mirror of
https://github.com/ThearchyHelios/AgentLab.git
synced 2026-11-15 06:51:59 +00:00
ThearchyHelios
9e33c0f801
eval.regressed、eval.failed 只在定时发起的评测收尾时推（runner._conclude 只认 trigger == "schedule"），手动发起的那次人自己 在看、不推，这和细案 01 做法 15 把这两个事件放在定时一节里一致。可订阅时的叫法却是「评测有退步」「评测失败」，没写「定时」： 有人给全局渠道勾上「评测失败」、再手动发起一次评测并止损停下，渠道里收不到任何消息，界面上也没处说为什么。 - notify/service.EVENTS 和前端 NOTIFY_TEXT.eventLabel 同一句改成「定时评测有退步」「定时评测没有评完」（test_notify_events 逐字对照两边）。eval.failed 不只是失败：到了上限、止损停下（partial）同样推，所以叫「没有评完」，和消息标题本来的说法一致。 - notify_eval 的消息标题也带上「定时」：定时评测「题集名」有退步 / 没有评完。 - 新测试：两个事件的叫法带「定时」；手动发起的评测止损停下，订阅了这两个事件的全局渠道收不到任何消息（去掉 trigger 的判断 这一条就失败）。 - 模块说明、接口说明里引用旧叫法的几处一并改。
817 lines
42 KiB
Python
817 lines
42 KiB
Python
"""定时任务：按周期自动发起工作流运行，结束、失败、停在审批时推送 Webhook，运行成功结束后和同一任务上一次成功的运行
|
||
自动对比。
|
||
|
||
**目标种类**（docs/design/next-round/00-总方案.md X4：定时只有一套）：target_kind 是 workflow（缺省，字段和行为一字不变）、
|
||
eval_suite（跑一次评测题集）或 source_maintenance（维护一个数据源）。非工作流目标写 target_id，不写 workflow_id、输入模板、
|
||
运行类别和 Webhook：核对目标交给各自的模块（scheduler/service.check_target → app/eval/scheduled、app/data/source_maintenance）；
|
||
它们的通知走通知渠道的订阅（定时评测有退步、定时评测没有评完……），不走定时任务的 Webhook——Webhook 的三种事件（运行
|
||
结束、运行失败、等待审批）说的是这一次发起的那次运行，非工作流目标没有那一次运行，配上了也收不到，所以照实拒收。接口返回带
|
||
target_kind、target_id、target_name（目标现在的名字，不在了用保存时记下的那一份）。
|
||
|
||
形状按前后端约定（Schedule、Fire 两种对象，见 _schedule_out、_fire_outs）。Webhook 地址和签名密钥加密存放，接口
|
||
只给主机和末 4 位（url_hint）、有没有密钥（has_secret）。请求体自己解析、自己校验：错误一律回中文 detail，表单
|
||
据 field 把报错落到那一格（422 {detail, code, field}，同 api/coded.py）。后台的调度循环在 app/scheduler 里。
|
||
"""
|
||
from __future__ import annotations
|
||
|
||
import re
|
||
from datetime import datetime, timedelta
|
||
from typing import Any
|
||
|
||
from fastapi import APIRouter, Depends, Header, HTTPException, Query, Request, Response
|
||
from sqlalchemy import and_, delete, func, select, update
|
||
from sqlalchemy.ext.asyncio import AsyncSession
|
||
|
||
from app.api.coded import CodedHTTPException
|
||
from app.api.runs import actor_of
|
||
from app.core.crypto import decrypt, encrypt
|
||
from app.db.base import get_session, utcnow
|
||
from app.db.models import Run, Schedule, ScheduleFire, WebhookDelivery, Workflow, WorkflowVersion
|
||
from app.scheduler import auto_compare, cron, service, webhook
|
||
|
||
router = APIRouter(prefix="/api/schedules", tags=["schedules"])
|
||
|
||
#: 请求不合规（422）的机器码。field 是出错的那一项：name、target_kind、target_id、workflow_id、cron、timezone、run_class、
|
||
#: input_template、enabled、overlap、misfire_grace_s、webhook、version、limit、before
|
||
SCHEDULE_INVALID = "schedule_invalid"
|
||
#: 非工作流目标配了 Webhook（见模块说明）
|
||
NO_WEBHOOK = ("目标是「{kind}」的定时任务不推 Webhook：Webhook 的事件说的是定时发起的那次运行。要收这类定时的通知，"
|
||
"请在「设置 → 通知」的渠道里订阅对应的事件")
|
||
_GONE = "定时任务不存在，可能已被删除"
|
||
#: 错过的触发在多少秒内补发：默认 1 小时，最长 7 天
|
||
DEFAULT_GRACE_S = 3600
|
||
MAX_GRACE_S = 7 * 24 * 3600
|
||
#: 预览最多给几次
|
||
MAX_PREVIEW = 20
|
||
|
||
|
||
def _invalid(detail: str, field: str | None = None) -> CodedHTTPException:
|
||
return CodedHTTPException(422, detail, SCHEDULE_INVALID, field=field)
|
||
|
||
|
||
def _iso(moment: datetime | None) -> str | None:
|
||
return service.as_utc(moment).isoformat() if moment is not None else None
|
||
|
||
|
||
#: 查询串把「+」解成空格之后，时间末尾的时区偏移：「…03.376320 00:00」
|
||
_SPACED_OFFSET = re.compile(r"(.*\d) (\d{2}:?\d{2})")
|
||
|
||
|
||
def _cursor(moment: datetime) -> str:
|
||
"""触发历史的翻页游标：UTC 写成 Z 结尾、精确到微秒。写成 +00:00 时照接口说明原样拼进地址，「+」在查询串里会被
|
||
解成空格，回 422（2026-10 验收）；Z 结尾的放进地址不用编码，和运行列表的游标一个写法。"""
|
||
return service.as_utc(moment).isoformat().replace("+00:00", "Z")
|
||
|
||
|
||
def _cursor_time(before: str) -> datetime:
|
||
"""before → 时间：Z、+00:00 都认，也认「+」被查询串解成空格的「 00:00」（老的游标照原样拼进地址时就是这样）。
|
||
不是时间的照样抛 ValueError。"""
|
||
text = before.strip()
|
||
try:
|
||
return datetime.fromisoformat(text.replace("Z", "+00:00"))
|
||
except ValueError:
|
||
spaced = _SPACED_OFFSET.fullmatch(text)
|
||
if spaced is None:
|
||
raise
|
||
return datetime.fromisoformat(f"{spaced[1]}+{spaced[2]}")
|
||
|
||
|
||
async def _body(request: Request) -> dict[str, Any]:
|
||
try:
|
||
data = await request.json()
|
||
except ValueError:
|
||
data = None
|
||
if not isinstance(data, dict):
|
||
raise _invalid("请求内容需要是 JSON 对象")
|
||
return data
|
||
|
||
|
||
# --------------------------------------------------------------------------
|
||
# 逐项校验：返回规范化之后的值，不对抛 422（带 field）
|
||
# --------------------------------------------------------------------------
|
||
|
||
|
||
def _name(value: Any) -> str:
|
||
text = value.strip() if isinstance(value, str) else ""
|
||
if not text:
|
||
raise _invalid("请填写定时任务的名称", "name")
