Files
headroom/tests/test_route_advice.py
gglucassandClaude Opus 5.5 762a337bbe fix(proxy): record a stream the client closes after its terminal event as completed (#4286)
## Description

Codex CLI (0.156.1) closes the HTTP connection as soon as it reads
`response.completed`. It does not wait for the upstream EOF. In
`StreamingMixin._stream_response`, `completed_normally` is set only
after `async for chunk in response.aiter_bytes()` ends. The client's
hang-up arrives as `asyncio.CancelledError` (the generator is waiting
for the next upstream read) or `GeneratorExit` (the generator is
suspended at a `yield`). Neither is an `Exception`, so both skip that
assignment and reach `finally` with `completed_normally=False`. Since
#3747 that records the turn as a 502 failure. `emit_request_outcome`
short-circuits on `>= 400`, so the turn's usage, cache metrics and
prefix-tracker update are dropped, even though the client got a complete
200 response.

A second ordering problem hits the `GeneratorExit` case. The read loop
yields each chunk to the client before `_parse_sse_usage_from_buffer`
parses it. If the client closes at the yield of the terminal chunk, the
terminal event and its usage are never parsed.

Reproduced on wheel 0.41.0 and on current main. `codex exec` through a
local proxy prints `ok`, while `/stats` shows `requests.total 0`,
`failed_by_provider.openai 1`, `tokens.input 0`. A
`HEADROOM_CODEX_WIRE_DEBUG=1` capture has every upstream chunk,
including the complete `response.completed` event, and no
`http_stream_upstream_complete` frame (written only after the loop ends
normally). `_sse_event_outcome` classifies the captured events as
terminal and not failed, so the classification is right. Only the timing
of the hang-up loses the turn.

## Type of Change

- [x] Bug fix (non-breaking change that fixes an issue)
- [ ] New feature (non-breaking change that adds functionality)
- [ ] Breaking change (fix or feature that would cause existing
functionality to change)
- [ ] Documentation update
- [ ] Performance improvement
- [ ] Code refactoring (no functional changes)

## Changes Made

- `_stream_response` (direct HTTP SSE):
- Each chunk is parsed (usage, `stream_terminal`, `stream_failed`)
before it is yielded. The memory filter and context guard still run on
exactly the bytes the client receives, in the same order. Nothing awaits
between the parse and the yield, so streaming latency is unchanged.
- If that parse raises (a complete event with invalid UTF-8), the error
is held until the chunk has been forwarded, then raised. The client gets
the upstream bytes and then the error event, as on main.
- A new `except (GeneratorExit, asyncio.CancelledError)` records success
only when a terminal event was handed to the client, then re-raises. The
outcome status uses `completed_normally or client_left_after_terminal`.
- That is tracked by `_client_frame`, which wraps `_guard_client_bytes`
at every client-facing yield: the main loop and the memory filter's
closing and continuation frames. Unfiltered, each yield carries the
whole parsed chunk, so it reuses the parser's `stream_terminal`. With
the memory tool filter on, the parser can see `message_stop` before the
client gets its first frame, and the filter holds `message_delta` /
`message_stop` until after the stream, so each forwarded frame (a whole
event) is classified as it goes out, until the terminal one.
- `completed_normally` itself still means "the loop finished". It also
gates draining mid-turn pending messages, which must not be yielded into
a closed stream.
- `_stream_response_bedrock`: same pattern. Each event is yielded right
after its own usage and terminal bookkeeping, so the terminal flag
always matches what the client got, and a hang-up after the terminal
event sets `completed_normally`. That `finally` never yields.
- `_stream_openai_via_backend`: it already parsed each chunk before
yielding all of it, with no filter. It now gets the same hang-up
handling.
- A hang-up before the terminal event reaches the client is still a 502,
as #3747 intended.
- Tests:
- `tests/test_stream_client_hangup_after_terminal.py` drives
`_stream_response` with a fake upstream that sends `response.created`,
then `response.completed` with usage, then never reaches EOF. The client
either `aclose()`s at the terminal chunk or is cancelled while the proxy
waits for EOF. Both must record status 200 with input 1,000 / cached 600
/ output 5. A hang-up before the terminal event must still record 502.
- Same file, memory tool filter on, one upstream chunk holding a whole
Anthropic message (`message_start` through `message_stop`):
`test_memory_filtered_hangup_before_the_client_gets_message_stop_is_a_failure[1-5]`
reads 1 to 5 frames (frames 1-4 come from that one chunk, frame 5 is the
held `message_delta`), closes, and must record 502.
`test_memory_filtered_hangup_after_message_stop_is_forwarded_records_success`
reads until `message_stop` is forwarded, closes, and must record 200
with its usage.
- `test_malformed_event_after_a_valid_one_still_forwards_the_chunk`: a
chunk with a valid event and then one with invalid UTF-8 reaches the
client verbatim, followed by the error event, and records 502.
-
`tests/test_route_advice.py::test_backend_client_hangup_after_terminal_event_is_completed`
covers both backend wrappers.
- `pyproject.toml` / `uv.lock` (review: Dependency audit): the `aiohttp`
constraint floor is now `>=3.14.4` (CVE-2026-111063, CVE-2026-111064 and
GHSA-2g87-pp6c-29x5, all fixed in 3.14.4), relocked with `uv lock
--upgrade-package aiohttp==3.14.4`. Only aiohttp (3.14.3 -> 3.14.4) and
yarl (1.22.0 -> 1.25.1, which 3.14.4 requires) move. Main still resolves
3.14.3, so the same commit is on #4299, #4286, #4285 and #4269.

