headroom/tests/test_tokenizer_count_offload.py
JD Davis 039cd2431a
fix(proxy): preserve merged session and quarantine contracts (#2943)
## Description

Forward-fixes two integration contracts exposed while auditing the large
August 12 merge batch on `main`.

The Codex WebSocket request-ID hardening correctly gave every emitted
dashboard/feed row a unique ID, but it also changed the human-readable
`PERF` prefix from the stable WebSocket session ID to that per-emission
ID. That broke operator correlation and the contract documented by the
original merge. This PR separates storage identity from log correlation:
rows remain unique, while `PERF` lines remain grouped under the session
ID.

The same audit found two tokenizer quarantine tests still modeling the
pre-time-cap behavior. Timeout debt no longer activates quarantine after
its deadline expires. The tests now establish a live deadline and
therefore continue to exercise the intended fail-open branch without
weakening the production guard.

## Changes

- Add an optional `RequestOutcome.perf_request_id` correlation field,
defaulting to the existing `request_id` behavior for all current
callers.
- Set that field to the stable session ID for both per-turn and residual
Codex WebSocket emissions.
- Strengthen the lifecycle regression test to prove the unique feed-row
ID is not used as the `PERF` prefix.
- Update tokenizer quarantine tests to model an active, time-capped
quarantine.

This is a forward fix; it does not revert the unique WebSocket request
IDs or the time-capped quarantine behavior.

## Merge-batch audit context

- Audited 48 squash merges from
`12149f7446` through
`941c25d31e`.
- Reviewed repeated-touch hotspots in the OpenAI/Anthropic handlers and
proxy server.
- Confirmed the batch has no deleted or renamed paths and no
revert/supersession commits.
- Reproduced the deterministic failures on the merged head and confirmed
the corresponding pre-batch tokenizer tests passed; the new WebSocket
assertion was introduced by the batch and exposed the dropped
correlation contract.

## Validation

- `ruff check` — passed
- `ruff format --check` — passed
- `mypy headroom` — passed (517 source files)
- Focused proxy/outcome/tokenizer suite — 99 passed
- CI-equivalent Python suite in four `pytest-split` shards — 10,797
passed, 603 skipped
- `cargo fmt --all -- --check` — passed
- `cargo clippy --workspace -- -D warnings` — passed
- New beta-sticky integration suite — 9 passed
- Simulator-backed proxy E2E — 7 passed
- Real local Headroom process against a real local FastAPI upstream:
  - `/livez` healthy
- `/readyz` ready (Kompress correctly optional/degraded in minimal
passthrough mode)
  - OpenAI `/v1/chat/completions` round trip passed
  - Anthropic `/v1/messages` round trip passed
- Local `act` exercised the push workflow's change-detection job and
entered the real lint job. `act` required local no-op cache/setup shims
because its runner action post-hooks/pip pairing are incompatible with
this macOS Docker environment; the actual Ruff/mypy validations above
were run natively and passed.

## Risk

Low. The new field is optional and preserves existing behavior by
default. Only Codex WebSocket outcome emitters opt into a separate log
correlation ID; dashboard/feed row identity remains unchanged.