|
||
if len(text) > 200:
|
||
raise _invalid("定时任务的名称不能超过 200 个字符", "name")
|
||
return text
|
||
|
||
|
||
def _workflow_id(value: Any) -> str:
|
||
if not isinstance(value, str) or not value.strip():
|
||
raise _invalid("请选择要定时运行的工作流", "workflow_id")
|
||
return value.strip()
|
||
|
||
|
||
def _target_kind(value: Any) -> str:
|
||
if value not in service.TARGET_KINDS:
|
||
labels = "、".join(service.TARGET_LABEL[k] for k in service.TARGET_KINDS)
|
||
raise _invalid(f"定时任务的目标需要是{labels}之一{_current(value)}", "target_kind")
|
||
return value
|
||
|
||
|
||
def _target_id(value: Any, kind: str) -> str:
|
||
if not isinstance(value, str) or not value.strip():
|
||
raise _invalid(f"请选择要定时的{service.TARGET_LABEL[kind]}", "target_id")
|
||
return value.strip()
|
||
|
||
|
||
def _target_extras(body: dict[str, Any], kind: str) -> None:
|
||
"""非工作流目标不写工作流、输入模板、Webhook（见模块说明）；运行类别带了也只认合法的写法，不存。"""
|
||
label = service.TARGET_LABEL[kind]
|
||
if not _blank(body.get("workflow_id")):
|
||
raise _invalid(f"目标是「{label}」的定时任务不指定工作流", "workflow_id")
|
||
if body.get("input_template") not in (None, {}):
|
||
raise _invalid(f"目标是「{label}」的定时任务没有输入模板", "input_template")
|
||
hook = body.get("webhook")
|
||
if isinstance(hook, dict) and not _blank(hook.get("url")):
|
||
raise _invalid(NO_WEBHOOK.format(kind=label), "webhook")
|
||
if hook is not None and not isinstance(hook, dict):
|
||
raise _invalid("Webhook 设置需要是一个对象，或者为空表示不推送", "webhook")
|
||
if "run_class" in body:
|
||
_run_class(body["run_class"])
|
||
|
||
|
||
def _cron(value: Any) -> str:
|
||
try:
|
||
return cron.parse(value if isinstance(value, str) else "").expr
|
||
except cron.CronError as e:
|
||
raise _invalid(str(e), "cron") from e
|
||
|
||
|
||
def _timezone(value: Any) -> str:
|
||
try:
|
||
return cron.timezone_name(value if isinstance(value, str) else "")
|
||
except cron.TimezoneError as e:
|
||
raise _invalid(str(e), "timezone") from e
|
||
|
||
|
||
def _run_class(value: Any) -> str:
|
||
if value not in service.RUN_CLASSES:
|
||
raise _invalid(f"运行类别需要选择正式运行或探索运行{_current(value)}", "run_class")
|
||
return value
|
||
|
||
|
||
def _template(value: Any) -> dict[str, str]:
|
||
try:
|
||
return service.check_template(value)
|
||
except service.TemplateError as e:
|
||
raise _invalid(str(e), "input_template") from e
|
||
|
||
|
||
def _current(value: Any) -> str:
|
||
return service.current_value(value)
|
||
|
||
|
||
def _enabled(value: Any) -> bool:
|
||
if not isinstance(value, bool):
|
||
raise _invalid(f"启用状态需要是 true 或 false{_current(value)}", "enabled")
|
||
return value
|
||
|
||
|
||
def _overlap(value: Any) -> str:
|
||
if value != "skip":
|
||
raise _invalid(f"重叠处理目前只支持「跳过」：上一次运行还没结束时跳过这次触发{_current(value)}", "overlap")
|
||
return value
|
||
|
||
|
||
def _grace(value: Any) -> int:
|
||
if isinstance(value, bool) or not isinstance(value, int) or not 0 <= value <= MAX_GRACE_S:
|
||
raise _invalid(f"补发宽限需要是 0 到 {MAX_GRACE_S} 之间的整数秒数（最长 7 天）{_current(value)}",
|
||
"misfire_grace_s")
|
||
return value
|
||
|
||
|
||
def _format(value: Any) -> str:
|
||
if value not in webhook.FORMATS:
|
||
labels = "、".join(webhook.FORMAT_LABEL[f] for f in webhook.FORMATS)
|
||
raise _invalid(f"Webhook 格式需要是{labels}之一{_current(value)}", "webhook")
|
||
return value
|
||
|
||
|
||
def _events(value: Any) -> list[str]:
|
||
labels = "、".join(webhook.EVENT_LABEL[e] for e in webhook.EVENTS)
|
||
if not isinstance(value, list):
|
||
raise _invalid(f"Webhook 的通知事件需要从{labels}中选择", "webhook")
|
||
unknown = [e for e in value if e not in webhook.EVENTS]
|
||
if unknown:
|
||
first = unknown[0]
|
||
named = f"「{first}」" if isinstance(first, str) and first.strip() else "中有一项"
|
||
raise _invalid(f"Webhook 的通知事件{named}不可选，只能从{labels}中选择", "webhook")
|
||
return [e for e in webhook.EVENTS if e in value]
|
||
|
||
|
||
def _url(value: Any) -> str:
|
||
try:
|
||
return webhook.check_url(value)
|
||
except ValueError as e:
|
||
raise _invalid(str(e), "webhook") from e
|
||
|
||
|
||
def _secret(value: Any) -> str | None:
|
||
if value is None or (isinstance(value, str) and not value.strip()):
|
||
return None
|
||
if not isinstance(value, str):
|
||
raise _invalid("Webhook 签名密钥需要是文本", "webhook")
|
||
return value.strip()
|
||
|
||
|
||
def _blank(value: Any) -> bool:
|
||
return value is None or (isinstance(value, str) and not value.strip())
|
||
|
||
|
||
def _matched_format(value: Any, url: str) -> str:
|
||
"""校验格式，并要求和地址对得上：群机器人的地址只认自己平台的格式（见 webhook.bot_format）。"""
|
||
fmt = _format(value)
|
||
if problem := webhook.format_problem(fmt, url):
|
||
raise _invalid(problem, "webhook")