## Testing

- [x] Unit tests pass (`pytest`)
- [x] Linting passes (`ruff check .`)
- [ ] Type checking passes (`mypy headroom`)
- [x] New tests added for new functionality
- [x] Manual testing performed

mypy was run on the touched module only (output below), not on all of
`headroom`.

### Test Output

```text
$ uv run --frozen --extra dev pytest <the 43 test files that reference _stream_response, handlers.streaming, _stream_openai_via_backend, context_guard, StreamingMixin or sse_buffer> tests/test_proxy_openai_responses_integration.py -q -p no:randomly
1111 passed, 14 skipped, 1 warning in 58.03s

# The new tests against 4eeb4619's streaming.py (this PR's previous head):
FAILED test_stream_client_hangup_after_terminal.py::test_memory_filtered_hangup_before_the_client_gets_message_stop_is_a_failure[1]
FAILED ...[2]  FAILED ...[3]  FAILED ...[4]  FAILED ...[5]      (each: assert 200 == 502)
FAILED test_stream_client_hangup_after_terminal.py::test_malformed_event_after_a_valid_one_still_forwards_the_chunk
6 failed, 4 passed   (the four passing: both Codex hang-up cases, the hang-up-before-terminal guard, the success after the filtered terminal)

# The same two test files against main's streaming.py:
FAILED test_stream_client_hangup_after_terminal.py::...[closed_at_terminal_yield]
FAILED test_stream_client_hangup_after_terminal.py::...[cancelled_awaiting_eof]
FAILED test_stream_client_hangup_after_terminal.py::test_memory_filtered_hangup_after_message_stop_is_forwarded_records_success
FAILED test_route_advice.py::test_backend_client_hangup_after_terminal_event_is_completed[anthropic]
FAILED test_route_advice.py::test_backend_client_hangup_after_terminal_event_is_completed[openai]
5 failed, 61 passed

$ uvx ruff@0.17.0 check headroom/proxy/handlers/streaming.py tests/test_stream_client_hangup_after_terminal.py tests/test_route_advice.py
All checks passed!
$ uvx ruff@0.17.0 format --check (same files)
3 files already formatted
$ uv run --frozen --extra dev mypy headroom/proxy/handlers/streaming.py
Success: no issues found in 1 source file
```

