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
refactor(proxy): remove dead aliases and wrappers from helpers, body_forwarding, forwarded_headers (#4080)
Tejas Chopra63cc9afa9d28
diff --git a/headroom/proxy/body_forwarding.py b/headroom/proxy/body_forwarding.pyindex deb07fe20..530284b1e 100644--- a/headroom/proxy/body_forwarding.py+++ b/headroom/proxy/body_forwarding.py@@ -23,8 +23,6 @@ _PYTHON_FORWARDER_MODE_ENV = python_forwarder_mode_policy.PYTHON_FORWARDER_MODE_PythonForwarderMode = python_forwarder_mode_policy.PythonForwarderModeOutboundBodySource = Literal["passthrough", "canonical", "legacy"]-_PYTHON_FORWARDER_MODE_DEFAULT = python_forwarder_mode_policy.PYTHON_FORWARDER_MODE_DEFAULT-@dataclass(frozen=True, slots=True)class OutboundBody:@@ -148,11 +146,6 @@ def _parse_original_body(original_body_bytes: bytes | None) -> dict[str, Any] |return original if isinstance(original, dict) else None-def _original_body_has_signed_thinking_blocks(original_body_bytes: bytes | None) -> bool:- original = _parse_original_body(original_body_bytes)- return original is not None and has_signed_thinking_blocks(original)--#: Allow canonical re-serialization on a thinking-bearing request when every#: thinking block is provably unchanged. **Default ON**; set this to ``0`` (or#: ``false``/``no``/``off``) to restore the previous blanket lock.diff --git a/headroom/proxy/forwarded_headers.py b/headroom/proxy/forwarded_headers.pyindex acf4f2260..a61c811cf 100644--- a/headroom/proxy/forwarded_headers.py+++ b/headroom/proxy/forwarded_headers.py@@ -59,7 +59,6 @@ from typing import TYPE_CHECKING, Anyfrom headroom.proxy.forwarded_policy import (ForwardedHeaderInputs,- header_first,normalize_ip,parse_cidr_list,peer_is_trusted_gateway,@@ -152,17 +151,6 @@ def _peer_host(request: Any) -> str | None:return getattr(client, "host", None)-def _header_first(value: str) -> str:- """Return the leftmost element of a comma-separated header value.-- ``X-Forwarded-For: client, proxy1, proxy2`` → ``"client"``. Empty- input returns ``""``. We intentionally do NOT walk the chain — the- leftmost hop is the only one whose authenticity the immediate- gateway can vouch for, and beyond that we have no trust signal.- """- return header_first(value)--def _read_header(request: Any, name: str) -> str:"""Read a header case-insensitively, ``""`` on miss."""headers = getattr(request, "headers", None)diff --git a/headroom/proxy/helpers.py b/headroom/proxy/helpers.pyindex 600476895..9f9683c99 100644--- a/headroom/proxy/helpers.py+++ b/headroom/proxy/helpers.py@@ -42,16 +42,13 @@ from headroom.proxy.beta_header_merge import (merge_anthropic_beta as merge_anthropic_beta,)from headroom.proxy.beta_header_merge import (- merge_beta_tokens,- split_beta_tokens,+ merge_openai_beta as merge_openai_beta,)from headroom.proxy.beta_header_merge import (- merge_openai_beta as merge_openai_beta,+ split_beta_tokens,)from headroom.proxy.beta_header_policy import (- BETA_HEADER_STICKY_DEFAULT,BETA_HEADER_STICKY_ENV,- BETA_TRACKER_MAX_SESSIONS_DEFAULT,BETA_TRACKER_MAX_SESSIONS_ENV,BetaHeaderStickyMode,resolve_beta_header_sticky_mode,@@ -81,8 +78,6 @@ from headroom.proxy.ccr_marker_policy import ()from headroom.proxy.ccr_session_tracker import SessionCcrTracker as _SessionCcrTrackerfrom headroom.proxy.internal_header_policy import (- INTERNAL_HEADER_PREFIX,- STRIP_INTERNAL_HEADERS_DEFAULT,STRIP_INTERNAL_HEADERS_ENV,StripInternalHeadersMode,resolve_strip_internal_headers_mode,@@ -125,8 +120,6 @@ logger = logging.getLogger("headroom.proxy")_CODEX_WIRE_DEBUG_ENV = "HEADROOM_CODEX_WIRE_DEBUG"_CODEX_WIRE_DEBUG_DIR_ENV = "HEADROOM_CODEX_WIRE_DEBUG_DIR"-_CODEX_WIRE_REDACTED = wire_debug_redaction_policy.WIRE_DEBUG_REDACTED-_CODEX_WIRE_SECRET_KEYS = wire_debug_redaction_policy.WIRE_DEBUG_SECRET_KEYSdef codex_wire_debug_enabled() -> bool:@@ -147,14 +140,6 @@ def _codex_wire_debug_dir() -> Path:return _paths.codex_wire_debug_dir()-def _should_redact_key(key: str) -> bool:- return wire_debug_redaction_policy.should_redact_key(key)---def _redact_value(value: Any) -> Any:- return wire_debug_redaction_policy.redact_for_wire_debug(value)--def redact_for_wire_debug(value: Any) -> Any:"""Redact obvious secrets while preserving request/response shape."""return