1

app/cli.py

@@ -30,6 +30,8 @@ def build_parser():
 
     event = subparsers.add_parser("event", help="运行事件驱动筛选")
     event.add_argument("--no-process-existing", action="store_true", help="只处理本轮新采集事件")
+    event.add_argument("--limit", type=int, default=None, help="单轮最多处理事件数")
+    event.add_argument("--max-seconds", type=int, default=None, help="单轮最大运行秒数")
 
     sentiment = subparsers.add_parser("sentiment", help="运行舆情任务")
     sentiment.add_argument("--collect", action="store_true", help="采集并存储")
@@ -73,7 +75,11 @@ def main():
     if args.command == "review":
         return review_engine.run_review(push_enabled=not args.no_push, compact=args.compact)
     if args.command == "event":
-        result = event_driven_screener.run_once(process_existing=not args.no_process_existing)
+        result = event_driven_screener.run_once(
+            process_existing=not args.no_process_existing,
+            limit=args.limit,
+            max_seconds=args.max_seconds,
+        )
         print(event_driven_screener.json.dumps(result, ensure_ascii=False, indent=2, default=str))
         return result
     if args.command == "sentiment":

app/db/scheduler_db.py

@@ -21,7 +21,7 @@ DEFAULT_JOBS = [
     {
         "job_name": "event",
         "command": "event",
-        "args": [],
+        "args": ["--limit", "4", "--max-seconds", "50"],
         "every_seconds": 60,
         "initial_delay": 5,
         "lock_group": "recommendation_write",

app/services/altcoin_screener.py

@@ -102,9 +102,21 @@ def _fetch_spot_24h_tickers():
     load_markets(), which is exactly the endpoint most likely to be IP-banned.
     The public spot 24h endpoint is enough for our broad universe scan.
     """
-    resp = requests.get(f"{BINANCE_SPOT_BASE_URL}/api/v3/ticker/24hr", timeout=15)
-    resp.raise_for_status()
-    data = resp.json()
+    try:
+        resp = requests.get(f"{BINANCE_SPOT_BASE_URL}/api/v3/ticker/24hr", timeout=15)
+        resp.raise_for_status()
+        data = resp.json()
+        _write_spot_24h_ticker_cache(data)
+    except Exception as exc:
+        data = _read_spot_24h_ticker_cache(max_age_seconds=0)
+        if data is None:
+            print(f"Binance 24h ticker 拉取失败且无缓存，跳过本轮行情快照: {exc}")
+            _fetch_spot_24h_tickers.last_raw_count = 0
+            _fetch_spot_24h_tickers.last_usdt_count = 0
+            _fetch_spot_24h_tickers.last_error = str(exc)[:500]
+            return {}
+        print(f"Binance 24h ticker 拉取失败，使用本地缓存兜底: {exc}")
+        _fetch_spot_24h_tickers.last_error = str(exc)[:500]
     if not isinstance(data, list):
         return {}
     tickers = {}
@@ -137,6 +149,38 @@ def _fetch_spot_24h_tickers():
     return tickers
 
 
+def _spot_24h_ticker_cache_path():
+    return EXCHANGE_CACHE_DIR / "binance_spot_24h_tickers.json"
+
+
+def _read_spot_24h_ticker_cache(max_age_seconds=300):
+    path = _spot_24h_ticker_cache_path()
+    try:
+        if not path.exists():
+            return None
+        payload = json.loads(path.read_text(encoding="utf-8"))
+        fetched_at = float(payload.get("fetched_at") or 0)
+        if max_age_seconds and time.time() - fetched_at > max_age_seconds:
+            return None
+        data = payload.get("data")
+        return data if isinstance(data, list) else None
+    except Exception:
+        return None
+
+
+def _write_spot_24h_ticker_cache(data):
+    try:
+        if not isinstance(data, list):
+            return
+        EXCHANGE_CACHE_DIR.mkdir(parents=True, exist_ok=True)
+        _spot_24h_ticker_cache_path().write_text(
+            json.dumps({"fetched_at": time.time(), "data": data}, ensure_ascii=False),
+            encoding="utf-8",
+        )
+    except Exception:
+        pass
+
+
 def _spot_exchange_info_cache_path():
     return EXCHANGE_CACHE_DIR / "binance_spot_exchange_info.json"
 

app/services/event_driven_screener.py

@@ -42,7 +42,10 @@ from app.services.altcoin_screener import (
 from app.services.altcoin_confirm import fetch_derivatives_context
 from app.core.pa_engine import full_pa_analysis, calc_atr
 
-exchange = ccxt.binance({"enableRateLimit": True})
+exchange = ccxt.binance({
+    "enableRateLimit": True,
+    "timeout": int(os.getenv("ALPHAX_EVENT_CCXT_TIMEOUT_MS", "8000")),
+})
 