|
||
return fmt
|
||
|
||
|
||
def _new_webhook(value: Any) -> dict[str, Any] | None:
|
||
"""创建时的 webhook：null 或 url 为空表示不配；否则校验格式、地址、密钥、事件（事件省略表示三种都要）。
|
||
格式省略时按地址认：群机器人的地址用它平台的格式，别的用通用格式。"""
|
||
if value is None:
|
||
return None
|
||
if not isinstance(value, dict):
|
||
raise _invalid("Webhook 设置需要是一个对象，或者为空表示不推送", "webhook")
|
||
if _blank(value.get("url")):
|
||
return None
|
||
url = _url(value.get("url"))
|
||
given = value.get("format")
|
||
fmt = _matched_format(given, url) if given is not None else webhook.bot_format(url) or "generic"
|
||
return {"format": fmt, "url": url, "secret": _secret(value.get("secret")),
|
||
"events": _events(value.get("events", list(webhook.EVENTS)))}
|
||
|
||
|
||
def _webhook_columns(hook: dict[str, Any] | None) -> dict[str, Any]:
|
||
if hook is None:
|
||
return {"webhook_format": None, "webhook_url": None, "webhook_secret": None, "webhook_events": []}
|
||
return {"webhook_format": hook["format"], "webhook_url": encrypt(hook["url"]),
|
||
"webhook_secret": encrypt(hook["secret"]) if hook["secret"] else None, "webhook_events": hook["events"]}
|
||
|
||
|
||
def _patched_webhook(row: Schedule, value: Any) -> dict[str, Any]:
|
||
"""PATCH 的 webhook → 要写的列。null 或 url 写空串：删除整个 Webhook；url 省略：地址不变；secret 省略：密钥不变、
|
||
写空串：去掉密钥。format 省略：换了地址的按新地址认（群机器人用它平台的格式），没换的不变。
|
||
|
||
地址或格式真的变了才查两者对不对得上：库里已有对不上的设置（这条校验之前存的）时，整份表单回写只为停用、
|
||
改名，不能因为 Webhook 被拦下。"""
|
||
if value is None:
|
||
return _webhook_columns(None)
|
||
if not isinstance(value, dict):
|
||
raise _invalid("Webhook 设置需要是一个对象，或者为空表示不推送", "webhook")
|
||
if "url" in value and _blank(value.get("url")):
|
||
return _webhook_columns(None)
|
||
has_url = bool(row.webhook_url)
|
||
if "url" not in value and not has_url:
|
||
raise _invalid("请填写 Webhook 地址", "webhook")
|
||
url = _url(value.get("url")) if "url" in value else None
|
||
if value.get("format") is not None:
|
||
fmt = _format(value["format"])
|
||
elif url is not None:
|
||
fmt = webhook.bot_format(url) or (row.webhook_format if has_url else "generic")
|
||
else:
|
||
fmt = row.webhook_format
|
||
if url is not None or fmt != row.webhook_format:
|
||
target = url or decrypt(row.webhook_url)
|
||
if target and (problem := webhook.format_problem(fmt, target)):
|
||
raise _invalid(problem, "webhook")
|
||
out: dict[str, Any] = {
|
||
"webhook_format": _format(fmt),
|
||
"webhook_events": _events(value.get("events", list(row.webhook_events or []) if has_url
|
||
else list(webhook.EVENTS))),
|
||
}
|
||
if url is not None:
|
||
out["webhook_url"] = encrypt(url)
|
||
if "secret" in value:
|
||
secret = _secret(value.get("secret"))
|
||
out["webhook_secret"] = encrypt(secret) if secret else None
|
||
return out
|
||
|
||
|
||
async def _check_workflow(session: AsyncSession, workflow_id: str, run_class: str, template: dict[str, str]) -> Workflow:
|
||
"""工作流要在；正式运行要有发布版本；模板里的字段名对得上入口字段，必填项都有（按这次要用的那张图查）。"""
|
||
workflow = await session.get(Workflow, workflow_id)
|
||
if workflow is None:
|
||
raise _invalid("工作流不存在，可能已被删除", "workflow_id")
|
||
graph: Any = workflow.graph
|
||
if run_class == "formal":
|
||
if not workflow.published_version:
|
||
raise _invalid(f"「{workflow.name}」还没有发布版本，正式运行只能从发布版本发起。"
|
||
"请先在编排页发布，或改为探索运行", "run_class")
|
||
graph = await _published_graph(session, workflow) or graph
|
||
try:
|
||
service.check_fields(template, service.input_fields(graph), workflow.name)
|
||
except service.TemplateError as e:
|
||
raise _invalid(str(e), "input_template") from e
|
||
return workflow
|
||
|
||
|
||
async def _published_graph(session: AsyncSession, workflow: Workflow) -> Any:
|
||
return await session.scalar(select(WorkflowVersion.graph).where(
|
||
WorkflowVersion.workflow_id == workflow.id, WorkflowVersion.version == workflow.published_version))
|
||
|
||
|
||
# --------------------------------------------------------------------------
|
||
# 对外的形状
|
||
# --------------------------------------------------------------------------
|
||
|
||
|
||
def _description(row: Schedule) -> str:
|
||
try:
|
||
return cron.describe(row.cron, row.timezone)
|
||
except cron.CronError:
|
||
return f"按 cron 表达式 {row.cron}（{row.timezone}）"
|
||
|
||
|
||
def _webhook_out(row: Schedule) -> dict[str, Any] | None:
|
||
if not row.webhook_url:
|
||
return None
|
||
plain = decrypt(row.webhook_url)
|
||
return {
|
||
"format": row.webhook_format,
|
||
# 解不开（加密密钥换过）时不猜：照实说，请人重新填
|
||
"url_hint": webhook.url_hint(plain) if plain else "（地址无法解密，请重新填写）",
|
||
"has_secret": bool(row.webhook_secret),
|
||
"events": [e for e in webhook.EVENTS if e in (row.webhook_events or [])],