wire_debug_redaction_policy.redact_for_wire_debug(value)@@ -268,7 +253,6 @@ def capture_codex_wire_debug(# "system_prompt" option — that path is permanently retired by I2 (cache hot# zone never modified). See REALIGNMENT/02-architecture.md §2.2._MEMORY_INJECTION_MODE_ENV = memory_injection_mode_policy.MEMORY_INJECTION_MODE_ENV-_MEMORY_INJECTION_MODE_DEFAULT = memory_injection_mode_policy.MEMORY_INJECTION_MODE_DEFAULTMemoryInjectionMode = memory_injection_mode_policy.MemoryInjectionMode@@ -1319,25 +1303,6 @@ _DECOMPRESS_CHUNK_SIZE = 64 * 1024# Maximum SSE buffer size (10MB - prevents memory exhaustion from malformed streams)MAX_SSE_BUFFER_SIZE = 10 * 1024 * 1024-# Per-event SSE size cap (PR-A8 / P1-8). Configurable via-# HEADROOM_SSE_BUFFER_MAX_BYTES. Guards against pathological huge events-# (a single event > 1 MB by default is treated as an upstream protocol bug-# and surfaces loudly rather than silently growing the buffer).-_SSE_EVENT_MAX_BYTES_ENV = request_limit_policy.SSE_EVENT_MAX_BYTES_ENV-_SSE_EVENT_MAX_BYTES_DEFAULT = request_limit_policy.SSE_EVENT_MAX_BYTES_DEFAULT---def get_sse_event_max_bytes() -> int:- """Return the per-event SSE size cap.-- Read at request time so operators can flip the env var without a- restart. Negative values are rejected loudly (no silent fallback).- """- return request_limit_policy.resolve_sse_event_max_bytes(- os.environ.get(_SSE_EVENT_MAX_BYTES_ENV)- )--# Well-known OpenAI-compatible upstreams, matched by host against the# configured ``--openai-api-url``. Used only to label the dashboard/stats# display provider — the internal provider key stays ``openai`` so pricing@@ -1404,7 +1369,6 @@ def resolve_display_provider(# Configurable via HEADROOM_PROXY_BODY_TOO_LARGE_STATUS for operators who need# to override (no expected production use; documentation knob)._BODY_TOO_LARGE_STATUS_ENV = request_limit_policy.BODY_TOO_LARGE_STATUS_ENV-_BODY_TOO_LARGE_STATUS_DEFAULT = request_limit_policy.BODY_TOO_LARGE_STATUS_DEFAULTdef get_body_too_large_status() -> int:@@ -1414,18 +1378,6 @@ def get_body_too_large_status() -> int:)-_SSE_EVENT_TERMINATORS = sse_byte_buffer_policy.SSE_EVENT_TERMINATORS---def _find_sse_event_terminator(buf: bytearray) -> tuple[int, int] | None:- """Return the earliest complete SSE event terminator in ``buf``."""- return sse_byte_buffer_policy.find_sse_event_terminator(buf)---_SSE_EVENT_LINE_PREFIX = b"event:"-_SSE_DATA_LINE_PREFIX = b"data:"--def safe_decode_for_logging(raw: bytes, *, max_bytes: int | None = None) -> str:"""Decode bytes to a string for **log/diagnostic display only**.@@ -2017,9 +1969,7 @@ def is_anthropic_auth(headers: dict[str, str]) -> bool:# tell its client about its own work. This helper only filters# request-side headers.-_INTERNAL_HEADER_PREFIX = INTERNAL_HEADER_PREFIX_STRIP_INTERNAL_HEADERS_ENV = STRIP_INTERNAL_HEADERS_ENV-_STRIP_INTERNAL_HEADERS_DEFAULT = STRIP_INTERNAL_HEADERS_DEFAULTdef get_strip_internal_headers_mode() -> StripInternalHeadersMode:@@ -2164,10 +2114,7 @@ def log_outbound_headers(# explicit per realignment build constraint #4 — NOT a silent fallback._BETA_HEADER_STICKY_ENV = BETA_HEADER_STICKY_ENV-_BETA_HEADER_STICKY_DEFAULT = BETA_HEADER_STICKY_DEFAULT-_BETA_TRACKER_MAX_SESSIONS_ENV = BETA_TRACKER_MAX_SESSIONS_ENV-_BETA_TRACKER_MAX_SESSIONS_DEFAULT = BETA_TRACKER_MAX_SESSIONS_DEFAULTdef get_beta_header_sticky_mode() -> BetaHeaderStickyMode:@@ -2188,9 +2135,6 @@ def get_beta_tracker_max_sessions() -> int:_split_beta_tokens = split_beta_tokens-_merge_beta_tokens = merge_beta_tokens--class SessionBetaTracker:"""Bounded LRU tracker of beta-header tokens observed per (provider, session).@@ -3742,17 +3686,6 @@ _core_tools_legacy_lock = threading.Lock()_core_tools_legacy_warned = False-def reset_core_tools_legacy_warn_state() -> None:- """Re-arm the legacy-variable warning. Test helper only.-- Without this the flag leaks between tests in a process, so a caplog- assertion on this warning passes alone and fails depending on ordering.- """- global _core_tools_legacy_warned- with _core_tools_legacy_lock:- _core_tools_legacy_warned = False--def _core_tools_override() -> str | None:"""Return the resident-tool override from either env var, canonical first."""global _core_tools_legacy_warned