scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-27 06:13:07 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 72
litellm/tests/integration/_support/hook_counts.py
yucheng-berri dd511a2b22
fix(logging)!: fire one CustomLogger failure hook per async call outside the async flag list (#47124) 
* fix(logging): fire one CustomLogger hook per async call outside the async flag list

aspeech, atranscription, aimage_generation, aimage_edit, aocr and asearch are
missing from is_sync_litellm_request's flag list, so the sync success and
failure handlers called a CustomLogger's sync hook on top of the async hook the
async handler had already called for the same request.

The sync handlers now skip a CustomLogger's log_success_event and
log_failure_event when that run is paired with an async handler run
(for_async_request) and the async handler gives the logger its own async hook.
The async client wrapper, dispatch_failure_handlers, the proxy-only failure
dispatch and the native bridge's sync failure step for an async call mark their
sync failure run as paired.

Resolves LIT-9486

* fix(logging): keep the sync CustomLogger hook when no async success runs

The async client wrapper skips async success logging for calls carrying a
fallbacks kwarg, and the native bridge skips it for internal and fallbacks
calls, yet the paired sync success run still dropped a both-hooks logger's
sync hook, so that logger got no success event at all.

The sync success run now suppresses the sync hook only when the async success
handler was actually scheduled for the call: the client wrapper passes
async_handler_runs=not is_completion_with_fallbacks, and the native host
records per call whether it enqueued or deferred the async success.

Tests that pinned the doubled failure hook on native aocr, messages and
responses calls and on proxy-rejected requests now expect the async hook
alone. The integration test adds /v1/messages and /v1/responses cells and
runs the key-auth rejection on every route.

* fix(logging): type dispatch_success_handlers kwargs

The new async_handler_runs keyword on success_handler made the untyped
**kwargs that dispatch_success_handlers forwards an unknown argument for it,
which grew the reportUnknownArgumentType count past its merge-base ceiling.
Typing the forwarded kwargs as Any, as _success_handler_body already does,
removes those diagnostics without changing what is forwarded.

* test(integration): cover CustomLogger hook counts across modes, chaos and router strategies

Moves the hook count rig into tests/integration/_support/hook_counts.py and
extends the cells it drives:

- streamed chat, Messages and Responses requests, success and failure
- loggers the async handler never reaches (sync-only overrides, sync event
  lists only) keep one sync hook per request on every surface outside the
  async flag list, key-auth rejections included
- a concurrent burst while the logger's destination stalls and then fails,
  and a burst while a worker is killed, never give a request two hooks
- least-busy keeps counting a request in flight while async speech requests
  finish around it, and usage-based routing v2 counts an async image
  request's tokens once

* test(integration): Datadog and LangSmith deliver one record per async request

Points the built-in Datadog, LangSmith and OpenTelemetry loggers at a local
sink and checks every async surface outside the flag list: one Datadog log
entry and one LangSmith run per request (two on the merge base), and never two
OpenTelemetry spans, which its own emit-once guard already ensured.

* fix(logging): give a both-hooks logger its sync hook when the async run never reaches it

The sync hook was skipped whenever the logger sat in the async callback list, so a
logger the async run never reached got no hook at all: a sibling stalled past the
logging worker timeout, a batch poll that skips default logging, or an async
failure run cut off. The sync run now holds the hook and the async run hands it
back unless it reached the logger

* fix(logging): run the held sync success hooks when the proxy drops a blocked response's success

A post-call guardrail block drops the parked async success, which left the sync
success hook held for it unrun. Base gave a both-hooks logger that hook on these
call types, so the drop now ends the pairing and the held hook runs as before

* fix(logging): run held sync hooks only after every sync run of the request has closed

A held hook released by the async run could run on another thread while the sync
run was still walking its other callbacks. The pairing now waits until the async
run has ended and every sync run has closed, and the last sync run to close runs
the held hooks itself

* fix(logging): end the async run when its coroutine is dropped before it starts

A logging queue that drops the async success coroutine, or a deferred success
that fails to schedule, left the sync hooks held for it unrun. The enqueued async
run now ends when its coroutine is discarded, and the proxy ends it when firing
the deferred success raises

* fix(logging): leave a still-queued async run to the exit flush

The discard finalizer ran at interpreter exit before the logging worker's exit
flush, so a queued request could get its held sync hook and then its async hook

* test: keep the router and file-stream tests in step with main

Main gives each owned proxy a private Redis by default, so the router cells
share the worker's Redis to read its counters, and the file-stream stub takes
the hook pairing the stream now passes

* fix(logging): run held sync hooks when a during_call guardrail raises or the exit flush gives up

A during_call guardrail that raised after the LLM call parked its deferred
success skipped the proxy's flush, and async runs the exit flush never started
were never ended, so a both-hooks logger lost its sync hook in both cases

* fix(logging): skip the sync failure hook when the async failure run logs to the logger

An async call outside is_sync_litellm_request's flag list ran both failure
handlers, so a CustomLogger with both failure hooks got both. The sync run
of an async request now skips log_failure_event for a logger that overrides
async_log_failure_event and is registered for async failures. The four sync
runs of async failures (dispatch_failure_handlers, the @client async wrapper,
the proxy-only failure thread and the Rust host) pass for_async_request=True.

Drops the HookPairing fallback. Successes are left to #47192, which already
gives such a logger only async_log_success_event.

---------

Co-authored-by: yucheng-berri <293926326+yucheng-berri@users.noreply.github.com>
2026-10-20 15:19:14 -07:00