|
||
}
|
||
|
||
|
||
async def _fire_outs(session: AsyncSession, fires: list[ScheduleFire]) -> list[dict[str, Any]]:
|
||
"""Fire 对象：run_status 是运行现在的状态，notified 是这次触发的推送记录 [{event, status, at}]（按时间先后），
|
||
compare 是运行成功结束后和同一任务上一次成功的运行的自动对比 {base_run_id, summary, previous}（summary 同对比接口
|
||
的计数，previous 说基准是不是紧挨着的上一次），没有为 null。
|
||
|
||
对比和比的时候两边的封存绑定（auto_compare）：哪一边现在不在终态（被继续执行）、被删掉、清单哈希和比的时候不一样
|
||
（重新封存过）都给 null——对比页按现在的封存现算，记下的计数和点进去看到的会对不上，或者干脆比不了。工件读不出来、
|
||
被改过的同样给 null。基准已经不是现在挑出来的那一次（auto_compare.outdated：中间隔着的运行事后接着运行成功了）也
|
||
给 null：运行详情里同名的入口比的是另一次。"""
|
||
compared = {f.id: f.compare_artifact for f in fires if f.compare_artifact and f.run_id}
|
||
compares = await auto_compare.briefs(list(compared.values()))
|
||
stale = await auto_compare.outdated(session, compares)
|
||
run_ids = {f.run_id for f in fires if f.run_id} | {brief["base_run_id"] for brief in compares.values()}
|
||
# 运行现在的状态和清单哈希：run_status 照写，对比按它核对两边的封存
|
||
current = {run_id: (status, seal) for run_id, status, seal in (await session.execute(
|
||
select(Run.id, Run.status, Run.manifest_hash).where(Run.id.in_(run_ids)))).all()} if run_ids else {}
|
||
|
||
def compare_of(f: ScheduleFire) -> dict[str, Any] | None:
|
||
artifact = compared.get(f.id) or ""
|
||
brief = compares.get(artifact)
|
||
if brief is None or brief["target_run_id"] != f.run_id or artifact in stale:
|
||
return None
|
||
for run_id, seal in ((brief["target_run_id"], brief["target_manifest"]),
|
||
(brief["base_run_id"], brief["base_manifest"])):
|
||
status, now_seal = current.get(run_id, (None, None))
|
||
if status not in auto_compare.TERMINAL or now_seal != seal:
|
||
return None
|
||
return auto_compare.public(brief)
|
||
|
||
notified: dict[str, list[dict[str, Any]]] = {}
|
||
if fires:
|
||
rows = (await session.execute(
|
||
select(WebhookDelivery).where(WebhookDelivery.fire_id.in_([f.id for f in fires]))
|
||
.order_by(WebhookDelivery.created_at)
|
||
)).scalars()
|
||
for d in rows:
|
||
notified.setdefault(d.fire_id, []).append({"event": d.event, "status": d.status, "at": _iso(d.created_at)})
|
||
# 正在发起的触发：库里是占位的「未能发起」（为了进程被强杀时照实说），进程还在、只是还没发起完，按已发起给
|
||
launching = {f.id for f in fires if service.launching(f.id)}
|
||
return [{
|
||
"id": f.id,
|
||
"fire_at": _iso(f.fire_at),
|
||
"kind": f.kind,
|
||
"status": "started" if f.id in launching else f.status,
|
||
"run_id": f.run_id,
|
||
"run_status": current.get(f.run_id, (None, None))[0] if f.run_id else None,
|
||
# 目标是评测题集时这一次触发建的那次评测（没有运行 id）
|
||
"eval_run_id": f.eval_run_id,
|
||
"reason": None if f.id in launching else f.reason,
|
||
"notified": notified.get(f.id, []),
|
||
"compare": compare_of(f),
|
||
"created_at": _iso(f.created_at),
|
||
} for f in fires]
|
||
|
||
|
||
async def _schedule_outs(session: AsyncSession, rows: list[Schedule]) -> list[dict[str, Any]]:
|
||
ids = [r.id for r in rows]
|
||
names: dict[str, str] = {}
|
||
targets: dict[str, str | None] = {}
|
||
last: dict[str, dict[str, Any]] = {}
|
||
if rows:
|
||
names = dict((await session.execute(
|
||
select(Workflow.id, Workflow.name).where(Workflow.id.in_({r.workflow_id for r in rows if r.workflow_id})))).all())
|
||
for r in rows:
|
||
kind, target_id = service.target_of(r)
|
||
if kind != "workflow":
|
||
targets[r.id] = await service.target_name(session, kind, target_id)
|
||
newest = (select(ScheduleFire.schedule_id, func.max(ScheduleFire.created_at).label("at"))
|
||
.where(ScheduleFire.schedule_id.in_(ids)).group_by(ScheduleFire.schedule_id).subquery())
|
||
fires = list((await session.execute(select(ScheduleFire).join(newest, and_(
|
||
ScheduleFire.schedule_id == newest.c.schedule_id, ScheduleFire.created_at == newest.c.at)))).scalars())
|
||
for out, fire in zip(await _fire_outs(session, fires), fires):
|
||
last.setdefault(fire.schedule_id, out)
|
||
def target(r: Schedule) -> dict[str, Any]:
|
||
kind, target_id = service.target_of(r)
|
||
name = names.get(r.workflow_id) if kind == "workflow" else targets.get(r.id)
|
||
# 目标还在就给它现在的名字；删掉之后用保存时记下的那一份
|
||
return {"target_kind": kind, "target_id": target_id, "target_label": service.TARGET_LABEL.get(kind, kind),
|
||
"target_name": name or r.workflow_name, "target_exists": name is not None}
|
||
|
||
return [{
|
||
"id": r.id,
|
||
"name": r.name,
|
||
**target(r),
|
||
"workflow_id": r.workflow_id,
|
||
# 工作流还在就给它现在的名字；删掉之后用保存时记下的那一份。非工作流目标这一项是目标的名字（同 target_name）
|
||
"workflow_name": names.get(r.workflow_id) or targets.get(r.id) or r.workflow_name,
|
||
"cron": r.cron,
|
||
"timezone": r.timezone,
|
||
"run_class": r.run_class,
|
||
