Yuzhiran Dev
5037d5d0ed
01:00 search → 01:10 trends → 01:30 collect → 02:00 create → 03:00 review → 05:00 optimize_sources → 06:00 metrics 各间隔: 10min/20min/30min/60min/2h/1h 采集到创作30min缓冲足够(采集≈5-15min)
337 lines
15 KiB
Python
337 lines
15 KiB
Python
"""
|
||
定时任务调度器
|
||
基于 APScheduler，支持在 FastAPI 生命周期内运行定时任务
|
||
"""
|
||
import os
|
||
import sys
|
||
import logging
|
||
from datetime import datetime
|
||
from pathlib import Path
|
||
from apscheduler.schedulers.background import BackgroundScheduler
|
||
from apscheduler.triggers.cron import CronTrigger
|
||
from .generator import run_creator_blocking
|
||
from .optimizer import run_optimizer_blocking
|
||
from .collector import run_collector_blocking
|
||
|
||
logger = logging.getLogger(__name__)
|
||
|
||
class TaskScheduler:
|
||
def __init__(self):
|
||
self.scheduler = BackgroundScheduler()
|
||
self._started = False
|
||
|
||
def start(self):
|
||
if self._started:
|
||
logger.warning("Scheduler already started")
|
||
return
|
||
# 使用 CronTrigger 设置每日固定时间点
|
||
# 顺序: 搜索缓存(01:00)→趋势(01:10)→采集(01:30)→创作(02:00)→审查(03:00)→源优化(05:00)→指标(06:00)
|
||
self.scheduler.add_job(
|
||
self._run_refresh_search_cache,
|
||
CronTrigger(hour=1, minute=0),
|
||
id='scheduled_refresh_search_cache',
|
||
replace_existing=True,
|
||
max_instances=1,
|
||
coalesce=True
|
||
)
|
||
self.scheduler.add_job(
|
||
self._run_fetch_trends,
|
||
CronTrigger(hour=1, minute=10),
|
||
id='scheduled_fetch_trends',
|
||
replace_existing=True,
|
||
max_instances=1,
|
||
coalesce=True
|
||
)
|
||
self.scheduler.add_job(
|
||
self._run_collect,
|
||
CronTrigger(hour=1, minute=30),
|
||
id='scheduled_collect',
|
||
)
|
||
self.scheduler.add_job(
|
||
self._run_generate,
|
||
CronTrigger(hour=2, minute=0),
|
||
id='scheduled_generate',
|
||
replace_existing=True,
|
||
max_instances=1,
|
||
coalesce=True
|
||
)
|
||
self.scheduler.add_job(
|
||
self._run_optimize,
|
||
CronTrigger(hour=3, minute=0),
|
||
id='scheduled_optimize',
|
||
replace_existing=True,
|
||
max_instances=1,
|
||
coalesce=True
|
||
)
|
||
self.scheduler.add_job(
|
||
self._run_optimize_sources,
|
||
CronTrigger(hour=5, minute=0),
|
||
id='scheduled_optimize_sources',
|
||
replace_existing=True,
|
||
max_instances=1,
|
||
coalesce=True
|
||
)
|
||
self.scheduler.add_job(
|
||
self._run_metrics_sync,
|
||
CronTrigger(hour=6, minute=0),
|
||
id='scheduled_metrics_sync',
|
||
replace_existing=True,
|
||
max_instances=1,
|
||
coalesce=True
|
||
)
|
||
self.scheduler.start()
|
||
self._started = True
|
||
logger.info("Scheduler started: 01:00 search 01:10 trends 01:30 collect 02:00 create 03:00 review 05:00 sources 06:00 metrics")
|
||
def shutdown(self):
|
||
if self.scheduler.running:
|
||
self.scheduler.shutdown()
|
||
logger.info("Scheduler shut down")
|
||
|
||
def _run_fetch_trends(self):
|
||
"""定时刷新热点趋势（百度/微博/知乎实时热搜 + LLM补充）"""
|
||
try:
|
||
logger.info("[Scheduled] Fetching hot trends...")