329 lines
14 KiB
Python
Raw Blame History

"""Scripted provider, proxy config and request helpers shared by the CustomLogger hook count tests."""
import base64
import json
import threading
from collections.abc import Callable, Mapping
from dataclasses import dataclass, field
from pathlib import Path
from typing import Final
import httpx
import yaml
from integration._support.client import Gateway, held
from integration._support.wire import Reply, Request, Wire
WAV_BYTES: Final = (
b"RIFF\x30\x00\x00\x00WAVEfmt \x10\x00\x00\x00\x01\x00\x01\x00"
b"\x40\x1f\x00\x00\x40\x1f\x00\x00\x01\x00\x08\x00data\x08\x00\x00\x00"
b"\x00\x01\x02\x03\x04\x05\x06\x07"
)
PNG_BYTES: Final = (
b"\x89PNG\r\n\x1a\n\x00\x00\x00\rIHDR\x00\x00\x00\x01\x00\x00\x00\x01\x08\x06\x00\x00\x00"
b"\x1f\x15\xc4\x89\x00\x00\x00\rIDAT\x08\xd7c\xf8\xcf\xc0\xf0\x1f\x00\x05\x00\x01\xff"
b"\x89\x99=\x1d\x00\x00\x00\x00IEND\xaeB`\x82"
)
_IMAGE_REPLY: Final = json.dumps({"created": 1700000000, "data": [{"b64_json": base64.b64encode(PNG_BYTES).decode()}]})
_CHAT_REPLY: Final = json.dumps(
{
"id": "chatcmpl-hook-count",
"object": "chat.completion",
"created": 1,
"model": "gpt-4o-mini",
"choices": [{"index": 0, "message": {"role": "assistant", "content": "ok"}, "finish_reason": "stop"}],
"usage": {"prompt_tokens": 3, "completion_tokens": 1, "total_tokens": 4},
}
)
_EMBEDDING_REPLY: Final = json.dumps(
{
"object": "list",
"model": "text-embedding-3-small",
"data": [{"object": "embedding", "index": 0, "embedding": [0.25, 0.5]}],
"usage": {"prompt_tokens": 3, "total_tokens": 3},
}
)
_MESSAGE_REPLY: Final = json.dumps(
{
"id": "msg_hook_count",
"type": "message",
"role": "assistant",
"model": "claude-haiku-4-5",
"content": [{"type": "text", "text": "ok"}],
"stop_reason": "end_turn",
"stop_sequence": None,
"usage": {"input_tokens": 3, "output_tokens": 1},
}
)
_RESPONSE_REPLY: Final = json.dumps(
{
"id": "resp_hook_count",
"object": "response",
"created_at": 1,
"status": "completed",
"model": "gpt-4o-mini",
"output": [
{
"type": "message",
"id": "msg_hook_count",
"status": "completed",
"role": "assistant",
"content": [{"type": "output_text", "text": "ok", "annotations": []}],
}
],
"usage": {"input_tokens": 3, "output_tokens": 1, "total_tokens": 4},
}
)
_OCR_REPLY: Final = json.dumps(
{
"pages": [
{"index": 0, "markdown": "page", "images": [], "dimensions": {"dpi": 200, "height": 1100, "width": 850}}
],
"model": "mistral-ocr-latest",
"usage_info": {"pages_processed": 1, "doc_size_bytes": 64},
"document_annotation": None,
}
)
_SEARCH_REPLY: Final = json.dumps(
{"results": [{"title": "hook count", "url": "https://result.invalid/a", "snippet": "hit"}]}
)
_CHAT_FRAMES: Final = (
'data: {"id":"chatcmpl-hook-count","object":"chat.completion.chunk","created":1,"model":"gpt-4o-mini",'
'"choices":[{"index":0,"delta":{"role":"assistant","content":"ok"},"finish_reason":null}]}',
'data: {"id":"chatcmpl-hook-count","object":"chat.completion.chunk","created":1,"model":"gpt-4o-mini",'
'"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],'
'"usage":{"prompt_tokens":3,"completion_tokens":1,"total_tokens":4}}',
"data: [DONE]",
)
_MESSAGE_FRAMES: Final = (
"event: message_start\n"
'data: {"type":"message_start","message":{"id":"msg_hook_count","type":"message","role":"assistant",'
'"model":"claude-haiku-4-5","content":[],"stop_reason":null,"stop_sequence":null,'
'"usage":{"input_tokens":3,"output_tokens":0}}}',
'event: content_block_start\ndata: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""}}',
"event: content_block_delta\n"
'data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"ok"}}',
'event: content_block_stop\ndata: {"type":"content_block_stop","index":0}',
"event: message_delta\n"
'data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"output_tokens":1}}',
'event: message_stop\ndata: {"type":"message_stop"}',
)
_RESPONSE_FRAMES: Final = (
"event: response.created\n"
'data: {"type":"response.created","sequence_number":0,"response":{"id":"resp_hook_count","object":"response",'
'"created_at":1,"status":"in_progress","model":"gpt-4o-mini","output":[],"usage":null}}',