"input_template": dict(r.input_template or {}),
|
||
"enabled": r.enabled,
|
||
"overlap": r.overlap,
|
||
"misfire_grace_s": r.misfire_grace_s,
|
||
"webhook": _webhook_out(r),
|
||
"next_run_at": _iso(r.next_run_at),
|
||
"last_fire": last.get(r.id),
|
||
"description_text": _description(r),
|
||
"created_by": r.created_by,
|
||
"created_at": _iso(r.created_at),
|
||
"updated_at": _iso(r.updated_at),
|
||
"version": r.version,
|
||
} for r in rows]
|
||
|
||
|
||
async def _schedule_out(session: AsyncSession, row: Schedule) -> dict[str, Any]:
|
||
return (await _schedule_outs(session, [row]))[0]
|
||
|
||
|
||
async def _get(session: AsyncSession, schedule_id: str) -> Schedule:
|
||
row = await session.get(Schedule, schedule_id)
|
||
if row is None:
|
||
raise HTTPException(404, _GONE)
|
||
return row
|
||
|
||
|
||
# --------------------------------------------------------------------------
|
||
# 固定路径（要排在 /{schedule_id} 前面）
|
||
# --------------------------------------------------------------------------
|
||
|
||
|
||
@router.get("/variables")
|
||
async def list_variables(tz: str | None = Query(default=None, alias="timezone")) -> list[dict[str, str]]:
|
||
"""输入模板能用的变量：[{name, label, example, note}]。示例按现在这一刻、timezone（默认 Asia/Shanghai）触发算；
|
||
note 是补充说明（ISO 周的跨年规则），没有的是空串。"""
|
||
zone = _timezone(tz) if tz is not None else service.DEFAULT_TIMEZONE
|
||
return service.variables(tz=zone)
|
||
|
||
|
||
@router.post("/preview")
|
||
async def preview(request: Request, session: AsyncSession = Depends(get_session)) -> dict[str, Any]:
|
||
"""{cron, timezone, input_template?, workflow_id?, n?=5} → {description_text, next: [ISO（UTC）…], inputs}。
|
||
|
||
inputs 是按下一次触发渲染的输入。带了 workflow_id 时对照入口字段的名字；必填项没填不在这里报——用户还在填。
|
||
"""
|
||
body = await _body(request)
|
||
cron_expr, tz = _cron(body.get("cron")), _timezone(body.get("timezone"))
|
||
template = _template(body.get("input_template"))
|
||
n = body.get("n", 5)
|
||
n = min(max(n, 1), MAX_PREVIEW) if isinstance(n, int) and not isinstance(n, bool) else 5
|
||
workflow_id = body.get("workflow_id")
|
||
if isinstance(workflow_id, str) and workflow_id.strip():
|
||
workflow = await session.get(Workflow, workflow_id.strip())
|
||
if workflow is not None:
|
||
# 还不知道按哪一类运行：发布版本和画布的字段都认
|
||
published = await _published_graph(session, workflow) if workflow.published_version else None
|
||
fields = service.input_fields(workflow.graph) + service.input_fields(published)
|
||
try:
|
||
service.check_fields(template, fields, workflow.name, require=False)
|
||
except service.TemplateError as e:
|
||
raise _invalid(str(e), "input_template") from e
|
||
upcoming = cron.preview(cron_expr, tz, n)
|
||
first = upcoming[0] if upcoming else utcnow()
|
||
return {
|
||
"description_text": cron.describe(cron_expr, tz),
|
||
"next": [_iso(m) for m in upcoming],
|
||
"inputs": service.render_inputs(template, service.template_values(first, tz, cron_expr)),
|
||
}
|
||
|
||
|
||
# --------------------------------------------------------------------------
|
||
# 增删改查
|
||
# --------------------------------------------------------------------------
|
||
|
||
|
||
@router.get("")
|
||
async def list_schedules(session: AsyncSession = Depends(get_session)) -> dict[str, Any]:
|
||
rows = list((await session.execute(select(Schedule).order_by(Schedule.created_at.desc()))).scalars())
|
||
return {"items": await _schedule_outs(session, rows)}
|
||
|
||
|
||
@router.post("", status_code=201)
|
||
async def create_schedule(request: Request, session: AsyncSession = Depends(get_session),
|
||
x_actor: str | None = Header(default=None)) -> dict[str, Any]:
|
||
row = await create_schedule_row(session, await _body(request), actor=actor_of(x_actor))
|
||
return await _schedule_out(session, row)
|
||
|
||
|
||
def schedule_values(body: dict[str, Any]) -> dict[str, Any]:
|
||
"""建定时任务的请求体 → 规范化之后的列（不含工作流名、下一次触发），逐项校验，不对抛 422（带 field）。只读，
|
||
不碰库：发布单的预检用它「只检查」，建任务的接口和发布单建任务都先过它。
|
||
|
||
target_kind 缺省 workflow：交回的键和以前一字不差。非工作流目标交回 target_kind、target_id，运行类别存探索（不用它）、
|
||
输入模板为空、没有 Webhook。"""
|
||
kind = _target_kind(body.get("target_kind", "workflow"))
|
||
if kind != "workflow":
|
||
_target_extras(body, kind)
|
||
return {"name": _name(body.get("name")), "target_kind": kind, "target_id": _target_id(body.get("target_id"), kind),
|
||
"cron": _cron(body.get("cron")), "timezone": _timezone(body.get("timezone")),
|
||
"run_class": "exploratory", "input_template": {}, "enabled": _enabled(body.get("enabled", True)),
|
||
"overlap": _overlap(body.get("overlap", "skip")),
|
||
"misfire_grace_s": _grace(body.get("misfire_grace_s", DEFAULT_GRACE_S)), "webhook": None}
