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.

fix(proxy): log response_content_length in proxy_inbound_response (#2701)

im47cncfa479af8e55
diff --git a/headroom/proxy/server.py b/headroom/proxy/server.py
index d5b758b0e..ca82ea3ad 100644
--- a/headroom/proxy/server.py
+++ b/headroom/proxy/server.py
@@ -3919,12 +3919,14 @@ def create_app(config: ProxyConfig | None = None) -> FastAPI:
except Exception:
logger.debug("record_inbound_response failed", exc_info=True)
logger.info(
- "event=proxy_inbound_response id=%s method=%s path=%s status=%s duration_ms=%.2f",
+ "event=proxy_inbound_response id=%s method=%s path=%s status=%s "
+ "duration_ms=%.2f response_content_length=%s",
inbound_id,
method,
path,
response.status_code,
(time.perf_counter() - started) * 1000.0,
+ response.headers.get("content-length", ""),
)
return response
diff --git a/tests/test_proxy/test_inbound_response_size.py b/tests/test_proxy/test_inbound_response_size.py
new file mode 100644
index 000000000..9e58e9edd
--- /dev/null
+++ b/tests/test_proxy/test_inbound_response_size.py
@@ -0,0 +1,116 @@
+"""Tests that ``proxy_inbound_response`` logs ``response_content_length``.
+
+The inbound response log line previously recorded only timing (``duration_ms``)
+and status, making it impossible to correlate slow responses with payload size.
+The fix adds ``response_content_length`` read from the Starlette response's
+``Content-Length`` header — the same approach the inbound request line uses for
+``content_length``.
+
+Contract pinned here:
+- response with Content-Length → value appears in the log line
+- response without Content-Length (streaming/chunked) → empty string, no crash
+"""
+
+from __future__ import annotations
+
+import logging
+import re
+
+import pytest
+
+fastapi = pytest.importorskip("fastapi")
+
+from fastapi.responses import JSONResponse # noqa: E402
+from fastapi.testclient import TestClient # noqa: E402
+
+from headroom.proxy.server import ProxyConfig, create_app # noqa: E402
+
+
+def _make_config(**overrides) -> ProxyConfig:
+ base = {
+ "optimize": False,
+ "cache_enabled": False,
+ "rate_limit_enabled": False,
+ "mode": "token",
+ }
+ base.update(overrides)
+ return ProxyConfig(**base)
+
+
+_RESPONSE_LOG_RE = re.compile(r"event=proxy_inbound_response\b")
+
+
+def _get_inbound_response_logs(records: list[logging.LogRecord]) -> list[str]:
+ return [r.getMessage() for r in records if _RESPONSE_LOG_RE.search(r.getMessage())]
+
+
+def capture_logs():
+ """Capture headroom.proxy log records via a dedicated handler."""
+ records: list[logging.LogRecord] = []
+
+ class _Handler(logging.Handler):
+ def emit(self, record):
+ records.append(record)
+
+ handler = _Handler()
+ handler.setLevel(logging.DEBUG)
+ lg = logging.getLogger("headroom.proxy")
+ lg.addHandler(handler)
+ lg.setLevel(logging.DEBUG)
+ try:
+ yield records
+ finally:
+ lg.removeHandler(handler)
+
+
+async def _json_handler(*a, **kw):
+ return JSONResponse({"ok": True, "data": "x" * 200})
+
+
+async def _stream_handler(*a, **kw):
+ from starlette.responses import StreamingResponse
+
+ async def _stream():
+ yield b'{"ok": true}'
+ yield b""
+
+ return StreamingResponse(_stream())
+
+
+def test_response_content_length_logged(capture_logs):
+ """A JSON response sets Content-Length; the value must appear in the log."""
+ app = create_app(_make_config())
+ with TestClient(app) as client:
+ client.app.state.proxy.handle_anthropic_messages = _json_handler
+ resp = client.post("/v1/messages", json={"model": "glm-5.2", "messages": []})
+ assert resp.status_code == 200
+
+ lines = _get_inbound_response_logs(capture_logs)
+ assert lines, "expected at least one proxy_inbound_response log line"
+ line = lines[-1]
+ assert "response_content_length=" in line
+ m = re.search(r"response_content_length=(\d+)", line)
+ assert m, f"expected numeric content_length in: {line}"
+ assert int(m.group(1)) > 0
+
+
+def test_streaming_response_no_content_length(capture_logs):
+ """A streaming response has no Content-Length header.
+
+ The log line must still appear with an empty ``response_content_length=``
+ value — not crash or omit the field.
+ """
+ app = create_app(_make_config())
+ with TestClient(app) as client:
+ client.app.state.proxy.handle_anthropic_messages = _stream_handler
+ resp = client.post("/v1/messages", json={"model": "glm-5.2", "messages": []})
+ assert resp.status_code == 200
+
+ lines = _get_inbound_response_logs(capture_logs)
+ assert lines, "expected at least one proxy_inbound_response log line"
+ line = lines[-1]
+ assert "response_content_length=" in line
+ m = re.search(r"response_content_length=(\S*)", line)
+ assert m, f"expected response_content_length field in: {line}"
+ assert m.group(1) in ("", '""'), f"expected empty for streaming, got: {m.group(1)}"