"event: response.output_text.delta\n"
'data: {"type":"response.output_text.delta","sequence_number":1,"item_id":"msg_hook_count","output_index":0,'
'"content_index":0,"delta":"ok"}',
"event: response.completed\n"
'data: {"type":"response.completed","sequence_number":2,"response":{"id":"resp_hook_count","object":"response",'
'"created_at":1,"status":"completed","model":"gpt-4o-mini","output":[{"type":"message","id":"msg_hook_count",'
'"status":"completed","role":"assistant","content":[{"type":"output_text","text":"ok","annotations":[]}]}],'
'"usage":{"input_tokens":3,"output_tokens":1,"total_tokens":4}}}',
)
_STREAM_FRAMES: Final = {
"/v1/chat/completions": _CHAT_FRAMES,
"/v1/messages": _MESSAGE_FRAMES,
"/v1/responses": _RESPONSE_FRAMES,
}
SURFACE_CALL_TYPES: Final = {
"chat": "acompletion",
"embedding": "aembedding",
"messages": "anthropic_messages",
"responses": "aresponses",
"speech": "aspeech",
"transcription": "atranscription",
"image_generation": "aimage_generation",
"image_edit": "aimage_edit",
"ocr": "aocr",
"search": "asearch",
}
IN_THE_ASYNC_FLAG_LIST: Final = frozenset({"chat", "embedding", "messages", "responses"})
STREAMING_SURFACES: Final = ("chat", "messages", "responses")
def _streams(request: Request) -> bool:
if request.headers.get("content-type", "").startswith("application/json"):
return json.loads(request.body or b"{}").get("stream") is True
return False
def provider(request: Request) -> Reply:
"""Answers every surface the hook count tests drive; any target under /fail/ is a provider 400."""
if request.method == "GET":
return Reply(body=b'{"object":"list","data":[]}')
if request.target.startswith("/fail/"):
return Reply(
status=400,
body=b'{"error":{"message":"scripted provider rejection","type":"invalid_request_error","code":null}}',
)
if request.target in _STREAM_FRAMES and _streams(request):
return Reply(
chunks=tuple(f"{frame}\n\n".encode() for frame in _STREAM_FRAMES[request.target]),
content_type="text/event-stream",
)
replies: Final[dict[str, Callable[[], Reply]]] = {
"/v1/chat/completions": lambda: Reply(body=_CHAT_REPLY.encode()),
"/v1/embeddings": lambda: Reply(body=_EMBEDDING_REPLY.encode()),
"/v1/messages": lambda: Reply(body=_MESSAGE_REPLY.encode()),
"/v1/responses": lambda: Reply(body=_RESPONSE_REPLY.encode()),
"/v1/audio/speech": lambda: Reply(body=WAV_BYTES, content_type="audio/wav"),
"/v1/audio/transcriptions": lambda: Reply(body=b'{"text":"hook count transcript"}'),
"/v1/images/generations": lambda: Reply(body=_IMAGE_REPLY.encode()),
"/v1/images/edits": lambda: Reply(body=_IMAGE_REPLY.encode()),
"/v1/ocr": lambda: Reply(body=_OCR_REPLY.encode()),
"/perplexity/search": lambda: Reply(body=_SEARCH_REPLY.encode()),
}
return replies[request.target]()
@dataclass
class HookRecords:
"""Every hook record the probe reported, kept across drains so cells can read their own call id."""
wire: Wire
function_callbacks: bool
seen: list[dict[str, str]] = field(default_factory=list)
lock: threading.Lock = field(default_factory=threading.Lock)
def drained(self) -> tuple[dict[str, str], ...]:
with self.lock:
self.seen.extend(json.loads(request.body) for request in self.wire.drain())
return tuple(self.seen)
def for_call(self, call_id: str) -> tuple[dict[str, str], ...]:
return tuple(record for record in self.drained() if record["litellm_call_id"] == call_id)
def _model(name: str, model: str, api_base: str, **extra: object) -> dict[str, object]:
return {
"model_name": name,
"litellm_params": {"model": model, "api_base": api_base, "api_key": "synthetic-provider-key"},
**extra,
}
def write_hook_count_config(
path: Path,
provider_url: str,
litellm_settings: Mapping[str, object],
guardrails: tuple[Mapping[str, object], ...] = (),
) -> Path:
"""One success and one failure deployment per surface, all answered by `provider`."""
models: Final = []
tools: Final = []
for outcome, prefix in (("success", ""), ("failure", "/fail")):
base: str = provider_url + prefix
models.extend(
(
_model(f"hooks-chat-{outcome}", "openai/gpt-4o-mini", base + "/v1"),
