fix: stop replaying OpenAI stream text

Avoid replaying buffered text at the end of OpenAI streams so text-only responses are emitted once while forced tool fallback behavior stays intact.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

app/main.py

@@ -650,9 +650,7 @@ async def v1_chat_completions(req: ChatCompletionsRequest, request: Request):
                             buffered_text_parts.clear()
                             yield f"data: {json.dumps(payload, ensure_ascii=False)}\n\n"
 
-                    if buffered_text_parts:
+                    if buffered_text_parts and forced_tool_name and saw_tool_call:
-                        for buffered_text in buffered_text_parts:
-                            yield _text_payload(buffered_text)
                         buffered_text_parts.clear()
 
                     done_payload = {

tests/test_tool_call_bridge.py

@@ -414,12 +414,38 @@ class ToolCallBridgeTests(unittest.IsolatedAsyncioTestCase):
             body = await _collect_stream(response)
 
         self.assertIn('"tool_calls"', body)
-        self.assertIn('"content": "hello"', body)
+        self.assertEqual(body.count('"content": "hello"'), 1)
         self.assertIn('"finish_reason": "tool_calls"', body)
         self.assertIn('"usage"', body)
         self.assertIn("data: [DONE]", body)
 
 
+    async def test_openai_stream_emits_text_delta_only_once_without_tools(self) -> None:
+        fake_client = _FakeClient(
+            stream_events=[
+                {"type": "text", "text": "你好"},
+            ],
+            complete_result={},
+        )
+        req = ChatCompletionsRequest(
+            model="org_auto",
+            messages=[{"role": "user", "content": "hi"}],
+            stream=True,
+        )
+
+        with (
+            patch.object(main, "pool", _FakePool(_FakeInstance(fake_client))),
+            patch.object(main, "chat_guard", _FakeGuard()),
+            patch.object(main, "_ensure_instance_logged_in", AsyncMock(return_value={"id": "u"})),
+            patch.object(main.stats_collector, "record_chat", AsyncMock(return_value=None)),
+        ):
+            response = await main.v1_chat_completions(req, _make_request("/v1/chat/completions"))
+            body = await _collect_stream(response)
+
+        self.assertEqual(body.count('"content": "你好"'), 1)
+        self.assertIn('"finish_reason": "stop"', body)
+        self.assertIn("data: [DONE]", body)
+
     async def test_openai_stream_filters_tool_events_by_allowlist(self) -> None:
         fake_client = _FakeClient(
             stream_events=[