datawhalechina/hello-agents · error · AgentException
任务不存在: {task_id}
Error message
任务不存在: {task_id} What it means
Lookup guard in ControllerAgent.execute_task: the given task_id is not a key in self.active_tasks. Tasks enter that dict only via submit_task and are removed in the finally block of execute_task, so the ID is unknown (never submitted, typo'd) or already finished — after completion the task moves to task_history and a second execute call with the same ID fails here.
Source
Thrown at Co-creation-projects/Apricity-InnocoreAI/agents/controller.py:107
"callback": callback,
"created_at": datetime.now(),
"started_at": None,
"completed_at": None,
"result": None,
"error": None,
"agent_results": {}
}
self.active_tasks[task_id] = task
await self.task_queue.put((priority, task))
logger.info(f"任务已提交: {task_id}, 类型: {task_type.value}")
return task_id
async def execute_task(self, task_id: str) -> Dict[str, Any]:
"""执行单个任务"""
if task_id not in self.active_tasks:
raise AgentException(f"任务不存在: {task_id}")
task = self.active_tasks[task_id]
async with self.semaphore: # 并发控制
try:
task["status"] = TaskStatus.RUNNING
task["started_at"] = datetime.now()
await self._trigger_event("task_started", task)
# 根据任务类型执行相应的逻辑
if task["type"] == TaskType.PAPER_HUNTING:
result = await self._execute_paper_hunting(task)
elif task["type"] == TaskType.PAPER_ANALYSIS:
result = await self._execute_paper_analysis(task)
elif task["type"] == TaskType.WRITING_ASSISTANCE:
result = await self._execute_writing_assistance(task)
elif task["type"] == TaskType.CITATION_VALIDATION:View on GitHub (pinned to 606a07d341)
Solutions
- Use the exact ID returned by submit_task, kept as a string, and execute it once.
- For re-execution, submit a new task instead of reusing a completed ID.
- Before raising, fall back: if task_id in self.task_history, report 'task already finished' instead of 'not found' for clearer UX.
- Persist task registry (or query task_history) if the controller outlives requests.
Example fix
# before
if task_id not in self.active_tasks:
raise AgentException(f"任务不存在: {task_id}")
# after
if task_id not in self.active_tasks:
done = next((t for t in self.task_history if t["id"] == task_id), None)
if done:
raise AgentException(f"任务 {task_id} 已结束(状态: {done['status']})，不能重复执行")
raise AgentException(f"任务不存在: {task_id}"); Defensive patterns
Strategy: validation
Validate before calling
# Caller-side: only execute ids you got from submit_task, exactly once
if not isinstance(task_id, str) or not task_id:
raise ValueError("task_id must be a non-empty string")
if task_id not in controller.active_tasks:
done = any(t["id"] == task_id for t in controller.task_history)
status = "already finished" if done else "unknown id"
raise ValueError(f"cannot execute {task_id!r}: {status}") Type guard
def is_active_task(controller, task_id: str) -> bool:
return isinstance(task_id, str) and task_id in controller.active_tasks Try / catch
try:
result = await controller.execute_task(task_id)
except AgentException as e:
if "任务不存在" in str(e):
task_id = await controller.submit_task(...) # resubmit and execute the new id
result = await controller.execute_task(task_id)
else:
raise Prevention
- Treat submit_task's return value as an opaque, single-use handle.
- Never retry execute_task with the same id after completion; submit a new task.
- Persist or query task_history for idempotent APIs.
When it happens
Trigger: Calling execute_task('abc') with an ID not returned by submit_task; retrying execute_task after a successful first run (ID moved to task_history in finally); racing a background worker that executed the queued task first; string-vs-int ID mismatch from JSON deserialization.
Common situations: API handler extracting task_id from URL with wrong type/whitespace; double-click on an 'execute' endpoint; frontend caching an old task ID after backend restart wiped in-memory state (active_tasks is not persisted).
Related errors
AI-assisted analysis of datawhalechina/hello-agents@606a07d341 (2026-08-14).
Data as JSON: /api/errors/d9b706fa8cd798d0.
Report an issue: GitHub.