fix: ensure responses stream always completes

Emit a fallback response.completed and [DONE] when upstream SSE closes early so OpenAI /v1/responses clients do not fail on incomplete streams.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
GitHub Actions
2026-04-20 13:23:43 +08:00
parent c9bd71f727
commit c08dea89a2
2 changed files with 36 additions and 0 deletions

View File

@@ -1193,6 +1193,21 @@ async def _responses_stream_from_chat_stream(
}
)
yield _sse_data(
{
"type": "response.completed",
"response": {
"id": response_id,
"object": "response",
"created_at": created_at,
"status": "completed",
"model": model,
"usage": usage,
},
}
)
yield "data: [DONE]\\n\\n"
@app.post("/v1/responses", dependencies=[Depends(auth_guard)])

View File

@@ -805,6 +805,27 @@ class ToolCallBridgeTests(unittest.IsolatedAsyncioTestCase):
self.assertIn('"output_tokens": 2', body)
self.assertIn('data: [DONE]', body)
async def test_responses_stream_emits_completed_when_upstream_closes_without_done(self) -> None:
async def _chat_sse_without_done():
yield b'data: {"choices": [{"delta": {"content": "partial"}}]}\n\n'
yield b'data: {"usage": {"prompt_tokens": 7, "completion_tokens": 1, "total_tokens": 8}, "choices": [{"delta": {}}]}\n\n'
req = ResponsesRequest(model="org_auto", input="hi", stream=True)
mock_chat = AsyncMock(
return_value=StreamingResponse(_chat_sse_without_done(), media_type="text/event-stream")
)
with patch.object(main, "v1_chat_completions", mock_chat):
response = await main.v1_responses(req, _make_request("/v1/responses"))
body = await _collect_stream(response)
self.assertIn('"type": "response.output_text.delta"', body)
self.assertIn('"delta": "partial"', body)
self.assertIn('"type": "response.completed"', body)
self.assertIn('"input_tokens": 7', body)
self.assertIn('"output_tokens": 1', body)
self.assertIn('data: [DONE]', body)
async def test_responses_non_stream_returns_502_on_invalid_upstream_json(self) -> None:
req = ResponsesRequest(model="org_auto", input="hi", stream=False)