|
||
values = {"name": _name(body.get("name")), "workflow_id": _workflow_id(body.get("workflow_id")),
|
||
"cron": _cron(body.get("cron")), "timezone": _timezone(body.get("timezone")),
|
||
"run_class": _run_class(body.get("run_class", "formal")),
|
||
"input_template": _template(body.get("input_template")),
|
||
"enabled": _enabled(body.get("enabled", True)), "overlap": _overlap(body.get("overlap", "skip")),
|
||
"misfire_grace_s": _grace(body.get("misfire_grace_s", DEFAULT_GRACE_S))}
|
||
return {**values, "webhook": _new_webhook(body.get("webhook"))}
|
||
|
||
|
||
def template_problem(template: dict[str, str], graph: Any, workflow_name: str) -> str | None:
|
||
"""输入模板对不上这张图的入口字段（字段名、必填项）：说明；对得上为 None。发布单的预检按将要发布的图查。"""
|
||
try:
|
||
service.check_fields(template, service.input_fields(graph), workflow_name)
|
||
except service.TemplateError as e:
|
||
return str(e)
|
||
return None
|
||
|
||
|
||
async def _check_target(session: AsyncSession, kind: str, target_id: str) -> str:
|
||
"""非工作流目标：交给它的模块核对（存在、可用），交回目标现在的名字。不对抛 422（field=target_id）。"""
|
||
if problem := await service.check_target(session, kind, target_id):
|
||
raise _invalid(problem, "target_id")
|
||
return await service.target_name(session, kind, target_id) or ""
|
||
|
||
|
||
async def create_schedule_row(session: AsyncSession, body: dict[str, Any], *, actor: str | None) -> Schedule:
|
||
"""校验并建一个定时任务（提交）。建任务的接口和发布单（停用着建，应用发布后再启用）共用。"""
|
||
values = schedule_values(body)
|
||
hook = values.pop("webhook")
|
||
if values.get("target_kind", "workflow") != "workflow":
|
||
kind, target_id = values.pop("target_kind"), values.pop("target_id")
|
||
name = await _check_target(session, kind, target_id)
|
||
# 目标的几列只经 scheduler/service 的一处封装写（非工作流目标：workflow_id 存空串）；workflow_name 记保存时的目标名
|
||
values |= service.target_columns(kind, target_id)
|
||
row = Schedule(**values, workflow_name=name,
|
||
next_run_at=service.next_run(values["cron"], values["timezone"], values["enabled"]),
|
||
created_by=actor, **_webhook_columns(None))
|
||
session.add(row)
|
||
await session.commit()
|
||
await session.refresh(row)
|
||
return row
|
||
workflow = await _check_workflow(session, values["workflow_id"], values["run_class"], values["input_template"])
|
||
# 目标的几列只经 scheduler/service 的一处封装写（工作流目标：target_id 同 workflow_id）
|
||
values |= service.target_columns("workflow", values["workflow_id"])
|
||
row = Schedule(**values, workflow_name=workflow.name,
|
||
next_run_at=service.next_run(values["cron"], values["timezone"], values["enabled"]),
|
||
created_by=actor, **_webhook_columns(hook))
|
||
session.add(row)
|
||
await session.commit()
|
||
await session.refresh(row)
|
||
return row
|
||
|
||
|
||
@router.get("/{schedule_id}")
|
||
async def get_schedule(schedule_id: str, session: AsyncSession = Depends(get_session)) -> dict[str, Any]:
|
||
return await _schedule_out(session, await _get(session, schedule_id))
|
||
|
||
|
||
@router.patch("/{schedule_id}")
|
||
async def update_schedule(schedule_id: str, request: Request,
|
||
session: AsyncSession = Depends(get_session)) -> dict[str, Any]:
|
||
"""改任意可改的项，带上读到的 version；对不上回 409。只读的项（id、next_run_at、last_fire……）带了也不理。
|
||
|
||
看的是值变没变，不是键在不在：编辑表单把整份字段带回来是正常用法。
|
||
- 工作流、运行类别、输入模板真改了才按改后的组合重新校验。只停用、改名不查工作流——工作流删了、撤了发布，
|
||
也要能先把任务停下来。
|
||
- cron、时区、启用状态真改了才重算下一次触发（见 _next_after_edit）。没改的不动它：到点了、这一拍还没来时
|
||
保存一次，这一次照常发起，不会被推到下一次、连记录都没有就丢了。
|
||
"""
|
||
body = await _body(request)
|
||
row = await _get(session, schedule_id)
|
||
version = body.get("version")
|
||
if isinstance(version, bool) or not isinstance(version, int):
|
||
raise _invalid("保存时需要带上读到的版本号 version，请刷新后重试", "version")
|
||
if version != row.version:
|
||
raise HTTPException(409, f"这个定时任务已被修改（当前是第 {row.version} 版，提交的是第 {version} 版），"
|
||
"请刷新后再保存")
|
||
values: dict[str, Any] = {}
|
||
kind, current_target = service.target_of(row)
|
||
if "target_kind" in body and _target_kind(body["target_kind"]) != kind:
|
||
raise _invalid("定时任务的目标种类建好之后不能改：请新建一个定时任务", "target_kind")
|
||
if kind != "workflow":
|
||
return await _update_target_schedule(session, row, body, kind, current_target, version)
|
||
if "name" in body:
|
||
values["name"] = _name(body["name"])
|
||
if "workflow_id" in body:
|
||
# 目标的几列一起改（scheduler/service.target_columns）：只改 workflow_id 的话 target_id 还指着原来的工作流
|
||
values |= service.target_columns("workflow", _workflow_id(body["workflow_id"]))
|
||
if "cron" in body:
|
||
values["cron"] = _cron(body["cron"])
|
||
if "timezone" in body:
|
||
values["timezone"] = _timezone(body["timezone"])
|
||