## Real Behavior Proof

- Environment: macOS, Python 3.11, Codex CLI 0.156.1 signed in to
ChatGPT. A local proxy on this branch's head (ef0773a7) started with
`python -m headroom.cli proxy --port 6821 --no-http2 --no-rate-limit`
and a clean HOME. The first revision of this PR also ran the Codex steps
against main's `streaming.py`: main reported `requests.total 0`,
`failed_by_provider.openai 1`, `tokens.input 0`.
- Exact command / steps: (1) Codex: `codex exec --skip-git-repo-check -c
'model_providers.headroom.base_url="http://127.0.0.1:6821/v1"' "Reply
with the single word: ok"` twice, then `GET /stats`. (2) Controlled
upstream (a local stub, not Anthropic): a server on 127.0.0.1:6831
answers `POST /v1/messages` with one write holding a whole message
(`message_start` through `message_stop`) and holds its EOF back 3 s. The
proxy runs with `--memory --memory-storage global --no-memory-context
--anthropic-api-url http://127.0.0.1:6831`, so the memory tool filter is
on. A raw-socket client POSTs a streaming request with one client tool
(memory tools are injected only beside client-declared tools) and closes
right after reading `message_start`. A second request is read through
`message_stop`. The early hang-up is repeated on a proxy running
4eeb4619's `streaming.py`.
- Observed result: (1) Codex printed `ok` both times. `/stats`:
`requests.total 2`, `failed 0`, `tokens.input 30198`, `tokens.output
10`, one PERF line per turn, no error or traceback in the log. (2) This
branch: the early hang-up client got only `message_start`, and `/stats`
went to `failed 1` (`failed_by_provider.anthropic 1`) with no PERF line
for that turn. The full read added 1 to `requests.total`, with a PERF
line. On 4eeb4619 the same early hang-up was recorded as a success
(`requests.total 1`, `failed 0`, PERF line). Without the client tool (no
memory tools, so no filter) the client got all six events from the one
chunk and hung up before the stub's EOF; that turn was recorded as
completed, `total_ms=3`.
- Not tested: the Bedrock and OpenAI-via-backend wrappers against a live
backend (unit tests only), a proxy-executed memory continuation round
through a real proxy (unit tests only), and Codex over WebSocket, which
uses a different handler.

## Runtime Rollout Safety

- Rollout-managed feature(s): None.
- Minimum rollout channel: Stable.
- Stable/default behavior changed: Yes, accounting only. A stream whose
terminal event arrived without a failure before the client hung up is
now recorded as completed: its usage, cache metrics and prefix-tracker
update are kept instead of being counted as a 502. Bytes sent to the
client are unchanged.
- Kill switch / disable path: Not applicable; no new behavior to switch
off.
- Unsafe override required: No.
- Qualification impact: None.
- Rollback path: Revert this commit.

## Review Readiness

- [x] I have performed a self-review
- [x] This PR is ready for human review

## Checklist

- [x] My code follows the project's style guidelines
- [x] I have performed a self-review of my code
- [x] I have commented my code, particularly in hard-to-understand areas
- [ ] I have made corresponding changes to the documentation
- [x] My changes generate no new warnings
- [x] I have added tests that prove my fix is effective or that my
feature works
- [x] New and existing unit tests pass locally with my changes
- [x] I did **not** edit `CHANGELOG.md` — it is generated by
release-please from my Conventional Commit PR title (a CI guard enforces
this)

## Screenshots (if applicable)

Not applicable.

## Additional Notes

- Malformed upstream content: a parse failure on a chunk (a complete
event with invalid UTF-8) is raised after that chunk is forwarded, so
the client gets the upstream bytes and then the error event, as on main.
`_stream_openai_via_backend` already parsed before yielding on main; its
order is unchanged here.
- "Terminal" means the terminal event was handed to the client, i.e. the
frame carrying it was yielded to the ASGI server. If a client
disconnects in the same instant that frame goes out, the turn counts as
completed. The upstream did bill it, so its usage belongs in the stats.
- Cost: the unfiltered path (Codex, plain Anthropic) adds no parsing.
With the memory filter on, each forwarded frame is classified once, and
only until the terminal one.
- Related: #3747, which introduced the failed-stream classification this
builds on.
- No documentation change needed.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

<!-- devin-review-badge-begin -->

---

<a href="https://app.devin.ai/review/headroomlabs-ai/headroom/pull/4286"
target="_blank"><picture><source media="(prefers-color-scheme: dark)"
srcset="https://static.devin.ai/assets/gh-devin-review-dark.svg?v=4"><img
src="https://static.devin.ai/assets/gh-devin-review-light.svg?v=4"
alt="Devin Review"></picture></a>
<!-- devin-review-badge-end -->

