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: format Codex compression changes
Tejas Chopra03d12fc14039
diff --git a/headroom/proxy/handlers/openai.py b/headroom/proxy/handlers/openai.pyindex 70596acf6..faad122d3 100644--- a/headroom/proxy/handlers/openai.py+++ b/headroom/proxy/handlers/openai.py@@ -45,9 +45,7 @@ from headroom.proxy.auth_mode import classify_auth_modelogger = logging.getLogger("headroom.proxy")_CODEX_WS_UNIT_ROUTER_MAX_WORKERS = 10-_CODEX_WS_UNIT_ROUTER_SEMAPHORE = threading.BoundedSemaphore(- _CODEX_WS_UNIT_ROUTER_MAX_WORKERS-)+_CODEX_WS_UNIT_ROUTER_SEMAPHORE = threading.BoundedSemaphore(_CODEX_WS_UNIT_ROUTER_MAX_WORKERS)def _codex_ws_unit_worker_count(unit_count: int) -> int:@@ -731,9 +729,7 @@ class OpenAIHandlerMixin:+ elapsed_ms)timing_sink[f"compression_unit_router_category_{result.reason_category}"] = (- timing_sink.get(- f"compression_unit_router_category_{result.reason_category}", 0.0- )+ timing_sink.get(f"compression_unit_router_category_{result.reason_category}", 0.0)+ elapsed_ms)record_unit = getattr(getattr(self, "metrics", None), "record_codex_ws_unit", None)@@ -1001,9 +997,7 @@ class OpenAIHandlerMixin:output_serialization_started = time.perf_counter()output_bytes = json.dumps(working).encode("utf-8")_add_timing("compression_output_json_dump", output_serialization_started)- output_context_budget = (- _openai_responses_context_budget(working) if debug_enabled else None- )+ output_context_budget = _openai_responses_context_budget(working) if debug_enabled else None# One-line summary at INFO — the single event a human reading# logs should scan first to understand "what happened on this# pass". All the verbose per-event debug data stays available@@ -3319,10 +3313,9 @@ class OpenAIHandlerMixin:stage_timer.record("compression", ws_overhead_ms_total)def _record_ws_compression_timing(name: str, duration_ms: float) -> None:- ws_compression_timing_totals[name] = (- ws_compression_timing_totals.get(name, 0.0)- + max(0.0, float(duration_ms))- )+ ws_compression_timing_totals[name] = ws_compression_timing_totals.get(+ name, 0.0+ ) + max(0.0, float(duration_ms))def _codex_ws_final_strategies(timing: dict[str, float]) -> list[str]:prefix = "compression_unit_router_strategy_"@@ -3608,9 +3601,7 @@ class OpenAIHandlerMixin:"compression_executor_wait_run",_first_frame_compression_elapsed_ms,)- _record_ws_compression_overhead(- _first_frame_compression_elapsed_ms- )+ _record_ws_compression_overhead(_first_frame_compression_elapsed_ms)record_frame = getattr(getattr(self, "metrics", None), "record_codex_ws_frame", None)@@ -3623,9 +3614,7 @@ class OpenAIHandlerMixin:tokens_saved=_ws_saved,modified=_modified,strategy_chain=_codex_ws_strategy_chain(_ws_transforms),- final_strategies=_codex_ws_final_strategies(- _ws_compression_timing- ),+ final_strategies=_codex_ws_final_strategies(_ws_compression_timing),)if _modified:if isinstance(_new_inner, dict):@@ -3681,9 +3670,7 @@ class OpenAIHandlerMixin:if record_frame is not None:record_frame(elapsed_ms=_first_frame_compression_elapsed_ms,- bytes_before=len(- first_msg_raw.encode("utf-8", errors="replace")- ),+ bytes_before=len(first_msg_raw.encode("utf-8", errors="replace")),failed=True,)logger.warning(@@ -3859,9 +3846,7 @@ class OpenAIHandlerMixin:try:model_for_frame = inner_payload.get("model") or ""_frame_auth_mode = classify_auth_mode(ws_headers)- _preflight_ms = (- time.perf_counter() - _preflight_started- ) * 1000.0+ _preflight_ms = (time.perf_counter() - _preflight_started) * 1000.0_record_ws_compression_timing("compression_preflight_serialization",_preflight_ms,@@ -3884,9 +3869,10 @@ class OpenAIHandlerMixin:model=model_for_frame,request_id=request_id,)- for _timing_name, _timing_ms in (- frame_compression_timing.items()- ):+ for (+ _timing_name,+ _timing_ms,+ ) in frame_compression_timing.items():_record_ws_compression_timing(_timing_name, _timing_ms)finally:frame_compression_elapsed_ms = (@@ -3896,9 +3882,7 @@ class OpenAIHandlerMixin:"compression_executor_wait_run",frame_compression_elapsed_ms,)- _record_ws_compression_overhead(- frame_compression_elapsed_ms- )+ _record_ws_compression_overhead(frame_compression_elapsed_ms)record_frame = getattr(getattr(self, "metrics", None),"record_codex_ws_frame",@@ -3912,9 +3896,7 @@ class OpenAIHandlerMixin:attempted_tokens=frame_attempted_tokens,tokens_saved=frame_saved,modified=modified,- strategy_chain=_codex_ws_strategy_chain(- frame_transforms- ),+ strategy_chain=_codex_ws_strategy_chain(frame_transforms),final_strategies=_codex_ws_final_strategies(frame_compression_timing),diff --git a/headroom/proxy/server.py b/headroom/proxy/server.pyindex 2432290d0..5319feba6 100644--- a/headroom/proxy/server.py+++ b/headroom/proxy/server.py@@ -2118,9 +2118,7 @@ def create_app(config: ProxyConfig | None = None) -> FastAPI:"frames_compressed_total": m.codex_ws_frames_compressed_total,"frames_failed_total": m.codex_ws_frames_failed_total,"frames_to_kompress_total": m.codex_ws_frames_to_kompress_total,- "frames_kompress_attempted_total": (- m.codex_ws_frames_kompress_attempted_total- ),+ "frames_kompress_attempted_total": (m.codex_ws_frames_kompress_attempted_total),"frames_to_kompress_percent": _pct(m.codex_ws_frames_to_kompress_total,m.codex_ws_frames_attempted_total,diff --git a/tests/test_openai_codex_ws_lifecycle.py b/tests/test_openai_codex_ws_lifecycle.pyindex 5beb90ee7..712facc7e 100644--- a/tests/test_openai_codex_ws_lifecycle.py+++ b/tests/test_openai_codex_ws_lifecycle.py@@ -380,7 +380,9 @@ async def test_ws_session_metrics_include_dashboard_performance_timings():assert recorded["pipeline_timing"]["codex_ws.compression_preflight_serialization"] > 0assert recorded["pipeline_timing"]["codex_ws.compression_executor_wait_run"] > 0assert recorded["pipeline_timing"]["codex_ws.compression_live_unit_extraction"] == 2.0- assert recorded["pipeline_timing"]["codex_ws.compression_unit_router_strategy_passthrough"] == 3.0+ assert (+ recorded["pipeline_timing"]["codex_ws.compression_unit_router_strategy_passthrough"] == 3.0+ )@pytest.mark.asynciodiff --git a/tests/test_openai_codex_ws_timings.py b/tests/test_openai_codex_ws_timings.pyindex 8dfb1d099..f1362439a 100644--- a/tests/test_openai_codex_ws_timings.py+++ b/tests/test_openai_codex_ws_timings.py@@ -280,6 +280,7 @@ def test_codex_ws_request_id_and_session_id_present_in_log(stage_log_capture):assert isinstance(payload["session_id"], str)assert len(payload["session_id"]) >= 16+def test_codex_compression_debug_noop_skips_expensive_payload_debug(monkeypatch):handler = _DummyOpenAIHandler()