stderr,
+        )
+        if detail:
+            print(f"Detail: {detail}", file=sys.stderr)
+        print(
+            "Try reducing diff size, disabling --full-registry, or "
+            "lowering --context to 'minimal'.",
+            file=sys.stderr,
+        )
+        sys.exit(1)
+
     if not content.strip():
         # No usable content — report the best diagnostic we have.
         status = result.get("status", "<missing>")
@@ -1244,7 +1262,7 @@ def _read_file(path: str, label: str) -> str:
 
 def main() -> None:
     parser = argparse.ArgumentParser(
-        description="Run local AI code review via OpenAI Chat Completions API."
+        description="Run local AI code review via OpenAI Responses API."
     )
     parser.add_argument(
         "--review-criteria",

tests/test_openai_review.py

@@ -1710,19 +1710,37 @@ def test_status_none_with_valid_output_succeeds(self, review_mod, mock_urlopen):
         content, _ = review_mod.call_openai("test", "gpt-5.4", "fake-key")
         assert content == "Good review."
 
-    def test_incomplete_status_with_valid_content_succeeds(self, review_mod, mock_urlopen):
-        """Non-completed status should still return content when output is usable."""
+    def test_incomplete_status_with_content_exits(self, review_mod, mock_urlopen):
+        """Truncated response (status=incomplete) should exit even if content exists."""
         mock_urlopen["response_data"] = {
             "status": "incomplete",
             "output_text": None,
             "output": [{
                 "type": "message",
-                "content": [{"type": "output_text", "text": "Partial but usable."}],
+                "content": [{"type": "output_text", "text": "Partial review."}],
             }],
             "usage": {"input_tokens": 10, "output_tokens": 5},
         }
-        content, _ = review_mod.call_openai("test", "gpt-5.4", "fake-key")
-        assert content == "Partial but usable."
+        with pytest.raises(SystemExit):
+            review_mod.call_openai("test", "gpt-5.4", "fake-key")
+
+    def test_incomplete_status_surfaces_details(self, review_mod, mock_urlopen, capsys):
+        """Incomplete response should print incomplete_details to stderr."""
+        mock_urlopen["response_data"] = {
+            "status": "incomplete",
+            "incomplete_details": {"reason": "max_output_tokens"},
+            "output_text": None,
+            "output": [{
+                "type": "message",
+                "content": [{"type": "output_text", "text": "Partial."}],
+            }],
+            "usage": {"input_tokens": 10, "output_tokens": 5},
+        }
+        with pytest.raises(SystemExit):
+            review_mod.call_openai("test", "gpt-5.4", "fake-key")
+        captured = capsys.readouterr()
+        assert "truncated" in captured.err.lower()
+        assert "max_output_tokens" in captured.err
 
     def test_output_text_convenience_field_used(self, review_mod, mock_urlopen):
         """When output_text is populated (SDK-style), use it directly."""