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-20 21:46:43 -05:00

729 lines
24 KiB
Python

"""Per-request backend selection published by an extension.
The property under test throughout is the one that makes this safe to merge:
with nothing published, every path is exactly what it was before.
"""
from __future__ import annotations
import asyncio
import json
from types import SimpleNamespace
import pytest
from headroom.proxy.route_advice import (
BackendResolver,
RouteAdvice,
advice_from,
)
class _Req:
def __init__(self, **state):
self.state = SimpleNamespace(**state)
DEFAULT = object() # stands in for the configured backend
def _sse_contains_error_event(payload: bytes) -> bool:
"""True when any complete SSE event in ``payload`` is a failure event."""
from headroom.proxy.handlers.streaming import _sse_event_outcome
from headroom.proxy.helpers import parse_sse_events_from_byte_buffer
return any(
_sse_event_outcome(event_name, data, "chat")[0]
for event_name, data in parse_sse_events_from_byte_buffer(bytearray(payload))
)
# --- reading what an extension published ------------------------------------
def test_no_extension_means_no_advice():
assert advice_from(_Req()) is None
assert advice_from(object()) is None # no .state at all
assert advice_from(None) is None
def test_advice_is_duck_typed_so_extensions_need_not_import_us():
a = advice_from(
_Req(
headroom_route=SimpleNamespace(
model="moonshot/kimi-k2", provider="moonshot", reason="cheaper"
)
)
)
assert a == RouteAdvice("moonshot/kimi-k2", "moonshot", "cheaper")
def test_an_extension_may_omit_everything_but_the_model():
a = advice_from(_Req(headroom_route=SimpleNamespace(model="gpt-5-mini")))
assert a.model == "gpt-5-mini" and a.provider == ""
def test_malformed_advice_is_ignored_rather_than_raised():
for bad in (
SimpleNamespace(),
SimpleNamespace(model=""),
SimpleNamespace(model=123),
"not an object",
):
assert advice_from(_Req(headroom_route=bad)) is None
def test_advice_needs_a_model():
with pytest.raises(ValueError):
RouteAdvice("")
# --- absent means unchanged, which is the whole safety argument -------------
def test_no_advice_returns_the_configured_backend():
r = BackendResolver(DEFAULT)
assert r.for_request(_Req()) is DEFAULT
def test_no_advice_and_no_configured_backend_stays_none():
"""None is not "no backend", it is the direct-API path. It must survive."""
assert BackendResolver(None).for_request(_Req()) is None
def test_a_native_provider_does_not_switch_backends():
"""Anthropic is the shape the proxy already holds, so a model rewrite is
enough and the extension has already done it."""
r = BackendResolver(DEFAULT)
req = _Req(headroom_route=SimpleNamespace(model="claude-haiku-4-5", provider="anthropic"))
assert r.for_request(req) is DEFAULT
def test_a_bare_anthropic_model_resolves_its_provider_and_stays_put():
r = BackendResolver(DEFAULT)
req = _Req(headroom_route=SimpleNamespace(model="claude-haiku-4-5"))
assert r.for_request(req) is DEFAULT
# --- switching, and refusing to switch --------------------------------------
def test_a_foreign_provider_gets_its_own_backend(monkeypatch):
built = []
class FakeBackend:
def __init__(self, provider):
built.append(provider)
self.provider = provider
monkeypatch.setattr(BackendResolver, "_build", lambda self, p: FakeBackend(p))
r = BackendResolver(DEFAULT)
body = {"model": "claude-opus-5"}
req = _Req(headroom_route=SimpleNamespace(model="moonshot/kimi-k2", provider="moonshot"))
got = r.for_request(req, body=body)
assert isinstance(got, FakeBackend) and got.provider == "moonshot"
# The extension could not safely write a foreign model id; we do it.
assert body["model"] == "moonshot/kimi-k2"
def test_backends_are_built_once_per_provider(monkeypatch):
built = []
monkeypatch.setattr(BackendResolver, "_build", lambda self, p: built.append(p) or object())
r = BackendResolver(DEFAULT)
req = _Req(headroom_route=SimpleNamespace(model="x", provider="moonshot"))
for _ in range(5):
r.for_request(req)
assert built == ["moonshot"], "construction is expensive; cache it"
def test_a_backend_that_will_not_build_falls_back_and_stops_retrying(monkeypatch):
calls = []
monkeypatch.setattr(BackendResolver, "_build", lambda self, p: calls.append(p) or None)
