Spaces:
Running
Running
File size: 10,013 Bytes
5f3e9f5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 | """Single-worker queue for long-running generation workflows."""
from __future__ import annotations
import threading
import traceback
import sys
from collections import deque
from dataclasses import dataclass
from typing import Callable
from core.ai_client import unregister_operation
from core import run_manager
from core.workflow_runner import WorkflowContext, publish_run_event
from utils.run_guard import RunRejected, begin_run, end_run
@dataclass
class QueuedJob:
run_id: str
operation_id: str
fn: Callable[[WorkflowContext], None]
cancel_event: threading.Event
label: str = "workflow"
pipeline_enabled: bool = False
_QUEUE: deque[QueuedJob] = deque()
_CURRENT: dict[str, QueuedJob] = {}
_CONDITION = threading.Condition()
_STARTED = False
def _prevent_sleep() -> bool:
if sys.platform != "win32":
return False
try:
import ctypes
ES_CONTINUOUS = 0x80000000
ES_SYSTEM_REQUIRED = 0x00000001
ES_DISPLAY_REQUIRED = 0x00000002
ctypes.windll.kernel32.SetThreadExecutionState(
ES_CONTINUOUS | ES_SYSTEM_REQUIRED | ES_DISPLAY_REQUIRED
)
return True
except Exception:
return False
def _allow_sleep() -> None:
if sys.platform != "win32":
return
try:
import ctypes
ES_CONTINUOUS = 0x80000000
ctypes.windll.kernel32.SetThreadExecutionState(ES_CONTINUOUS)
except Exception:
pass
def _queue_positions() -> dict[str, int]:
return {job.run_id: index + 1 for index, job in enumerate(_QUEUE)}
def _publish_queue_positions() -> None:
positions = _queue_positions()
for job in list(_QUEUE):
position = positions.get(job.run_id)
message = f"Queued at position {position}" if position else "Queued"
run_manager.update_run(
job.run_id,
status="queued",
stage="queued",
message=message,
progress=0,
queue_position=position,
)
run_manager.add_event(
job.run_id,
event_type="queued",
stage="queued",
progress=0,
message=message,
data={"queue_position": position},
)
publish_run_event(job.run_id, {
"type": "queued",
"run_id": job.run_id,
"operation_id": job.operation_id,
"message": message,
"progress": 0,
"queue_position": position,
})
def _worker_loop() -> None:
while True:
with _CONDITION:
while not _QUEUE:
_CONDITION.wait()
job = _QUEUE.popleft()
_CURRENT[job.run_id] = job
_publish_queue_positions()
_run_job(job)
def _run_job(job: QueuedJob) -> None:
ctx = WorkflowContext(job.run_id, job.operation_id, job.cancel_event)