if "run_class" in body:
|
||
values["run_class"] = _run_class(body["run_class"])
|
||
if "input_template" in body:
|
||
values["input_template"] = _template(body["input_template"])
|
||
if "enabled" in body:
|
||
values["enabled"] = _enabled(body["enabled"])
|
||
if "overlap" in body:
|
||
values["overlap"] = _overlap(body["overlap"])
|
||
if "misfire_grace_s" in body:
|
||
values["misfire_grace_s"] = _grace(body["misfire_grace_s"])
|
||
if "webhook" in body:
|
||
values |= _patched_webhook(row, body["webhook"])
|
||
|
||
def merged(key: str) -> Any:
|
||
return values.get(key, getattr(row, key))
|
||
|
||
def changed(key: str) -> bool:
|
||
if key not in values:
|
||
return False
|
||
stored = getattr(row, key)
|
||
return values[key] != (dict(stored or {}) if key == "input_template" else stored)
|
||
|
||
if changed("workflow_id") or changed("run_class") or changed("input_template"):
|
||
workflow = await _check_workflow(session, merged("workflow_id"), merged("run_class"), merged("input_template"))
|
||
values["workflow_name"] = workflow.name
|
||
# cron、时区在上面已经规范化（字段之间一个空格、时区的标准写法），和库里存的同一种写法，直接比
|
||
if changed("cron") or changed("timezone") or changed("enabled"):
|
||
values["next_run_at"] = _next_after_edit(row, merged("cron"), merged("timezone"), merged("enabled"))
|
||
# 条件 UPDATE 才是真的乐观锁：两个人同时读到第 3 版、同时保存，只有先到的那个改得动
|
||
result = await session.execute(
|
||
update(Schedule).where(Schedule.id == schedule_id, Schedule.version == version)
|
||
.values(**values, version=version + 1).execution_options(synchronize_session=False))
|
||
if result.rowcount != 1:
|
||
await session.rollback()
|
||
raise HTTPException(409, "这个定时任务刚被别人修改过，请刷新后再保存")
|
||
await session.commit()
|
||
session.expire_all()
|
||
return await _schedule_out(session, await _get(session, schedule_id))
|
||
|
||
|
||
async def _update_target_schedule(session: AsyncSession, row: Schedule, body: dict[str, Any], kind: str,
|
||
current_target: str, version: int) -> dict[str, Any]:
|
||
"""改一个非工作流目标的定时任务：名称、目标（同一种）、cron、时区、启用、重叠、补发宽限。工作流、输入模板、运行类别、
|
||
Webhook 不适用（带了照实拒；整份表单带回来的空值不算）。目标真改了才重新核对。"""
|
||
probe = {k: body[k] for k in ("workflow_id", "input_template", "webhook", "run_class") if k in body}
|
||
if probe.get("webhook") is None:
|
||
probe.pop("webhook", None)
|
||
_target_extras(probe, kind)
|
||
values: dict[str, Any] = {}
|
||
if "name" in body:
|
||
values["name"] = _name(body["name"])
|
||
if "target_id" in body:
|
||
target_id = _target_id(body["target_id"], kind)
|
||
if target_id != current_target:
|
||
values["workflow_name"] = await _check_target(session, kind, target_id)
|
||
values |= service.target_columns(kind, target_id)
|
||
for key, check in (("cron", _cron), ("timezone", _timezone), ("enabled", _enabled), ("overlap", _overlap),
|
||
("misfire_grace_s", _grace)):
|
||
if key in body:
|
||
values[key] = check(body[key])
|
||
if any(k in values and values[k] != getattr(row, k) for k in ("cron", "timezone", "enabled")):
|
||
values["next_run_at"] = _next_after_edit(row, values.get("cron", row.cron), values.get("timezone", row.timezone),
|
||
values.get("enabled", row.enabled))
|
||
result = await session.execute(
|
||
update(Schedule).where(Schedule.id == row.id, Schedule.version == version)
|
||
.values(**values, version=version + 1).execution_options(synchronize_session=False))
|
||
if result.rowcount != 1:
|
||
await session.rollback()
|
||
raise HTTPException(409, "这个定时任务刚被别人修改过，请刷新后再保存")
|
||
schedule_id = row.id
|
||
await session.commit()
|
||
session.expire_all()
|
||
return await _schedule_out(session, await _get(session, schedule_id))
|
||
|
||
|
||
async def retime(session: AsyncSession, row: Schedule, body: dict[str, Any]) -> Schedule:
|
||
"""按新的设置改一个已有的定时任务（提交）：名称、cron、时区、输入模板，校验同建任务；启用状态不动。cron、时区真改了才
|
||
按 PATCH 同一个规则重算下一次触发。发布单给已经发布过的蓝图再发一版时用它沿用原来的任务，不另建一个（两个任务会
|
||
在同一时刻各发起一次）。"""
|
||
values = schedule_values({**body, "workflow_id": row.workflow_id, "enabled": row.enabled})
|
||
workflow = await _check_workflow(session, row.workflow_id, row.run_class, values["input_template"])
|
||
changed = {"name": values["name"], "cron": values["cron"], "timezone": values["timezone"],
|
||
"input_template": values["input_template"], "workflow_name": workflow.name}
|
||
if values["cron"] != row.cron or values["timezone"] != row.timezone:
|
||
changed["next_run_at"] = _next_after_edit(row, values["cron"], values["timezone"], row.enabled)
|
||
await session.execute(update(Schedule).where(Schedule.id == row.id)
|
||
.values(**changed, version=row.version + 1).execution_options(synchronize_session=False))
|
||
await session.commit()