r = BackendResolver(DEFAULT)
req = _Req(headroom_route=SimpleNamespace(model="x", provider="nope"))
for _ in range(5):
assert r.for_request(req) is DEFAULT
assert calls == ["nope"], "a broken provider must not be retried per request"
def test_a_routing_preference_can_never_take_traffic_down(monkeypatch):
"""Missing credentials, missing optional dependency -- whatever the reason,
the request still has to be served."""
def boom(self, provider):
raise RuntimeError("no credentials")
monkeypatch.setattr(BackendResolver, "_build", boom)
r = BackendResolver(DEFAULT)
req = _Req(headroom_route=SimpleNamespace(model="x", provider="moonshot"))
assert r.for_request(req) is DEFAULT
def test_an_unknown_provider_is_rejected_at_resolve_time():
"""LiteLLMBackend accepts ANY provider string -- the registry falls through
to a generic pass-through -- so a typo silently builds a backend that only
fails later, at request time, with an error pointing nowhere near the typo.
Validate the name up front instead."""
r = BackendResolver(DEFAULT)
assert r._build("definitely-not-a-provider-name") is None
req = _Req(headroom_route=SimpleNamespace(model="x", provider="definitely-not-a-provider-name"))
assert r.for_request(req) is DEFAULT
def test_a_real_provider_name_is_accepted():
assert BackendResolver(DEFAULT)._build("moonshot") is not None
# --- streaming, which is the path agents actually take ----------------------
class _Backend:
"""Records that it, and not some other backend, served the request."""
def __init__(self, name):
self.name = name
self.served = False
async def stream_message(self, body, headers):
self.served = True
return
yield # pragma: no cover -- makes this an async generator
async def stream_openai_message(self, body, headers):
self.served = True
return
yield # pragma: no cover
class _FailingBackend(_Backend):
async def stream_message(self, body, headers):
self.served = True
yield SimpleNamespace(
event_type="message_start",
raw_sse="event: message_start\ndata: {}\n\n",
data={},
)
raise RuntimeError("stream finalization failed")
class _ErrorEventBackend(_Backend):
async def stream_message(self, body, headers):
self.served = True
yield SimpleNamespace(
event_type="error",
raw_sse='event: error\ndata: {"error": {"type": "api_error"}}\n\n',
data={"error": {"type": "api_error"}},
)
async def stream_openai_message(self, body, headers):
self.served = True
yield 'data: {"type": "er'
yield 'ror", "message": "provider failed"}\n\n'
yield "data: [DONE]\n\n"
async def _drive_with_outcome(handler, backend):
from headroom.proxy.handlers.streaming import StreamingMixin
response = await StreamingMixin._stream_response_bedrock(
handler,
{"messages": []},
{},
"anthropic",
"m",
"rid",
0,
0,
0,
[],
{},
0.0,
backend=backend,
)
async for _ in response.body_iterator:
pass
async def _drive_openai_with_outcome(handler, backend):
from headroom.proxy.handlers.streaming import StreamingMixin
response = await StreamingMixin._stream_openai_via_backend(
handler,
{"messages": []},
{},
"m",
"rid",
0.0,
0,
0,
0,
[],
{},
0.0,
backend=backend,
)
async for _ in response.body_iterator:
pass
async def _drive(handler, **kw):
from headroom.proxy.handlers.streaming import StreamingMixin
resp = await StreamingMixin._stream_response_bedrock(
handler,
{"messages": []},
{},
"anthropic",
"m",
"rid",
0,
0,
0,
[],
{},
0.0,
**kw,
)
async for _ in resp.body_iterator:
pass
class _Config:
"""Every proxy flag off. Named individually the list would rot; the test
is about which backend served the request, not about config."""
def __getattr__(self, name):
return False
def _handler(default):
return SimpleNamespace(
anthropic_backend=default,
config=_Config(),
_record_request_outcome=lambda outcome: _noop(),
)
async def _noop():
return None
def test_streaming_honours_the_routed_backend():
"""The non-streaming branch was the easy half. body["model"] has already
been rewritten to a foreign id by the time we get here, so streaming to
the configured backend would send e.g. moonshot/kimi-k2 to Anthropic."""
configured, routed = _Backend("anthropic"), _Backend("moonshot")
asyncio.run(_drive(_handler(configured), backend=routed))
assert routed.served and not configured.served