# In classic mode, hold the global single-flight slot for the
# duration of the job. Pipeline mode deliberately releases that
# coarse lock and lets the workflow's own phase locks protect
# screenshots / PowerPoint while AI can overlap.
slot_held = False
sleep_guard = False
try:
sleep_guard = _prevent_sleep()
if not job.pipeline_enabled:
try:
begin_run(job.operation_id, "/job-queue", {"run_id": job.run_id})
slot_held = True
except RunRejected as rr:
ctx.fail(f"Backend busy: {rr.message}")
return
if job.cancel_event.is_set():
ctx.cancel()
else:
ctx.started(f"{job.label} started")
job.fn(ctx)
except Exception as exc:
traceback.print_exc()
run = run_manager.get_run(job.run_id)
if run and run.get("status") not in {"completed", "failed", "cancelled"}:
ctx.fail(f"Error: {exc}")
finally:
if sleep_guard:
_allow_sleep()
if slot_held:
end_run(job.operation_id)
with _CONDITION:
_CURRENT.pop(job.run_id, None)
_publish_queue_positions()
def _ensure_worker() -> None:
global _STARTED
with _CONDITION:
if _STARTED:
return
thread = threading.Thread(target=_worker_loop, daemon=True, name="workflow-job-queue")
thread.start()
_STARTED = True
def enqueue(
run_id: str,
operation_id: str,
fn: Callable[[WorkflowContext], None],
cancel_event: threading.Event,
label: str = "workflow",
pipeline_enabled: bool = False,
) -> int:
job = QueuedJob(
run_id=run_id,
operation_id=operation_id,
fn=fn,
cancel_event=cancel_event,
label=label,
pipeline_enabled=pipeline_enabled,
)
_ensure_worker()
with _CONDITION:
if pipeline_enabled and not _QUEUE and not any(
not current.pipeline_enabled for current in _CURRENT.values()
):
_CURRENT[job.run_id] = job
_publish_queue_positions()
thread = threading.Thread(
target=_run_job,
args=(job,),
daemon=True,
name=f"workflow-{run_id}",
)
thread.start()
return 1
_QUEUE.append(job)
position = len(_QUEUE)
_publish_queue_positions()
_CONDITION.notify()
return position
_SOFT_CANCEL_STAGES = {
"after_html": ("html_saved", "Cancellation requested. The process will stop after the HTML file is saved."),
"after_screenshots": ("screenshots_done", "Cancellation requested. The process will stop after screenshots are captured."),
"after_pptx": ("pptx_built", "Cancellation requested. The process will stop after the PowerPoint deck is saved."),
"after_video": ("video_export_done", "Cancellation requested. The process will stop after the MP4 export finishes."),
}
def cancel_job(run_id: str, mode: str = "now", delete_outputs: bool = False) -> bool:
mode = str(mode or "now").strip().lower()
delete_outputs = bool(delete_outputs)
with _CONDITION:
for job in list(_QUEUE):
if job.run_id == run_id or job.operation_id == run_id:
job.cancel_event.set()
_QUEUE.remove(job)
run_manager.update_run(job.run_id, settings={"delete_outputs_on_cancel": delete_outputs})
run_manager.finish_run(job.run_id, status="cancelled", message="Operation cancelled before start", progress=0)
unregister_operation(job.operation_id)
publish_run_event(job.run_id, {
"type": "cancelled",
"run_id": job.run_id,
"operation_id": job.operation_id,
"message": "Operation cancelled before start",
})
_publish_queue_positions()
return True
current = next(
(job for job in _CURRENT.values() if job.run_id == run_id or job.operation_id == run_id),
None,
)
if current:
if mode in _SOFT_CANCEL_STAGES:
stage, message = _SOFT_CANCEL_STAGES[mode]
current_run = run_manager.get_run(current.run_id) or {}
run_manager.update_run(
current.run_id,
status="running",
stage="cancelling",
message=message,
progress=None,
settings={
"cancel_after_stage": stage,
"delete_outputs_on_cancel": False,
},
)
publish_run_event(current.run_id, {
"type": "progress",
"run_id": current.run_id,
"operation_id": current.operation_id,
"message": message,
"stage": "cancelling",
"progress": current_run.get("progress", 0),
})
return True
current.cancel_event.set()
current_run = run_manager.get_run(current.run_id) or {}
run_manager.update_run(
current.run_id,
status="running",
stage="cancelling",
message="Cancellation requested. Waiting for the running step to stop.",
progress=None,
settings={"delete_outputs_on_cancel": delete_outputs},
)
publish_run_event(current.run_id, {
"type": "progress",
"run_id": current.run_id,
"operation_id": current.operation_id,
"message": "Cancellation requested",
"stage": "cancelling",
"progress": current_run.get("progress", 0),
})
return True
return False
def get_queue_snapshot() -> dict:
with _CONDITION:
running = list(_CURRENT.values())
return {
"running": {
"run_id": running[0].run_id,
"operation_id": running[0].operation_id,
"label": running[0].label,
} if running else None,
"running_all": [
{
"run_id": job.run_id,
"operation_id": job.operation_id,
"label": job.label,
"pipeline_enabled": job.pipeline_enabled,
}
for job in running
],
"queued": [
{
"run_id": job.run_id,
"operation_id": job.operation_id,
"label": job.label,
"position": index + 1,
}
for index, job in enumerate(_QUEUE)
],
}
|