|
||
await session.refresh(row)
|
||
return row
|
||
|
||
|
||
async def set_enabled(session: AsyncSession, row: Schedule, enabled: bool) -> Schedule:
|
||
"""启用、停用一个定时任务（提交），下一次触发按 PATCH 同一个规则算，版本号加一。发布单在应用发布之后用它启用
|
||
先停用着建的任务；已经是这个状态的不动。"""
|
||
if row.enabled == enabled:
|
||
return row
|
||
await session.execute(
|
||
update(Schedule).where(Schedule.id == row.id)
|
||
.values(enabled=enabled, next_run_at=_next_after_edit(row, row.cron, row.timezone, enabled),
|
||
version=row.version + 1).execution_options(synchronize_session=False))
|
||
await session.commit()
|
||
# 只重读这一行：调用方（发布单）同一个会话里还拿着别的对象，不能整个会话过期
|
||
await session.refresh(row)
|
||
return row
|
||
|
||
|
||
def _next_after_edit(row: Schedule, cron_expr: str, tz: str, enabled: bool) -> datetime | None:
|
||
"""改了 cron、时区或启用状态之后的下一次触发。
|
||
|
||
停用：没有下一次。刚启用：从现在算，不补停用期间错过的。启用着改了 cron、时区：原来那一次已经到点、这一拍还没来
|
||
的，从那一刻接着按新的表达式算——新表达式也覆盖那一刻（每天 → 工作日）就还发起这一次，只属于旧表达式的不再发起。
|
||
"""
|
||
if not enabled:
|
||
return None
|
||
now = utcnow()
|
||
due = service.as_utc(row.next_run_at) if row.enabled and row.next_run_at is not None else None
|
||
if due is not None and due <= now:
|
||
return cron.next_fire(cron_expr, tz, due - timedelta(microseconds=1))
|
||
return service.next_run(cron_expr, tz, True, now)
|
||
|
||
|
||
@router.delete("/{schedule_id}", status_code=204)
|
||
async def delete_schedule(schedule_id: str, session: AsyncSession = Depends(get_session)) -> Response:
|
||
"""删除任务，连同它的触发历史和推送记录。它发起过的运行留着（运行上的来源 id 也留着）：那是追溯凭证。"""
|
||
await _get(session, schedule_id)
|
||
# 外键是 ON DELETE CASCADE；这里照样按顺序删一遍，不依赖连接上开没开外键约束
|
||
await session.execute(delete(WebhookDelivery).where(WebhookDelivery.schedule_id == schedule_id))
|
||
await session.execute(delete(ScheduleFire).where(ScheduleFire.schedule_id == schedule_id))
|
||
await session.execute(delete(Schedule).where(Schedule.id == schedule_id))
|
||
await session.commit()
|
||
return Response(status_code=204)
|
||
|
||
|
||
# --------------------------------------------------------------------------
|
||
# 立即运行、触发历史、测试 Webhook
|
||
# --------------------------------------------------------------------------
|
||
|
||
|
||
@router.post("/{schedule_id}/run")
|
||
async def run_now(schedule_id: str, session: AsyncSession = Depends(get_session),
|
||
x_actor: str | None = Header(default=None)) -> dict[str, Any]:
|
||
"""立即触发一次（kind=manual），停用的任务也可以。规则和到点触发相同：上一次运行没结束就跳过，正式运行只从当前
|
||
发布版本发起。运行记在点按钮的人名下，没有署名时记在「定时任务：<名字>」名下。返回 {fire, run_id}；目标是评测题集的，
|
||
发起的是一次评测，fire.eval_run_id 是它，run_id 为空。
|
||
|
||
触发时间取到秒：连点两下落在同一秒的，交回同一条触发记录，不会发起两次。"""
|
||
await _get(session, schedule_id)
|
||
fire = await service.fire_once(schedule_id, utcnow().replace(microsecond=0), "manual", actor=actor_of(x_actor))
|
||
if fire is None:
|
||
raise HTTPException(404, _GONE)
|
||
return {"fire": (await _fire_outs(session, [fire]))[0], "run_id": fire.run_id}
|
||
|
||
|
||
@router.get("/{schedule_id}/fires")
|
||
async def list_fires(schedule_id: str, limit: str | None = None, before: str | None = None,
|
||
session: AsyncSession = Depends(get_session)) -> dict[str, Any]:
|
||
"""触发历史，新的在前：{items: [Fire], next_before}。翻页把 next_before 原样作为 before 传回来（Z 结尾，拼进地址
|
||
不用编码）；没有更多时为空。"""
|
||
await _get(session, schedule_id)
|
||
size = 20
|
||
if limit is not None:
|
||
if not limit.strip().isdecimal() or not 1 <= int(limit) <= 100:
|
||
raise _invalid(f"每页条数需要是 1 到 100 之间的整数，当前为「{limit}」", "limit")
|
||
size = int(limit)
|
||
stmt = select(ScheduleFire).where(ScheduleFire.schedule_id == schedule_id)
|
||
if before is not None:
|
||
try:
|
||
cursor = _cursor_time(before)
|
||
except ValueError as e:
|
||
raise _invalid(f"翻页位置需要是接口返回的 next_before，当前为「{before}」", "before") from e
|
||
stmt = stmt.where(ScheduleFire.created_at < service.as_utc(cursor))
|
||
rows = list((await session.execute(
|
||
stmt.order_by(ScheduleFire.created_at.desc(), ScheduleFire.id.desc()).limit(size + 1))).scalars())
|
||
items = rows[:size]
|
||
return {"items": await _fire_outs(session, items),
|
||
"next_before": _cursor(items[-1].created_at) if len(rows) > size else None}
|
||
|
||
|
||
@router.post("/{schedule_id}/webhook/test")
|
||
async def test_webhook(schedule_id: str, session: AsyncSession = Depends(get_session)) -> dict[str, Any]:
|
||
"""按保存的 Webhook 发一条测试消息（不重试）：{ok, http_status, message}。失败的说明里只有主机名。"""
|
||
await _get(session, schedule_id)
|
||
try:
|
||
return await webhook.send_test(schedule_id)
|
||
except KeyError as e:
|
||
raise HTTPException(404, _GONE) from e
|
||
except ValueError as e:
|
||
raise _invalid(str(e), "webhook") from e
|