def test_streaming_without_a_route_uses_the_configured_backend():
configured = _Backend("anthropic")
asyncio.run(_drive(_handler(configured)))
assert configured.served
def test_failed_stream_finalization_is_not_counted_as_completed():
backend = _FailingBackend("anthropic")
outcomes = []
async def record_outcome(outcome):
outcomes.append(outcome)
handler = _handler(backend)
handler._extract_anthropic_cache_ttl_metrics = lambda usage: (0, 0)
handler._record_request_outcome = record_outcome
asyncio.run(_drive_with_outcome(handler, backend))
assert len(outcomes) == 1
assert outcomes[0].status_code == 502
def test_stream_outcome_status_preserves_success_and_reclassifies_failure():
from headroom.proxy.handlers.streaming import _stream_outcome_status
assert _stream_outcome_status(200, True) == 200
assert _stream_outcome_status(200, False) == 502
assert _stream_outcome_status(529, False) == 502
def test_sse_error_detection_handles_native_and_split_events():
assert _sse_contains_error_event(b'event: error\ndata: {"type":"error"}\n\n')
assert _sse_contains_error_event(b'data: {"type":"er' + b'ror"}\n\n')
assert not _sse_contains_error_event(b'data: {"type":"message"}\n\n')
def test_sse_error_text_in_response_content_is_not_an_error():
assert not _sse_contains_error_event(b'data: {"text":"event: error"}\n\n')
def test_responses_failed_envelope_is_an_error():
assert _sse_contains_error_event(
b'data: {"type":"response.failed","response":{"status":"failed",'
b'"error":{"code":"server_error"}}}\n\n'
)
@pytest.mark.parametrize("provider", ["anthropic", "openai"])
@pytest.mark.parametrize("complete", [False, True])
def test_backend_clean_eof_requires_protocol_completion(provider, complete):
from types import MethodType
from headroom.proxy.handlers.streaming import StreamingMixin
class Backend(_Backend):
async def stream_message(self, body, headers):
yield SimpleNamespace(
event_type="message_start", raw_sse="event: message_start\ndata: {}\n\n", data={}
)
if complete:
yield SimpleNamespace(
event_type="message_stop",
raw_sse='event: message_stop\ndata: {"type":"message_stop"}\n\n',
data={"type": "message_stop"},
)
async def stream_openai_message(self, body, headers):
yield 'data: {"choices":[{"delta":{"content":"partial"}}]}\n\n'
if complete:
yield "data: [DONE]\n\n"
outcomes = []
async def record(outcome):
outcomes.append(outcome)
handler = _handler(Backend(provider))
handler._record_request_outcome = record
handler._extract_anthropic_cache_ttl_metrics = lambda usage: (0, 0)
handler._parse_sse_usage_from_buffer = MethodType(
StreamingMixin._parse_sse_usage_from_buffer, handler
)
drive = _drive_with_outcome if provider == "anthropic" else _drive_openai_with_outcome
asyncio.run(drive(handler, handler.anthropic_backend))
assert len(outcomes) == 1
assert outcomes[0].status_code == (200 if complete else 502)
@pytest.mark.parametrize("provider", ["anthropic", "openai"])
def test_backend_client_hangup_after_terminal_event_is_completed(provider):
"""A client may close on the terminal event, before the backend's EOF."""
from types import MethodType
from headroom.proxy.handlers.streaming import StreamingMixin
class Backend(_Backend):
async def stream_message(self, body, headers):
yield SimpleNamespace(
event_type="message_stop",
raw_sse='event: message_stop\ndata: {"type":"message_stop"}\n\n',
data={"type": "message_stop"},
)
await asyncio.Event().wait()
async def stream_openai_message(self, body, headers):
yield "data: [DONE]\n\n"
await asyncio.Event().wait()
outcomes = []
async def record(outcome):
outcomes.append(outcome)
handler = _handler(Backend(provider))
handler._record_request_outcome = record
handler._extract_anthropic_cache_ttl_metrics = lambda usage: (0, 0)
handler._parse_sse_usage_from_buffer = MethodType(
StreamingMixin._parse_sse_usage_from_buffer, handler
)
async def drive():
arguments = {
"body": {"messages": []},
"headers": {},
"model": "m",
"request_id": "rid",
"original_tokens": 0,
"optimized_tokens": 0,
"tokens_saved": 0,
"transforms_applied": [],
"tags": {},
"optimization_latency": 0.0,
"backend": handler.anthropic_backend,
}
if provider == "anthropic":
response = await StreamingMixin._stream_response_bedrock(
