Skip to content

Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.

test(litellm): cover standalone max_completion_tokens forwarding (#3558)

AxelRayc32b315537fe
diff --git a/tests/test_litellm_openai_passthrough.py b/tests/test_litellm_openai_passthrough.py
index 1604102ad..79ea524a2 100644
--- a/tests/test_litellm_openai_passthrough.py
+++ b/tests/test_litellm_openai_passthrough.py
@@ -145,6 +145,42 @@ async def test_standard_params_still_forwarded() -> None:
assert kwargs["extra_body"] == {"chat_template_kwargs": {"enable_thinking": False}}
+async def test_max_completion_tokens_forwarded_as_standard_param() -> None:
+ backend = make_backend()
+ body = request_body(max_completion_tokens=64)
+ body.pop("max_tokens")
+
+ with patch("headroom.backends.litellm.acompletion", new_callable=AsyncMock) as mock_acomp:
+ mock_acomp.return_value = make_response()
+
+ await backend.send_openai_message(body, {})
+
+ kwargs = mock_acomp.await_args.kwargs
+ assert kwargs["max_completion_tokens"] == 64
+ assert "extra_body" not in kwargs
+
+
+async def test_max_completion_tokens_forwarded_as_standard_param_streaming() -> None:
+ backend = make_backend()
+ body = request_body(max_completion_tokens=64)
+ body.pop("max_tokens")
+ stream = FakeAsyncStream(
+ [SimpleNamespace(model_dump=lambda **kwargs: {"id": "chunk1", "choices": []})]
+ )
+
+ with patch("headroom.backends.litellm.acompletion", new_callable=AsyncMock) as mock_acomp:
+ mock_acomp.return_value = stream
+
+ chunks = [chunk async for chunk in backend.stream_openai_message(body, {})]
+
+ kwargs = mock_acomp.await_args.kwargs
+ assert kwargs["max_completion_tokens"] == 64
+ assert "extra_body" not in kwargs
+ assert chunks[-1] == "data: [DONE]\n\n"
+
+
@pytest.mark.asyncio
async def test_reasoning_params_forwarded_top_level_not_extra_body() -> None:
# reasoning_effort / max_completion_tokens are litellm-mapped params: they