|
||
import subprocess
|
||
result = subprocess.run(
|
||
[sys.executable, str(Path(__file__).parent.parent.parent.parent / "scripts" / "trends.py")],
|
||
capture_output=True, text=True, timeout=120
|
||
)
|
||
if result.returncode == 0:
|
||
for line in result.stdout.strip().split("\n"):
|
||
if line.strip():
|
||
logger.info("[Trends] %s", line.strip())
|
||
logger.info("[Scheduled] Trends refreshed successfully")
|
||
else:
|
||
logger.warning("[Scheduled] Trends refresh failed: %s", result.stderr[-500:])
|
||
except Exception as e:
|
||
logger.exception("[Scheduled] Trends refresh error: %s", e)
|
||
|
||
def _run_refresh_search_cache(self):
|
||
"""定时刷新搜索缓存（通过 opencode webfetch）"""
|
||
try:
|
||
logger.info("[Scheduled] Refreshing search cache via opencode...")
|
||
import subprocess
|
||
result = subprocess.run(
|
||
[sys.executable, str(Path(__file__).parent.parent.parent.parent / "scripts" / "opencode_search.py"), "--refresh-cache"],
|
||
capture_output=True, text=True, timeout=600
|
||
)
|
||
for line in result.stdout.strip().split("\n"):
|
||
if line.strip():
|
||
logger.info("[SearchCache] %s", line.strip())
|
||
for line in result.stderr.strip().split("\n"):
|
||
if line.strip():
|
||
logger.warning("[SearchCache] %s", line.strip())
|
||
if result.returncode == 0:
|
||
logger.info("[Scheduled] Search cache refreshed")
|
||
else:
|
||
logger.warning("[Scheduled] Search cache refresh may have partial failures")
|
||
except subprocess.TimeoutExpired:
|
||
logger.warning("[Scheduled] Search cache refresh timed out")
|
||
except Exception as e:
|
||
logger.exception("[Scheduled] Search cache refresh error: %s", e)
|
||
|
||
def _run_generate(self):
|
||
try:
|
||
logger.info("[Scheduled] Starting content generation...")
|
||
result = run_creator_blocking()
|
||
logger.info("[Scheduled] Generation completed: %s", result)
|
||
created_id = result.get("topic_id") if isinstance(result, dict) else None
|
||
if created_id:
|
||
logger.info("[Scheduled] Running compliance review on %s...", created_id)
|
||
review_result = run_optimizer_blocking([created_id])
|
||
if review_result.get("ok"):
|
||
logger.info("[Scheduled] Review completed for %s", created_id)
|
||
else:
|
||
logger.warning("[Scheduled] Review failed: %s", review_result.get("error"))
|
||
except Exception as e:
|
||
logger.exception("[Scheduled] Generation pipeline failed: %s", e)
|
||
|
||
def _run_optimize(self):
|
||
try:
|
||
logger.info("[Scheduled] Starting compliance review...")
|
||
result = run_optimizer_blocking()
|
||
logger.info("[Scheduled] Review completed: %s", result)
|
||
except Exception as e:
|
||
logger.exception("[Scheduled] Review failed: %s", e)
|
||
|
||
def _run_collect(self):
|
||
try:
|
||
logger.info("[Scheduled] Starting topic collection...")
|
||
result = run_collector_blocking()
|
||
logger.info("[Scheduled] Collection completed: %s", result.get("output", "")[-200:])
|
||
except Exception as e:
|
||
logger.exception("[Scheduled] Collection failed: %s", e)
|
||
|
||
def _run_optimize_sources(self):
|
||
"""AI自动优化采集类别与信息源：对比市场热点和当前配置，给出调整建议"""
|
||
try:
|
||
logger.info("[Scheduled] Starting source optimization with AI...")