 LEVEL_RANK = {"S": 4, "A": 3, "B": 2, "C": 1, "D": 0, "RISK": 5}
 
@@ -60,6 +63,22 @@ def _cfg():
     return load_rules(force_reload=True).get("event_driven", {})
 
 
+def _positive_int(value, default):
+    try:
+        n = int(value)
+        return n if n > 0 else default
+    except Exception:
+        return default
+
+
+def _process_limit(override=None):
+    return _positive_int(override, _positive_int(_cfg().get("max_process_events_per_run"), 4))
+
+
+def _max_run_seconds(override=None):
+    return _positive_int(override, _positive_int(_cfg().get("max_run_seconds"), 50))
+
+
 def _parse_binance_time(ms):
     try:
         return datetime.fromtimestamp(int(ms) / 1000)
@@ -249,19 +268,27 @@ def _event_copy_for_symbol(event, symbol, theme_name, theme, expanded=True):
 
 def expand_theme_events(events):
     """重大生态/主题事件扩散到同生态币，解决 TON/DOGS 这类联动行情漏选。"""
-    if not _theme_cfg().get("enabled", False):
+    theme_cfg = _theme_cfg()
+    if not theme_cfg.get("enabled", False):
         return events
     expanded = list(events)
     seen = {(e.get("source"), e.get("title"), e.get("symbol")) for e in expanded}
+    max_expanded = _positive_int(theme_cfg.get("max_expanded_symbols"), 12)
     for e in events:
+        expanded_for_event = 0
         for theme_name, theme in _matched_themes(e.get("title", ""), e.get("symbol", "")):
+            if expanded_for_event >= max_expanded:
+                break
             direct = _event_copy_for_symbol(e, e.get("symbol"), theme_name, theme, expanded=False)
             key = (direct.get("source"), direct.get("title"), direct.get("symbol"))
             if direct.get("symbol") and key not in seen and _tradable_symbol(direct.get("symbol")):
                 expanded.append(direct)
                 seen.add(key)
+                expanded_for_event += 1
 
             for base in theme.get("symbols", []):
+                if expanded_for_event >= max_expanded:
+                    break
                 symbol = f"{str(base).upper()}/USDT"
                 if symbol == e.get("symbol") or not _tradable_symbol(symbol):
                     continue
@@ -270,6 +297,7 @@ def expand_theme_events(events):
                 if key not in seen:
                     expanded.append(child)
                     seen.add(key)
+                    expanded_for_event += 1
     return expanded
 
 
@@ -866,15 +894,25 @@ def load_unprocessed_events(limit=20):
     return events
 
 
-def run_once(process_existing=True):
+def run_once(process_existing=True, limit=None, max_seconds=None):
     started = _now()
     init_db()
     init_event_tables()
+    process_limit = _process_limit(limit)
+    runtime_limit = _max_run_seconds(max_seconds)
     collected = collect_events()
     stored = store_events(collected)
-    to_process = stored if stored else (load_unprocessed_events() if process_existing else [])
+    if stored:
+        to_process = stored[:process_limit]
+    else:
+        to_process = load_unprocessed_events(limit=process_limit) if process_existing else []
     processed = []
+    skipped_due_to_limit = max(len(stored) - len(to_process), 0) if stored else 0
+    stopped_by_runtime = False
     for e in to_process:
+        if (_now() - started).total_seconds() >= runtime_limit:
+            stopped_by_runtime = True
+            break
         if isinstance(e.get("published_at"), str):
             e["published_at"] = datetime.fromisoformat(e["published_at"])
         processed.append(process_event(e))
@@ -884,6 +922,10 @@ def run_once(process_existing=True):
         "collected_count": len(collected),
         "stored_count": len(stored),
         "processed_count": len(processed),
+        "skipped_due_to_limit": skipped_due_to_limit,
+        "stopped_by_runtime": stopped_by_runtime,
+        "process_limit": process_limit,
+        "max_run_seconds": runtime_limit,
         "decisions": {k: sum(1 for p in processed if p["result"]["decision"] == k) for k in ["recommend", "observe", "risk", "ignore"]},
         "events": [
             {
@@ -908,7 +950,15 @@ def run_once(process_existing=True):
         started_at=started.isoformat(),
         finished_at=_now().isoformat(),
         duration_ms=int((_now() - started).total_seconds() * 1000),
-        summary={"stored_count": len(stored), "processed_count": len(processed), "decisions": output["decisions"]},
+        summary={
+            "stored_count": len(stored),
+            "processed_count": len(processed),
+            "skipped_due_to_limit": skipped_due_to_limit,
+            "stopped_by_runtime": stopped_by_runtime,
+            "process_limit": process_limit,
+            "max_run_seconds": runtime_limit,
+            "decisions": output["decisions"],
+        },
         error_message="",
     )
     return output
@@ -919,8 +969,14 @@ def main():
     parser = argparse.ArgumentParser(description="事件驱动舆情触发选币")
     parser.add_argument("--once", action="store_true")
     parser.add_argument("--no-process-existing", action="store_true")
+    parser.add_argument("--limit", type=int, default=None, help="单轮最多处理事件数")
+    parser.add_argument("--max-seconds", type=int, default=None, help="单轮最大运行秒数")
     args = parser.parse_args()
-    out = run_once(process_existing=not args.no_process_existing)
+    out = run_once(
+        process_existing=not args.no_process_existing,
+        limit=args.limit,
+        max_seconds=args.max_seconds,
+    )
     print(json.dumps(out, ensure_ascii=False, indent=2, default=str))
 
