Merge branch 'main' into gl/fix/task-planning-order

This commit is contained in:
Greyson LaLonde
2025-11-30 19:46:04 -05:00
committed by GitHub
4 changed files with 386 additions and 0 deletions

View File

@@ -71,6 +71,7 @@ from crewai.events.types.reasoning_events import (
AgentReasoningFailedEvent,
AgentReasoningStartedEvent,
)
from crewai.events.types.system_events import SignalEvent, on_signal
from crewai.events.types.task_events import (
TaskCompletedEvent,
TaskFailedEvent,
@@ -159,6 +160,7 @@ class TraceCollectionListener(BaseEventListener):
self._register_flow_event_handlers(crewai_event_bus)
self._register_context_event_handlers(crewai_event_bus)
self._register_action_event_handlers(crewai_event_bus)
self._register_system_event_handlers(crewai_event_bus)
self._listeners_setup = True
@@ -458,6 +460,15 @@ class TraceCollectionListener(BaseEventListener):
) -> None:
self._handle_action_event("knowledge_query_failed", source, event)
def _register_system_event_handlers(self, event_bus: CrewAIEventsBus) -> None:
"""Register handlers for system signal events (SIGTERM, SIGINT, etc.)."""
@on_signal
def handle_signal(source: Any, event: SignalEvent) -> None:
"""Flush trace batch on system signals to prevent data loss."""
if self.batch_manager.is_batch_initialized():
self.batch_manager.finalize_batch()
def _initialize_crew_batch(self, source: Any, event: Any) -> None:
"""Initialize trace batch.

View File

@@ -0,0 +1,102 @@
"""System signal event types for CrewAI.
This module contains event types for system-level signals like SIGTERM,
allowing listeners to perform cleanup operations before process termination.
"""
from collections.abc import Callable
from enum import IntEnum
import signal
from typing import Annotated, Literal, TypeVar
from pydantic import Field, TypeAdapter
from crewai.events.base_events import BaseEvent
class SignalType(IntEnum):
"""Enumeration of supported system signals."""
SIGTERM = signal.SIGTERM
SIGINT = signal.SIGINT
SIGHUP = signal.SIGHUP
SIGTSTP = signal.SIGTSTP
SIGCONT = signal.SIGCONT
class SigTermEvent(BaseEvent):
"""Event emitted when SIGTERM is received."""
type: Literal["SIGTERM"] = "SIGTERM"
signal_number: SignalType = SignalType.SIGTERM
reason: str | None = None
class SigIntEvent(BaseEvent):
"""Event emitted when SIGINT is received."""
type: Literal["SIGINT"] = "SIGINT"
signal_number: SignalType = SignalType.SIGINT
reason: str | None = None
class SigHupEvent(BaseEvent):
"""Event emitted when SIGHUP is received."""
type: Literal["SIGHUP"] = "SIGHUP"
signal_number: SignalType = SignalType.SIGHUP
reason: str | None = None
class SigTStpEvent(BaseEvent):
"""Event emitted when SIGTSTP is received.
Note: SIGSTOP cannot be caught - it immediately suspends the process.
"""
type: Literal["SIGTSTP"] = "SIGTSTP"
signal_number: SignalType = SignalType.SIGTSTP
reason: str | None = None
class SigContEvent(BaseEvent):
"""Event emitted when SIGCONT is received."""
type: Literal["SIGCONT"] = "SIGCONT"
signal_number: SignalType = SignalType.SIGCONT
reason: str | None = None
SignalEvent = Annotated[
SigTermEvent | SigIntEvent | SigHupEvent | SigTStpEvent | SigContEvent,
Field(discriminator="type"),
]
signal_event_adapter: TypeAdapter[SignalEvent] = TypeAdapter(SignalEvent)
SIGNAL_EVENT_TYPES: tuple[type[BaseEvent], ...] = (
SigTermEvent,
SigIntEvent,
SigHupEvent,
SigTStpEvent,
SigContEvent,
)
T = TypeVar("T", bound=Callable[[object, SignalEvent], None])
def on_signal(func: T) -> T:
"""Decorator to register a handler for all signal events.
Args:
func: Handler function that receives (source, event) arguments.
Returns:
The original function, registered for all signal event types.
"""
from crewai.events.event_bus import crewai_event_bus
for event_type in SIGNAL_EVENT_TYPES:
crewai_event_bus.on(event_type)(func)
return func

View File

@@ -9,12 +9,14 @@ data is collected. Users can opt-in to share more complete data using the
from __future__ import annotations
import asyncio
import atexit
from collections.abc import Callable
from importlib.metadata import version
import json
import logging
import os
import platform
import signal
import threading
from typing import TYPE_CHECKING, Any
@@ -31,6 +33,14 @@ from opentelemetry.sdk.trace.export import (
from opentelemetry.trace import Span
from typing_extensions import Self
from crewai.events.event_bus import crewai_event_bus
from crewai.events.types.system_events import (
SigContEvent,
SigHupEvent,
SigIntEvent,
SigTStpEvent,
SigTermEvent,
)
from crewai.telemetry.constants import (
CREWAI_TELEMETRY_BASE_URL,
CREWAI_TELEMETRY_SERVICE_NAME,
@@ -121,6 +131,7 @@ class Telemetry:
)
self.provider.add_span_processor(processor)
self._register_shutdown_handlers()
self.ready = True
except Exception as e:
if isinstance(
@@ -155,6 +166,71 @@ class Telemetry:
self.ready = False
self.trace_set = False
def _register_shutdown_handlers(self) -> None:
"""Register handlers for graceful shutdown on process exit and signals."""
atexit.register(self._shutdown)
self._original_handlers: dict[int, Any] = {}
self._register_signal_handler(signal.SIGTERM, SigTermEvent, shutdown=True)
self._register_signal_handler(signal.SIGINT, SigIntEvent, shutdown=True)
self._register_signal_handler(signal.SIGHUP, SigHupEvent, shutdown=False)
self._register_signal_handler(signal.SIGTSTP, SigTStpEvent, shutdown=False)
self._register_signal_handler(signal.SIGCONT, SigContEvent, shutdown=False)
def _register_signal_handler(
self,
sig: signal.Signals,
event_class: type,
shutdown: bool = False,
) -> None:
"""Register a signal handler that emits an event.
Args:
sig: The signal to handle.
event_class: The event class to instantiate and emit.
shutdown: Whether to trigger shutdown on this signal.
"""
try:
original_handler = signal.getsignal(sig)
self._original_handlers[sig] = original_handler
def handler(signum: int, frame: Any) -> None:
crewai_event_bus.emit(self, event_class())
if shutdown:
self._shutdown()
if original_handler not in (signal.SIG_DFL, signal.SIG_IGN, None):
if callable(original_handler):
original_handler(signum, frame)
elif shutdown:
raise SystemExit(0)
signal.signal(sig, handler)
except ValueError as e:
logger.warning(
f"Cannot register {sig.name} handler: not running in main thread",
exc_info=e,
)
except OSError as e:
logger.warning(f"Cannot register {sig.name} handler: {e}", exc_info=e)
def _shutdown(self) -> None:
"""Flush and shutdown the telemetry provider on process exit.
Uses a short timeout to avoid blocking process shutdown.
"""
if not self.ready:
return
try:
self.provider.force_flush(timeout_millis=5000)
self.provider.shutdown()
self.ready = False
except Exception as e:
logger.debug(f"Telemetry shutdown failed: {e}")
def _safe_telemetry_operation(
self, operation: Callable[[], Span | None]
) -> Span | None: