mirror of
https://github.com/headroomlabs-ai/headroom.git
synced 2026-08-27 14:17:10 -04:00
PR-A6 of the Phase A cache-safety lockdown. Eliminates P5-50 and preps
P0-6 (memory tool injection toggling).
Two cache-killer patterns the merge + tracker defeat:
1. Mid-session mutation: when memory was enabled the proxy did an
ad-hoc concat of `context-management-2025-06-27` onto the client
value (anthropic.py:1244-1248). The order varied with the client
value, breaking byte-stable headers across turns.
2. Token drop-out across turns: clients (Claude Code, Codex CLI) MAY
drop a beta token between turn N and turn N+1 even when the proxy
mutated turn N to add it. The cache hot zone is positional, so the
next turn's prefix bytes hash differently and the prefix-cache
read misses.
Changes
-------
`headroom/proxy/helpers.py`
* `merge_anthropic_beta` / `merge_openai_beta`: pure, deterministic,
order-preserving merge. Client tokens first (in their original
order), then Headroom-required tokens (in the order passed). Dedupe
is case-insensitive but preserves the original casing of the first
occurrence. No regex.
* `SessionBetaTracker`: bounded LRU keyed by (provider, session_id),
unioning client tokens with previously-seen tokens. OrderedDict
LRU; threading.RLock for thread safety (mirrors the
CompressionCache pattern from compression_cache.py).
* `get_session_beta_tracker` / `_reset_session_beta_tracker_for_test`
process-wide singleton with test reset.
* `log_beta_header_merge`: structured log per cache-affecting merge.
* Env-var knobs (NO HARDCODES):
- HEADROOM_BETA_HEADER_STICKY=enabled|disabled (default enabled).
- HEADROOM_BETA_TRACKER_MAX_SESSIONS (default 1000).
`headroom/proxy/handlers/anthropic.py`
* After `compute_session_id` (line ~744): record client
`anthropic-beta` against the session tracker, write the sticky
value back into `headers` if changed. Order matters: sticky-merge
FIRST so memory-injection has the canonical baseline.
* Memory-injection site (line ~1244): replace the ad-hoc concat with
`merge_anthropic_beta(headers["anthropic-beta"], required_tokens)`.
`headroom/proxy/handlers/openai.py`
* Chat-completions (line ~360): record/merge `openai-beta`.
* /v1/responses HTTP (line ~1213): compute `_responses_session_id`
and record/merge `openai-beta`.
* /v1/responses WS (line ~1711): replace the ad-hoc absent-only
inject with `merge_openai_beta(sticky, ["responses_websockets=
2026-02-06"])`. Replaces any case-variants of the existing key.
Tests
-----
`tests/test_anthropic_beta_session_sticky.py` (26 tests):
* Pure helper: empty inputs, only-client, only-headroom, ordering,
dedupe casing, deterministic memory-injection order, no-double-
inject when token already present.
* Tracker: sticky-on across turns even when client drops, casing
preservation, provider namespace independence, LRU eviction at
max_sessions, env-var validation (loud failures), thread safety
under 16-thread concurrent access, blank-input rejection.
`tests/test_openai_beta_session_sticky.py` (17 tests):
* Mirror of the anthropic suite for `OpenAI-Beta`.
* Plus WS-specific coverage: sticky-then-merge of
`responses_websockets=2026-02-06` against client baseline.
`tests/test_openai_codex_routing.py`
* Add `session_tracker_store` stub to `_DummyOpenAIHandler` so the
routing tests still exercise the responses HTTP handler now that
it computes a session_id for beta-merge.
Notes
-----
Build constraints honored:
* Configurable: HEADROOM_BETA_HEADER_STICKY,
HEADROOM_BETA_TRACKER_MAX_SESSIONS.
* No regex, no hardcodes (env-var bounds), no fallbacks (disabled
mode is operator opt-in for diagnostics, loud failures on invalid
values).
* Structured tracing log via `log_beta_header_merge`.
Acceptance:
* 43 new tests pass.
* `cargo test --workspace` green (no Rust changes).
* `make ci-precheck` green.