|
||
from .nvidia_client import call_llm
|
||
from ..database import SessionLocal
|
||
from ..models import CollectorCategory, CollectorSource
|
||
from datetime import date
|
||
|
||
db = SessionLocal()
|
||
try:
|
||
cats = db.query(CollectorCategory).filter(CollectorCategory.is_active == True).all()
|
||
sources = db.query(CollectorSource).filter(CollectorSource.is_active == True).all()
|
||
except Exception:
|
||
logger.warning("[Scheduled] DB not ready for source optimization")
|
||
db.close()
|
||
return
|
||
|
||
cat_names = [c.name for c in cats]
|
||
src_summary = "\n".join(f"- [{s.source_type}] {s.name}: {s.query or s.url or ''}" for s in sources)
|
||
|
||
prompt = f"""你是一个内容策略分析师。分析当前中文互联网可持续生活领域的真实热点，与以下配置进行对比。
|
||
|
||
当前配置的类别（{len(cat_names)}个）：
|
||
{chr(10).join(f'- {n}' for n in cat_names)}
|
||
|
||
当前配置的信息源（{len(sources)}个）：
|
||
{src_summary}
|
||
|
||
请完成以下任务：
|
||
1. 评估每个类别是否仍符合2026年中国市场真实热点（基于你的知识）
|
||
2. 评估每个信息源是否可能在中国正常访问
|
||
3. 建议新增或删除的类别（最多2条）
|
||
4. 建议新增的信息源搜索词（最多3条，包含具体搜索词）
|
||
|
||
输出 JSON 格式：
|
||
{{
|
||
"category_assessment": [{{"name": "类别名", "status": "保留/淘汰/合并", "reason": "原因"}}],
|
||
"source_assessment": [{{"name": "源名", "status": "保留/淘汰/替换", "reason": "原因"}}],
|
||
"suggested_new_categories": [{{"name": "类别名", "search_query": "搜索词", "reason": "推荐原因"}}],
|
||
"suggested_new_sources": [{{"name": "源名", "type": "web_search", "query": "搜索词", "focus": "聚焦领域"}}],
|
||
"summary": "一句话总结本次优化建议"
|
||
}}
|
||
|
||
只输出JSON，不要其他文字。"""
|
||
|
||
resp = call_llm(prompt, temperature=0.5, max_tokens=2000)
|
||
if resp.startswith("```"):
|
||
resp = resp.split("\n", 1)[1].rsplit("\n", 1)[0]
|
||
result = json.loads(resp)
|
||
|
||
# 将AI建议写入系统配置（供运营参考，不自动执行）
|
||
from ..models import SystemConfig
|
||
sc = db.query(SystemConfig).filter(SystemConfig.key == "collector_ai_advice").first()
|
||
if sc:
|
||
sc.value = json.dumps(result, ensure_ascii=False)
|
||
else:
|
||
db.add(SystemConfig(key="collector_ai_advice", value=json.dumps(result, ensure_ascii=False), description="AI每日采集优化建议"))
|
||
db.commit()
|
||
logger.info("[Scheduled] Source AI optimization completed: %s", result.get("summary", ""))
|
||
db.close()
|
||
except Exception as e:
|
||
logger.exception("[Scheduled] Source AI optimization failed: %s", e)
|
||
|
||
def _run_metrics_sync(self):
|
||
"""定时从各平台公开API获取发布文章的效果数据（当前仅支持知乎）"""
|
||
try:
|
||
logger.info("[Scheduled] Starting metrics sync (zhihu auto-fetch)...")