_model(f"hooks-embedding-{outcome}", "openai/text-embedding-3-small", base + "/v1"),
_model(f"hooks-messages-{outcome}", "anthropic/claude-haiku-4-5", base),
_model(f"hooks-responses-{outcome}", "openai/gpt-4o-mini", base + "/v1"),
_model(f"hooks-speech-{outcome}", "openai/gpt-4o-mini-tts", base + "/v1"),
_model(f"hooks-transcription-{outcome}", "openai/whisper-1", base + "/v1"),
_model(f"hooks-image_generation-{outcome}", "openai/gpt-image-1", base + "/v1"),
_model(f"hooks-image_edit-{outcome}", "openai/gpt-image-1", base + "/v1"),
_model(f"hooks-ocr-{outcome}", "mistral/mistral-ocr-latest", base + "/v1", model_info={"mode": "ocr"}),
)
)
tools.append(
{
"search_tool_name": f"hooks-search-{outcome}",
"litellm_params": {
"search_provider": "perplexity",
"api_key": "synthetic-provider-key",
"api_base": base + "/perplexity",
},
}
)
path.write_text(
yaml.safe_dump(
{
"model_list": models,
"search_tools": tools,
"litellm_settings": dict(litellm_settings),
**({"guardrails": [dict(guardrail) for guardrail in guardrails]} if guardrails else {}),
"general_settings": {
"master_key": "os.environ/LITELLM_MASTER_KEY",
"database_url": "os.environ/DATABASE_URL",
"disable_model_info_refresh": True,
},
"router_settings": {"num_retries": 0},
}
)
)
return path
def send(
proxy: Gateway, surface: str, outcome: str, call_id: str, *, key: str | None = None, stream: bool = False
) -> httpx.Response:
model: Final = f"hooks-{surface}-{outcome}"
nonce: Final = f"hook-count-{call_id}"
headers: Final = {"Authorization": f"Bearer {key or proxy.key}", "x-litellm-call-id": call_id}
if surface == "transcription":
return proxy.client.post(
"/v1/audio/transcriptions",
data={"model": model, "prompt": nonce},
files={"file": ("a.wav", WAV_BYTES + call_id.encode(), "audio/wav")},
headers=headers,
)
if surface == "image_edit":
return proxy.client.post(
"/v1/images/edits",
data={"model": model, "prompt": nonce},
files={"image": ("a.png", PNG_BYTES, "image/png")},
headers=headers,
)
streaming: Final[dict[str, object]] = {"stream": True} if stream else {}
routes: Final[dict[str, tuple[str, dict[str, object]]]] = {
"chat": (
"/v1/chat/completions",
{"model": model, "messages": [{"role": "user", "content": nonce}], **streaming},
),
"embedding": ("/v1/embeddings", {"model": model, "input": nonce}),
"messages": (
"/v1/messages",
{"model": model, "max_tokens": 16, "messages": [{"role": "user", "content": nonce}], **streaming},
),
"responses": ("/v1/responses", {"model": model, "input": nonce, **streaming}),
"speech": ("/v1/audio/speech", {"model": model, "input": nonce, "voice": "alloy", "response_format": "wav"}),
"image_generation": ("/v1/images/generations", {"model": model, "prompt": nonce, "n": 1}),
"ocr": (
"/v1/ocr",
{
"model": model,
"document": {
"type": "document_url",
"document_url": "data:application/pdf;base64," + base64.b64encode(nonce.encode()).decode(),
},
},
),
"search": ("/v1/search", {"search_tool_name": f"hooks-search-{outcome}", "query": nonce}),
}
path, body = routes[surface]
return proxy.client.post(path, json=body, headers=headers)
def counts_function_hooks(records: HookRecords, surface: str) -> bool:
return records.function_callbacks and surface not in IN_THE_ASYNC_FLAG_LIST
def probe_hooks(records: HookRecords, surface: str, found: tuple[dict[str, str], ...]) -> list[str]:
counted: Final = counts_function_hooks(records, surface)
return sorted(record["hook"] for record in found if counted or not record["hook"].startswith("function_"))
def one_async_hook(records: HookRecords, surface: str, outcome: str) -> list[str]:
function_hook: Final = (f"function_{outcome}",) if counts_function_hooks(records, surface) else ()
return sorted((f"async_{outcome}", *function_hook))
def assert_hooks(records: HookRecords, surface: str, call_id: str, expected: list[str]) -> None:
settled: Final = held(
lambda: records.for_call(call_id),
lambda found: probe_hooks(records, surface, found) == expected,
holding=3,
seconds=20,
)
assert {record["call_type"] for record in settled} == {SURFACE_CALL_TYPES[surface]}, settled
Reference in a new issue View git blame Copy permalink