226 lines
7.9 KiB
Python
226 lines
7.9 KiB
Python
"""Session-sticky `OpenAI-Beta` tests for PR-A6 (P5-50, preps P0-6).
|
|
|
|
OpenAI's `OpenAI-Beta` follows the same comma-separated convention as
|
|
Anthropic's `anthropic-beta`. The proxy auto-injects
|
|
`responses_websockets=2026-02-06` on the WS path when absent
|
|
(handlers/openai.py:~1711). PR-A6 routes that injection through
|
|
`merge_openai_beta` so the client's tokens are preserved and the
|
|
auto-injected token is appended deterministically.
|
|
|
|
The same `SessionBetaTracker` (provider-aware, keyed by
|
|
``(provider, session_id)``) backs both providers — one tracker, two
|
|
namespaces.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import threading
|
|
|
|
import pytest
|
|
|
|
from headroom.proxy.helpers import (
|
|
SessionBetaTracker,
|
|
_reset_session_beta_tracker_for_test,
|
|
merge_openai_beta,
|
|
)
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Pure helper: `merge_openai_beta`
|
|
# ---------------------------------------------------------------------------
|
|
|
|
|
|
def test_merge_helper_pure_function() -> None:
|
|
"""Same inputs → same output, no global state."""
|
|
a = merge_openai_beta("a,b", ["c"])
|
|
b = merge_openai_beta("a,b", ["c"])
|
|
assert a == b == "a,b,c"
|
|
|
|
|
|
def test_merge_helper_empty_inputs_returns_empty_string() -> None:
|
|
assert merge_openai_beta(None, []) == ""
|
|
assert merge_openai_beta("", []) == ""
|
|
|
|
|
|
def test_merge_helper_only_client() -> None:
|
|
assert merge_openai_beta("alpha=1,beta=2", []) == "alpha=1,beta=2"
|
|
|
|
|
|
def test_merge_helper_only_headroom() -> None:
|
|
assert merge_openai_beta(None, ["responses_websockets=2026-02-06"]) == (
|
|
"responses_websockets=2026-02-06"
|
|
)
|
|
|
|
|
|
def test_merge_helper_preserves_client_order_appends_headroom() -> None:
|
|
out = merge_openai_beta(
|
|
"alpha=1,beta=2",
|
|
["responses_websockets=2026-02-06"],
|
|
)
|
|
assert out == "alpha=1,beta=2,responses_websockets=2026-02-06"
|
|
|
|
|
|
def test_merge_helper_no_double_inject_when_already_present() -> None:
|
|
out = merge_openai_beta(
|
|
"responses_websockets=2026-02-06,alpha=1",
|
|
["responses_websockets=2026-02-06"],
|
|
)
|
|
assert out == "responses_websockets=2026-02-06,alpha=1"
|
|
|
|
|
|
def test_dedupe_case_insensitive_preserves_first_casing() -> None:
|
|
# OpenAI tokens are usually `kebab=value` so casing rarely differs in
|
|
# practice, but the helper's contract is provider-agnostic.
|
|
assert (
|
|
merge_openai_beta("Responses_Websockets=2026-02-06", ["responses_websockets=2026-02-06"])
|
|
== "Responses_Websockets=2026-02-06"
|
|
)
|
|
|
|
|
|
def test_merge_helper_skips_empty_tokens() -> None:
|
|
assert merge_openai_beta("a, ,b", []) == "a,b"
|
|
assert merge_openai_beta("a", ["", "b", " "]) == "a,b"
|
|
|
|
|
|
def test_test_memory_injection_appends_deterministic_order() -> None:
|
|
"""Auto-injected `responses_websockets` appends AFTER client tokens."""
|
|
out = merge_openai_beta("client-token", ["responses_websockets=2026-02-06"])
|
|
assert out == "client-token,responses_websockets=2026-02-06"
|
|
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# `SessionBetaTracker` — provider="openai"
|
|
# ---------------------------------------------------------------------------
|
|
|
|
|
|
@pytest.fixture(autouse=True)
|
|
def _isolate_tracker(monkeypatch: pytest.MonkeyPatch) -> None:
|
|
monkeypatch.delenv("HEADROOM_BETA_HEADER_STICKY", raising=False)
|
|
monkeypatch.delenv("HEADROOM_BETA_TRACKER_MAX_SESSIONS", raising=False)
|
|
_reset_session_beta_tracker_for_test()
|
|
yield
|
|
_reset_session_beta_tracker_for_test()
|
|
|
|
|
|
def test_beta_seen_turn_1_present_in_turn_2_even_if_client_drops() -> None:
|
|
tracker = SessionBetaTracker(max_sessions=10)
|
|
|
|
out1 = tracker.record_and_get_sticky_betas(
|
|
provider="openai",
|
|
session_id="s-1",
|
|
client_value="responses_websockets=2026-02-06,extra-beta=1",
|
|
)
|
|
assert out1 == "responses_websockets=2026-02-06,extra-beta=1"
|
|
|
|
out2 = tracker.record_and_get_sticky_betas(
|
|
provider="openai",
|
|
session_id="s-1",
|
|
client_value="responses_websockets=2026-02-06",
|
|
)
|
|
assert out2 == "responses_websockets=2026-02-06,extra-beta=1"
|
|
|
|
|
|
def test_client_value_preserved_when_no_injection() -> None:
|
|
tracker = SessionBetaTracker(max_sessions=10)
|
|
out = tracker.record_and_get_sticky_betas(
|
|
provider="openai", session_id="s-1", client_value="alpha,beta"
|
|
)
|
|
assert out == "alpha,beta"
|
|
|
|
|
|
def test_lru_eviction_at_max_sessions() -> None:
|
|
tracker = SessionBetaTracker(max_sessions=2)
|
|
tracker.record_and_get_sticky_betas(provider="openai", session_id="s-1", client_value="a")
|
|
tracker.record_and_get_sticky_betas(provider="openai", session_id="s-2", client_value="b")
|
|
tracker.record_and_get_sticky_betas(provider="openai", session_id="s-1", client_value=None)
|
|
tracker.record_and_get_sticky_betas(provider="openai", session_id="s-3", client_value="c")