---------

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
2026-08-12 12:27:05 -05:00

283 lines
12 KiB
Python

"""Token counting must run off the event loop (GH #1701): the Anthropic messages
handler resolved the tokenizer and counted the conversation inline in the async
handler. For HF-backed models (e.g. deepseek-*) first use triggers an unbounded
network download, freezing the whole server (610s request, then /livez, /readyz
and /health hang until kill). The fix routes resolution + counting through
HeadroomProxy._count_tokens_offloaded (compression executor, bounded by
COMPRESSION_TIMEOUT_SECONDS, fail-open to estimation) — shared by every provider
handler (Anthropic, OpenAI, Gemini), since the OpenAI passthrough endpoints
receive the same HF-backed models — and offloads the inline batch
pipeline.apply() calls the same way.
"""
from __future__ import annotations
import asyncio
import inspect
import threading
import time
from headroom.proxy.handlers.anthropic import AnthropicHandlerMixin
from headroom.proxy.handlers.batch import BatchHandlerMixin
from headroom.proxy.handlers.gemini import GeminiHandlerMixin
from headroom.proxy.handlers.openai import OpenAIHandlerMixin
from headroom.proxy.server import (
CompressionQuarantinedError,
ProxyConfig,
create_app,
)
from headroom.proxy.token_counting import (
_count_offloaded,
count_texts_offloaded,
count_tokens_offloaded,
)
from headroom.tokenizers import EstimatingTokenCounter
def _make_proxy(): # noqa: ANN202 — returns the internal HeadroomProxy
app = create_app(
ProxyConfig(
optimize=True,
cache_enabled=False,
rate_limit_enabled=False,
cost_tracking_enabled=False,
)
)
return app.state.proxy
def test_handlers_offload_token_counting_and_batch_apply() -> None:
"""Wiring guard: the request paths must use the offloaded helpers, not inline
get_tokenizer/count_messages or pipeline.apply on the event loop."""
# Every provider handler that counts the original conversation must route
# resolution + counting through the shared fail-open helper, never inline on
# the loop. OpenAI /chat + /responses are multi-provider passthroughs, so an
# HF-routed model (qwen, deepseek, llama, ...) can reach them and cold-load.
for mixin, method in (
(AnthropicHandlerMixin, "handle_anthropic_messages"),
(OpenAIHandlerMixin, "handle_openai_chat"),
(OpenAIHandlerMixin, "handle_openai_responses"),
(GeminiHandlerMixin, "handle_gemini_generate_content"),
(GeminiHandlerMixin, "handle_google_cloudcode_stream"),
(GeminiHandlerMixin, "handle_gemini_count_tokens"),
):
fn = getattr(mixin, method)
assert inspect.iscoroutinefunction(fn), f"{method} must be async"
src = inspect.getsource(fn)
assert "_count_tokens_offloaded(" in src, f"{method}: token counting not offloaded"
assert "tokenizer = get_tokenizer(" not in src, (
f"{method}: tokenizer resolved inline on the loop"
)
fn = GeminiHandlerMixin.handle_gemini_stream_generate_content
assert inspect.iscoroutinefunction(fn)
src = inspect.getsource(fn)
assert "_count_texts_offloaded(" in src, "streaming Gemini text counting not offloaded"
assert "tokenizer = get_tokenizer(" not in src, "tokenizer resolved inline on the loop"
assert "count_text(" not in src, "streaming Gemini count_text still runs on the loop"
assert "_dict_parts(" in src, "streaming Gemini must reuse the shared _dict_parts coercion"
assert 'isinstance(part.get("text"), str)' in src, (
"streaming Gemini must skip non-str text so count_text can't 500"
)
for mixin, method in (
(AnthropicHandlerMixin, "handle_anthropic_batch_create"),
(BatchHandlerMixin, "handle_google_batch_create"),
(BatchHandlerMixin, "_compress_batch_jsonl"),
):
fn = getattr(mixin, method)
assert inspect.iscoroutinefunction(fn), f"{method} must be async"
src = inspect.getsource(fn)
if "pipeline.apply(" in src:
assert "_run_compression_in_executor(" in src, f"{method}: apply() not offloaded"
assert "COMPRESSION_TIMEOUT_SECONDS" in src, f"{method}: offload missing timeout"
helper_src = inspect.getsource(_count_offloaded)
assert "COMPRESSION_TIMEOUT_SECONDS" in helper_src
assert "EstimatingTokenCounter" in helper_src, "helper must fail open to estimation"
async def test_count_tokens_offloaded_runs_on_worker_thread(monkeypatch) -> None: # noqa: ANN001
proxy = _make_proxy()
loop_thread = threading.current_thread().name
seen: dict[str, str] = {}
class _SpyTokenizer(EstimatingTokenCounter):
def count_messages(self, messages): # noqa: ANN001, ANN201
seen["thread"] = threading.current_thread().name
return super().count_messages(messages)
monkeypatch.setattr("headroom.tokenizers.get_tokenizer", lambda *a, **k: _SpyTokenizer())
_, tokens = await proxy._count_tokens_offloaded("gpt-4", [{"role": "user", "content": "hi"}])
assert tokens > 0
assert seen["thread"].startswith("headroom-compress")
assert seen["thread"] != loop_thread
async def test_count_tokens_offloaded_keeps_loop_responsive(monkeypatch) -> None: # noqa: ANN001
"""A slow tokenizer (stand-in for an HF network load) must not starve the loop —
the pre-fix inline call yielded ~0 ticks here."""
proxy = _make_proxy()
ticks = 0
async def _ticker() -> None:
nonlocal ticks
while True:
await asyncio.sleep(0.01)
ticks += 1
class _SlowTokenizer(EstimatingTokenCounter):
def count_messages(self, messages): # noqa: ANN001, ANN201
time.sleep(0.3)
return super().count_messages(messages)
monkeypatch.setattr("headroom.tokenizers.get_tokenizer", lambda *a, **k: _SlowTokenizer())
tick_task = asyncio.create_task(_ticker())
try:
_, tokens = await proxy._count_tokens_offloaded("m", [{"role": "user", "content": "hi"}])
finally:
tick_task.cancel()
assert tokens > 0
assert ticks >= 5
async def test_count_tokens_offloaded_fails_open(monkeypatch) -> None: # noqa: ANN001
"""Resolution errors and timeouts downgrade to estimation instead of raising."""
proxy = _make_proxy()
def _boom(*a, **k): # noqa: ANN002, ANN003, ANN202
raise RuntimeError("tokenizer backend exploded")
monkeypatch.setattr("headroom.tokenizers.get_tokenizer", _boom)
tokenizer, tokens = await proxy._count_tokens_offloaded(
"deepseek-chat", [{"role": "user", "content": "hello world"}]
)
assert isinstance(tokenizer, EstimatingTokenCounter)
assert tokens > 0
# Logged-once bookkeeping records the downgraded model.
assert "deepseek-chat" in proxy._token_count_fallback_models
async def test_count_tokens_offloaded_fails_open_on_executor_quarantine() -> None:
"""Now that OpenAI/Gemini counting shares the compression executor, an
unrelated request's compression timeout can quarantine it — the next
``_run_compression_in_executor`` call raises ``CompressionQuarantinedError``
immediately (process-wide state). A request that is only counting tokens
must not 500 on that; it fails open to estimation like any other error."""
# The executor's ``except Exception`` fail-open only catches the quarantine
# error because it subclasses Exception — pin that contract.
assert issubclass(CompressionQuarantinedError, Exception)
proxy = _make_proxy()
# Record a concurrent compression as timed out so the real executor guard
# quarantines the next call — no mock of the helper itself. Since the
# quarantine became time-capped (#2412), standing debt alone no longer
# quarantines: the deadline armed by the fresh timeout must still be in
# the future, so arm it the way a real timeout would.
proxy._compression_timed_out_in_flight = 1
proxy._compression_quarantine_deadline = time.monotonic() + 60.0
tokenizer, tokens = await proxy._count_tokens_offloaded(
"qwen2.5-coder", [{"role": "user", "content": "hello world"}]
)
assert isinstance(tokenizer, EstimatingTokenCounter)
assert tokens > 0
assert "qwen2.5-coder" in proxy._token_count_fallback_models
async def test_count_tokens_offloaded_returns_count_text_capable_tokenizer() -> None:
"""The fail-open tokenizer should still support text counting for callers
that need per-fragment accounting."""
proxy = _make_proxy()
# Quarantine forces the fail-open branch (an EstimatingTokenCounter).
# Post-#2412 the quarantine is time-capped, so the deadline must be armed
# alongside the standing debt.
proxy._compression_timed_out_in_flight = 1
proxy._compression_quarantine_deadline = time.monotonic() + 60.0
# The empty-messages count is intentionally discarded by that handler
# (it sums text parts itself), so only the tokenizer matters here.
tokenizer, _ = await proxy._count_tokens_offloaded("qwen2.5-coder", [])
assert isinstance(tokenizer, EstimatingTokenCounter)
# The streaming handler's per-part loop must not raise on the fallback.
assert tokenizer.count_text("hello world") > 0
async def test_count_texts_offloaded_runs_on_worker_thread(monkeypatch) -> None: # noqa: ANN001
proxy = _make_proxy()
loop_thread = threading.current_thread().name
seen: dict[str, str] = {}
class _SpyTokenizer(EstimatingTokenCounter):
def count_text(self, text): # noqa: ANN001, ANN201
seen["thread"] = threading.current_thread().name
return super().count_text(text)
monkeypatch.setattr("headroom.tokenizers.get_tokenizer", lambda *a, **k: _SpyTokenizer())
_, tokens = await proxy._count_texts_offloaded("gemini-pro", ["hello", "world"])
assert tokens > 0
assert seen["thread"].startswith("headroom-compress")
assert seen["thread"] != loop_thread
async def test_count_texts_offloaded_fails_open(monkeypatch) -> None: # noqa: ANN001
"""The texts variant downgrades to estimation on a resolution error, the same
as the messages variant (its fail-open branch was previously uncovered)."""
proxy = _make_proxy()
def _boom(*a, **k): # noqa: ANN002, ANN003, ANN202
raise RuntimeError("tokenizer backend exploded")
monkeypatch.setattr("headroom.tokenizers.get_tokenizer", _boom)
tokenizer, tokens = await proxy._count_texts_offloaded("deepseek-chat", ["hello", "world"])
assert isinstance(tokenizer, EstimatingTokenCounter)
assert tokens > 0
assert "deepseek-chat" in proxy._token_count_fallback_models
async def test_count_offloaded_without_executor_estimates() -> None:
"""An owner with no compression executor (a lightweight caller or test double)
fails open to estimation inline instead of crashing on the missing runner."""
class _NoExecutorOwner:
pass
owner = _NoExecutorOwner()
tok, n_msg = await count_tokens_offloaded(
owner, "gpt-4", [{"role": "user", "content": "hello world"}]
)
assert isinstance(tok, EstimatingTokenCounter)
assert n_msg > 0
tok2, n_txt = await count_texts_offloaded(owner, "gemini-pro", ["hello", "world"])
assert isinstance(tok2, EstimatingTokenCounter)
assert n_txt > 0
async def test_count_texts_offloaded_sums_fragments(monkeypatch) -> None: # noqa: ANN001
"""The streaming rewrite sums per-fragment counts, matching the old per-part
count_text loop it replaced."""
proxy = _make_proxy()
monkeypatch.setattr(
"headroom.tokenizers.get_tokenizer", lambda *a, **k: EstimatingTokenCounter()
)
fragments = ["hello", "world", "foo"]
_, total = await proxy._count_texts_offloaded("gemini-pro", fragments)
est = EstimatingTokenCounter()
assert total == sum(est.count_text(f) for f in fragments)
assert total > 0