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.
Python agentaianthropicclaude-codecompressioncontext-engineeringcontext-windowcursorfastapilangchainllmmcpopenaiprompt-engineeringproxypythonragtoken-optimizationtokenstypescript
fix(proxy): give each Codex /v1/responses WS turn a unique request_id (#2164)
Rod Boevd02df1075894
diff --git a/headroom/proxy/handlers/openai.py b/headroom/proxy/handlers/openai.pyindex 4b096e070..e8e4b4e18 100644--- a/headroom/proxy/handlers/openai.py+++ b/headroom/proxy/handlers/openai.py@@ -7779,7 +7779,8 @@ class OpenAIHandlerMixin:# as HTTP turns.await self._record_request_outcome(RequestOutcome(- request_id=request_id,+ # Per-emission ids keep dashboard request-log keys unique.+ request_id=await self._next_request_id(),provider="openai",model=model_for_metrics,original_tokens=max(0, input_delta) + max(0, saved_delta),@@ -8354,7 +8355,8 @@ class OpenAIHandlerMixin:ws_messages_for_log.append({"role": "user", "content": ws_input_for_log})await self._record_request_outcome(RequestOutcome(- request_id=request_id,+ # Per-emission ids keep dashboard request-log keys unique.+ request_id=await self._next_request_id(),provider="openai",model=model_name,original_tokens=residual_input_tokens + residual_tokens_saved,diff --git a/tests/test_openai_codex_ws_lifecycle.py b/tests/test_openai_codex_ws_lifecycle.pyindex 481496b4e..b89a81a7e 100644--- a/tests/test_openai_codex_ws_lifecycle.py+++ b/tests/test_openai_codex_ws_lifecycle.py@@ -114,6 +114,14 @@ class _DummyOpenAIHandler(OpenAIHandlerMixin):await emit_request_outcome(self, outcome)+class _CapturingLogger:+ def __init__(self) -> None:+ self.entries = []++ def log(self, entry) -> None: # noqa: ANN001+ self.entries.append(entry)++class _FakeWebSocketDisconnect(Exception):"""Mirrors the ``WebSocketDisconnect`` type-name check in the handler.@@ -894,6 +902,138 @@ async def test_ws_session_metrics_include_dashboard_performance_timings():)+async def test_ws_multi_turn_request_ids_are_unique():+ upstream_events = [+ json.dumps({"type": "response.created", "response": {"id": "r_1"}}),+ json.dumps(+ {+ "type": "response.completed",+ "response": {+ "id": "r_1",+ "usage": {+ "input_tokens": 100,+ "input_tokens_details": {"cached_tokens": 75},+ "output_tokens": 12,+ },+ },+ }+ ),+ json.dumps({"type": "response.created", "response": {"id": "r_2"}}),+ json.dumps(+ {+ "type": "response.completed",+ "response": {+ "id": "r_2",+ "usage": {+ "input_tokens": 160,+ "input_tokens_details": {"cached_tokens": 120},+ "output_tokens": 20,+ },+ },+ }+ ),+ ]+ upstream = _FakeUpstream(upstream_events)+ fake_ws_mod = _make_fake_websockets_module(upstream)++ client_ws = _FakeWebSocket(frames=[_first_frame()])+ handler = _DummyOpenAIHandler()+ handler.logger = _CapturingLogger()++ counter = 0++ async def _next_request_id() -> str:+ nonlocal counter+ counter += 1+ return f"req-ws-{counter}"++ handler._next_request_id = _next_request_id # type: ignore[method-assign]++ with patch.dict(sys.modules, {"websockets": fake_ws_mod}):+ await handler.handle_openai_responses_ws(client_ws)++ logged = handler.logger.entries+ assert len(logged) == 2+ request_ids = [entry.request_id for entry in logged]+ assert len(set(request_ids)) == len(request_ids)+ assert [entry.input_tokens_optimized for entry in logged] == [100, 160]+ assert [entry.output_tokens for entry in logged] == [12, 20]+++async def test_ws_no_delta_turn_emits_no_extra_request_log():+ upstream_events = [+ json.dumps({"type": "response.created", "response": {"id": "r_1"}}),+ json.dumps({"type": "response.completed", "response": {"id": "r_1"}}),+ ]+ upstream = _FakeUpstream(upstream_events)+ fake_ws_mod = _make_fake_websockets_module(upstream)++ client_ws = _FakeWebSocket(frames=[_first_frame()])+ handler = _DummyOpenAIHandler()+ handler.logger = _CapturingLogger()++ counter = 0++ async def _next_request_id() -> str:+ nonlocal counter+ counter += 1+ return f"req-ws-{counter}"++ handler._next_request_id = _next_request_id # type: ignore[method-assign]++ with patch.dict(sys.modules, {"websockets": fake_ws_mod}):+ await handler.handle_openai_responses_ws(client_ws)++ assert len(handler.logger.entries) == 1+ assert all(entry.input_tokens_optimized == 0 for entry in handler.logger.entries)+ assert all(entry.output_tokens == 0 for entry in handler.logger.entries)+++async def test_ws_session_log_prefix_uses_session_id(caplog: pytest.LogCaptureFixture):+ upstream_events = [+ json.dumps({"type": "response.created", "response": {"id": "r_1"}}),+ json.dumps(+ {+ "type": "response.completed",+ "response": {+ "id": "r_1",+ "usage": {+ "input_tokens": 100,+ "input_tokens_details": {"cached_tokens": 75},+ "output_tokens": 12,+ },+ },+ }+ ),+ ]+ upstream = _FakeUpstream(upstream_events)+ fake_ws_mod = _make_fake_websockets_module(upstream)++ client_ws = _FakeWebSocket(frames=[_first_frame()])+ handler = _DummyOpenAIHandler()+ handler.logger = _CapturingLogger()++ counter = 0++ async def _next_request_id() -> str:+ nonlocal counter+ counter += 1+ return f"req-ws-{counter}"++ handler._next_request_id = _next_request_id # type: ignore[method-assign]+ caplog.set_level(logging.INFO, logger="headroom.proxy")++ with patch.dict(sys.modules, {"websockets": fake_ws_mod}):+ await handler.handle_openai_responses_ws(client_ws)++ assert handler.logger.entries+ assert handler.logger.entries[0].request_id != "req-ws-1"+ assert "[req-ws-1] PERF" in caplog.text++@pytest.mark.asyncioasync def test_ws_opt_in_flattens_response_create_for_openai_compatible_upstream(monkeypatch):"""Some OpenAI-compatible WS gateways expect top-level response.create payloads."""