|
|
# s-2 evicted.
|
|
out = tracker.record_and_get_sticky_betas(
|
|
provider="openai", session_id="s-2", client_value=None
|
|
)
|
|
assert out == ""
|
|
|
|
|
|
def test_disabled_mode_passes_through(monkeypatch: pytest.MonkeyPatch) -> None:
|
|
monkeypatch.setenv("HEADROOM_BETA_HEADER_STICKY", "disabled")
|
|
tracker = SessionBetaTracker(max_sessions=10)
|
|
|
|
out1 = tracker.record_and_get_sticky_betas(
|
|
provider="openai", session_id="s-1", client_value="alpha"
|
|
)
|
|
assert out1 == "alpha"
|
|
out2 = tracker.record_and_get_sticky_betas(
|
|
provider="openai", session_id="s-1", client_value=None
|
|
)
|
|
assert out2 == ""
|
|
|
|
|
|
def test_thread_safe_concurrent_access() -> None:
|
|
tracker = SessionBetaTracker(max_sessions=10)
|
|
n_threads = 16
|
|
iterations = 50
|
|
errors: list[BaseException] = []
|
|
|
|
def worker(idx: int) -> None:
|
|
try:
|
|
for i in range(iterations):
|
|
tracker.record_and_get_sticky_betas(
|
|
provider="openai",
|
|
session_id="shared",
|
|
client_value=f"o-t{idx}-i{i}",
|
|
)
|
|
except BaseException as e: # noqa: BLE001
|
|
errors.append(e)
|
|
|
|
threads = [threading.Thread(target=worker, args=(idx,)) for idx in range(n_threads)]
|
|
for t in threads:
|
|
t.start()
|
|
for t in threads:
|
|
t.join()
|
|
assert errors == []
|
|
final = tracker.record_and_get_sticky_betas(
|
|
provider="openai", session_id="shared", client_value=None
|
|
)
|
|
final_tokens = set(final.split(","))
|
|
expected = {f"o-t{idx}-i{i}" for idx in range(n_threads) for i in range(iterations)}
|
|
assert expected.issubset(final_tokens)
|
|
|
|
|
|
def test_provider_namespaces_are_independent() -> None:
|
|
tracker = SessionBetaTracker(max_sessions=10)
|
|
tracker.record_and_get_sticky_betas(
|
|
provider="openai", session_id="shared", client_value="o-token"
|
|
)
|
|
tracker.record_and_get_sticky_betas(
|
|
provider="anthropic", session_id="shared", client_value="a-token"
|
|
)
|
|
out_openai = tracker.record_and_get_sticky_betas(
|
|
provider="openai", session_id="shared", client_value=None
|
|
)
|
|
out_anth = tracker.record_and_get_sticky_betas(
|
|
provider="anthropic", session_id="shared", client_value=None
|
|
)
|
|
assert out_openai == "o-token"
|
|
assert out_anth == "a-token"
|
|
|
|
|
|
def test_ws_required_token_appended_deterministically() -> None:
|
|
"""Mirrors the WS handler logic — record client value, then merge required."""
|
|
tracker = SessionBetaTracker(max_sessions=10)
|
|
sticky = tracker.record_and_get_sticky_betas(
|
|
provider="openai",
|
|
session_id="ws-1",
|
|
client_value="custom-beta=1",
|
|
)
|
|
merged = merge_openai_beta(sticky, ["responses_websockets=2026-02-06"])
|
|
assert merged == "custom-beta=1,responses_websockets=2026-02-06"
|
|
|
|
|
|
def test_ws_required_token_no_double_when_client_already_has_it() -> None:
|
|
tracker = SessionBetaTracker(max_sessions=10)
|
|
sticky = tracker.record_and_get_sticky_betas(
|
|
provider="openai",
|
|
session_id="ws-2",
|
|
client_value="responses_websockets=2026-02-06",
|
|
)
|
|
merged = merge_openai_beta(sticky, ["responses_websockets=2026-02-06"])
|
|
assert merged == "responses_websockets=2026-02-06"
|