|
||
from ..database import SessionLocal
|
||
from ..models import Topic, ContentMetrics
|
||
import re, requests as http_requests
|
||
|
||
db = SessionLocal()
|
||
try:
|
||
topics = db.query(Topic).filter(
|
||
Topic.status.in_(["published", "已发布"])
|
||
).all()
|
||
except Exception:
|
||
logger.warning("[Scheduled] DB not ready for metrics sync")
|
||
db.close()
|
||
return
|
||
|
||
ua = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36"
|
||
count = 0
|
||
for topic in topics:
|
||
platform_urls = topic.platform_urls or {}
|
||
zhihu_url = platform_urls.get("zhihu", "")
|
||
if not zhihu_url:
|
||
continue
|
||
m = re.search(r'zhuanlan\.zhihu\.com/p/(\d+)', zhihu_url)
|
||
if not m:
|
||
continue
|
||
post_id = m.group(1)
|
||
api_url = f"https://zhuanlan.zhihu.com/api/posts/{post_id}"
|
||
try:
|
||
resp = http_requests.get(api_url, headers={"User-Agent": ua}, timeout=10)
|
||
if resp.status_code != 200:
|
||
continue
|
||
raw = resp.json()
|
||
existing = db.query(ContentMetrics).filter(
|
||
ContentMetrics.topic_id == topic.id,
|
||
ContentMetrics.platform == "zhihu"
|
||
).first()
|
||
metric_data = {
|
||
"views": raw.get("voteup_count", raw.get("views_count", 0)),
|
||
"likes": raw.get("voteup_count", 0),
|
||
"favorites": raw.get("favorite_count", 0),
|
||
"comments": raw.get("comment_count", raw.get("comments_count", 0)),
|
||
"shares": raw.get("share_count", 0),
|
||
"last_fetched": datetime.now(),
|
||
"publish_url": zhihu_url,
|
||
"data_snapshot": raw,
|
||
}
|
||
if existing:
|
||
for k, v in metric_data.items():
|
||
setattr(existing, k, v)
|
||
else:
|
||
db.add(ContentMetrics(topic_id=topic.id, platform="zhihu", **metric_data))
|
||
count += 1
|
||
except Exception:
|
||
continue
|
||
if count:
|
||
db.commit()
|
||
logger.info("[Scheduled] Metrics sync completed: synced %d zhihu articles", count)
|
||
# 生成指标反馈：按 field 聚合表现，写入 metrics_feedback.json 供 collector 读取
|
||
try:
|
||
import json as json_mod
|
||
from sqlalchemy import func as sql_func
|
||
feedback = db.query(
|
||
Topic.field,
|
||
sql_func.avg(ContentMetrics.likes).label("avg_likes"),
|
||
sql_func.avg(ContentMetrics.views).label("avg_views"),
|
||
sql_func.avg(ContentMetrics.comments).label("avg_comments"),
|
||
sql_func.count(ContentMetrics.id).label("article_count"),
|
||
).join(ContentMetrics, ContentMetrics.topic_id == Topic.id
|
||
).filter(Topic.field.isnot(None), Topic.field != ""
|
||
).group_by(Topic.field).all()
|
||
if feedback:
|
||
scored = []
|
||
for row in feedback:
|
||
score = (row.avg_likes or 0) + (row.avg_views or 0) * 0.01 + (row.avg_comments or 0) * 2
|
||
scored.append((row.field, round(score, 1), int(row.article_count)))
|
||
scored.sort(key=lambda x: x[1], reverse=True)
|
||
feedback_data = {
|
||
"updated_at": datetime.now().isoformat(),
|
||
"top_domains": [(f, s) for f, s, _ in scored[:5]],
|
||
"detail": [{"field": f, "score": s, "articles": c} for f, s, c in scored],
|
||
}
|
||
feedback_file = Path(__file__).parent.parent.parent.parent / "automation" / "data" / "metrics_feedback.json"
|
||
feedback_file.parent.mkdir(parents=True, exist_ok=True)
|
||
feedback_file.write_text(json_mod.dumps(feedback_data, ensure_ascii=False, indent=2), encoding='utf-8')
|
||
logger.info("[Scheduled] Metrics feedback written: top domain %s (score %.1f)", scored[0][0], scored[0][1])
|
||
except Exception as e_fb:
|
||
logger.warning("[Scheduled] Metrics feedback generation failed: %s", e_fb)
|
||
else:
|
||
logger.info("[Scheduled] Metrics sync: no zhihu articles to sync")
|
||
db.close()
|
||
except Exception as e:
|
||
logger.exception("[Scheduled] Metrics sync failed: %s", e)
|
||
|
||
def get_jobs(self):
|
||
"""返回当前所有定时任务的状态"""
|
||
jobs = []
|
||
for job in self.scheduler.get_jobs():
|
||
jobs.append({
|
||
"id": job.id,
|
||
"next_run_time": job.next_run_time.isoformat() if job.next_run_time else None,
|
||
"trigger": str(job.trigger),
|
||
})
|
||
return jobs
|
||
|
||
# 全局单例
|
||
scheduler = TaskScheduler() |