handler, provider=provider, **arguments
)
else:
response = await StreamingMixin._stream_openai_via_backend(
handler, start_time=0.0, **arguments
)
body = response.body_iterator
while b"message_stop" not in (chunk := await body.__anext__()) and b"[DONE]" not in chunk:
pass
await body.aclose()
asyncio.run(drive())
assert len(outcomes) == 1
assert outcomes[0].status_code == 200
@pytest.mark.parametrize("provider", ["anthropic", "openai"])
@pytest.mark.parametrize("complete", [False, True])
def test_backend_prefix_tracker_advances_only_after_completion(provider, complete):
from types import MethodType
from unittest.mock import MagicMock
from headroom.proxy.handlers.streaming import StreamingMixin
class Backend(_Backend):
async def stream_message(self, body, headers):
yield SimpleNamespace(event_type="message_start", raw_sse="", data={})
if complete:
yield SimpleNamespace(event_type="message_stop", raw_sse="", data={})
async def stream_openai_message(self, body, headers):
yield 'data: {"choices":[{"delta":{"content":"partial"}}]}\n\n'
if complete:
yield "data: [DONE]\n\n"
backend = Backend(provider)
handler = _handler(backend)
handler._extract_anthropic_cache_ttl_metrics = lambda usage: (0, 0)
handler._parse_sse_usage_from_buffer = MethodType(
StreamingMixin._parse_sse_usage_from_buffer, handler
)
handler._parse_sse_to_response = lambda *args: None
handler._assistant_message_from_response_json = lambda response: None
tracker = MagicMock()
tracker.classify_cache_miss.return_value = SimpleNamespace(is_miss=False)
async def drive():
arguments = {
"body": {"messages": []},
"headers": {},
"model": "m",
"request_id": "rid",
"original_tokens": 0,
"optimized_tokens": 0,
"tokens_saved": 0,
"transforms_applied": [],
"tags": {},
"optimization_latency": 0.0,
"backend": backend,
"prefix_tracker": tracker,
}
if provider == "anthropic":
response = await StreamingMixin._stream_response_bedrock(
handler, provider=provider, **arguments
)
else:
response = await StreamingMixin._stream_openai_via_backend(
handler, start_time=0.0, **arguments
)
async for _ in response.body_iterator:
pass
asyncio.run(drive())
assert tracker.update_from_response.call_count == int(complete)
@pytest.mark.parametrize(
"event, data, expected_failed, expected_terminal",
[
("response.completed", '{"type":"response.completed"}', False, True),
("response.incomplete", '{"type":"response.incomplete"}', False, True),
("response.failed", '{"type":"response.failed"}', True, False),
(None, "[DONE]", False, False),
],
)
def test_responses_terminal_semantics(event, data, expected_failed, expected_terminal):
from headroom.proxy.handlers.streaming import _sse_event_outcome
assert _sse_event_outcome(event, data, "responses") == (expected_failed, expected_terminal)
@pytest.mark.parametrize("event_type", ['["error"]', '{"unexpected":true}', "7", "true"])
def test_malformed_event_type_is_not_interpreted_as_a_protocol_marker(event_type):
payload = ('data: {"type":' + event_type + "}\n\ndata: [DONE]\n\n").encode()
assert not _sse_contains_error_event(payload)
@pytest.mark.parametrize("wrapped", [False, True])
@pytest.mark.parametrize("finish_reason", ["STOP", "MAX_TOKENS"])
def test_gemini_finish_reason_is_a_terminal_event(wrapped, finish_reason):
import json
from headroom.proxy.handlers.streaming import _sse_event_outcome
payload = {"candidates": [{"index": 0, "finishReason": finish_reason}]}
if wrapped:
payload = {"response": payload}
assert _sse_event_outcome(None, json.dumps(payload), "gemini") == (False, True)
@pytest.mark.parametrize("wrapped", [False, True])
def test_gemini_prompt_block_is_a_failed_outcome(wrapped):
import json
from headroom.proxy.handlers.streaming import _sse_event_outcome
payload = {"promptFeedback": {"blockReason": "SAFETY"}}
if wrapped:
payload = {"response": payload}
failed, _ = _sse_event_outcome(None, json.dumps(payload), "gemini")
assert failed
def test_gemini_completion_requires_all_requested_candidates():
from headroom.proxy.handlers.streaming import StreamingMixin
parser = StreamingMixin()
state = {"sse_buffer": bytearray(), "gemini_candidate_count": 2}
state["sse_buffer"].extend(b'data: {"candidates":[{"index":0,"finishReason":"STOP"}]}\n\n')
parser._parse_sse_usage_from_buffer(state, "gemini")
assert not state.get("stream_terminal", False)
state["sse_buffer"].extend(b'data: {"candidates":[{"index":1,"finishReason":"STOP"}]}\n\n')
parser._parse_sse_usage_from_buffer(state, "gemini")
assert state["stream_terminal"]
def test_outcome_classification_runs_once_for_a_chunked_complete_event(monkeypatch):
from headroom.proxy.handlers import streaming
seen = []
classify = streaming._sse_event_outcome
def record(*args):
seen.append(args)
return classify(*args)
monkeypatch.setattr(streaming, "_sse_event_outcome", record)
state = {"sse_buffer": bytearray()}
parser = streaming.StreamingMixin()
payload = b'data: {"text":"' + b"x" * 10_000 + b'"}\n\n'
for offset in range(0, len(payload), 101):
state["sse_buffer"].extend(payload[offset : offset + 101])
parser._parse_sse_usage_from_buffer(state, "openai")
assert len(seen) == 1
assert not state["sse_buffer"]
def test_backend_error_event_is_not_counted_as_completed():
backend = _ErrorEventBackend("anthropic")
outcomes = []
async def record_outcome(outcome):
outcomes.append(outcome)
handler = _handler(backend)
handler._extract_anthropic_cache_ttl_metrics = lambda usage: (0, 0)
handler._record_request_outcome = record_outcome
asyncio.run(_drive_with_outcome(handler, backend))
assert backend.served
assert len(outcomes) == 1
assert outcomes[0].status_code == 502
def test_openai_backend_error_event_is_not_counted_as_completed():
backend = _ErrorEventBackend("openai")
outcomes = []
async def record_outcome(outcome):
outcomes.append(outcome)
handler = _handler(backend)
handler._parse_sse_usage_from_buffer = lambda state, provider: None
handler._record_request_outcome = record_outcome
asyncio.run(_drive_openai_with_outcome(handler, backend))
assert backend.served
assert len(outcomes) == 1
assert outcomes[0].status_code == 502
async def _drive_openai(handler, **kw):
from headroom.proxy.handlers.streaming import StreamingMixin
resp = await StreamingMixin._stream_openai_via_backend(
handler,
{"messages": []},
{},
"m",
"rid",
0.0,
0,
0,
0,
[],
{},
0.0,
**kw,
)
async for _ in resp.body_iterator:
pass
def test_openai_streaming_honours_the_routed_backend():
"""opencode and pi can speak either protocol, so the OpenAI chat path
needs the same treatment as the Anthropic one."""
configured, routed = _Backend("openai"), _Backend("moonshot")
asyncio.run(_drive_openai(_handler(configured), backend=routed))
assert routed.served and not configured.served
def test_openai_streaming_without_a_route_uses_the_configured_backend():
configured = _Backend("openai")
asyncio.run(_drive_openai(_handler(configured)))
assert configured.served
def test_the_resolver_follows_a_reassigned_default():
class H:
anthropic_backend = None
h = H()
from headroom.proxy.route_advice import BackendResolver as BR
first = BR(h.anthropic_backend)
assert first.default is None
h.anthropic_backend = DEFAULT
assert BR(h.anthropic_backend).default is DEFAULT
@pytest.mark.parametrize(("written", "inferred"), [(0, False), (20, False), (None, True)])
def test_backend_openai_stream_reads_reported_cache_writes(written, inferred):
"""A reported write, zero included, is the provider's word; only an absent
field falls back to inferring writes from the uncached slice."""
from types import MethodType
from headroom.proxy.handlers.streaming import StreamingMixin
details = {"cached_tokens": 75}
if written is not None:
details["cache_write_tokens"] = written
usage = {"input_tokens": 100, "output_tokens": 5, "input_tokens_details": details}
class Backend(_Backend):
async def stream_openai_message(self, body, headers):
yield f"data: {json.dumps({'type': 'response.completed', 'response': {'usage': usage}})}\n\n"
yield "data: [DONE]\n\n"
outcomes = []
async def record(outcome):
outcomes.append(outcome)
handler = _handler(Backend("openai"))
handler._record_request_outcome = record
handler._parse_sse_usage_from_buffer = MethodType(
StreamingMixin._parse_sse_usage_from_buffer, handler
)
asyncio.run(_drive_openai_with_outcome(handler, handler.anthropic_backend))
o = outcomes[0]
assert o.cache_inferred is inferred
assert o.cache_write_tokens == (25 if written is None else written)