 

docker/scheduler.py

@@ -214,8 +214,10 @@ def finish_running_jobs(running: dict[str, RunningJob]) -> None:
         proc = item.proc
         timeout = max(item.job.every_seconds * 2, 600)
         elapsed = time.time() - item.started_at
+        killed_by_timeout = False
         if proc.poll() is None and elapsed > timeout:
             proc.kill()
+            killed_by_timeout = True
             print(f"[{now_str()}] [scheduler] timeout {name} after {elapsed:.1f}s", flush=True)
         if proc.poll() is None:
             continue
@@ -241,14 +243,30 @@ def finish_running_jobs(running: dict[str, RunningJob]) -> None:
         if output_tail:
             print(output_tail, flush=True)
         if exit_code != 0:
+            timed_out = killed_by_timeout or elapsed >= timeout
+            error_type = f"{name}_exit_{exit_code}"
+            message = f"scheduler job {name} failed with exit={exit_code}"
+            if timed_out and exit_code == -9:
+                error_type = f"{name}_timeout_killed"
+                message = f"scheduler job {name} exceeded timeout={int(timeout)}s and was killed"
+            elif exit_code == -9:
+                error_type = f"{name}_sigkill"
+                message = f"scheduler job {name} received SIGKILL before scheduler timeout; check container resources and external API stalls"
             record_system_error(
                 source="scheduler",
                 level="error",
-                error_type=f"{name}_exit_{exit_code}",
-                message=f"scheduler job {name} failed with exit={exit_code}",
+                error_type=error_type,
+                message=message,
                 stack_trace=output_tail,
                 status_code=exit_code,
-                context={"job_name": name, "run_kind": item.run_kind, "trigger_id": item.trigger_id},
+                context={
+                    "job_name": name,
+                    "run_kind": item.run_kind,
+                    "trigger_id": item.trigger_id,
+                    "duration_ms": duration_ms,
+                    "timeout_seconds": int(timeout),
+                    "killed_by_timeout": timed_out,
+                },
             )
         update_runtime(
             name,

rules.yaml

@@ -303,6 +303,8 @@ event_driven:
   enabled: true
   poll_interval_min: 1
   decision_target_seconds: 60
+  max_process_events_per_run: 4
+  max_run_seconds: 50
   news_time_window_hours: 3
   max_event_age_hours: 6
   dedup_window_hours: 24

tests/test_event_driven_screener.py

@@ -90,6 +90,28 @@ def test_theme_expansion_spreads_ton_news_to_ecosystem_symbols():
     assert "主题扩散:ton_ecosystem" in by_symbol["DOGS/USDT"]["title"]
 
 
+def test_theme_expansion_respects_max_expanded_symbols(monkeypatch):
+    event = {
+        "source": "coingecko_trending",
+        "symbol": "TON/USDT",
+        "title": "Telegram becomes the main driver of the TON ecosystem and cuts TON fees",
+        "url": "https://example.com/ton",
+        "published_at": datetime.now(),
+        "importance": "B",
+        "event_type": "market_heat",
+        "raw": {},
+    }
+
+    original_cfg = ed._cfg()
+    limited_cfg = json.loads(json.dumps(original_cfg))
+    limited_cfg["theme_expansion"]["max_expanded_symbols"] = 2
+    monkeypatch.setattr(ed, "_cfg", lambda: limited_cfg)
+
+    expanded = ed.expand_theme_events([event])
+
+    assert len(expanded) <= 3  # 原始事件 + 最多 2 个扩散事件
+
+
 def test_wublock_atom_feed_events_are_parsed_and_symbolized(monkeypatch):
     xml = """<?xml version="1.0" encoding="utf-8"?>
     <feed xmlns="http://www.w3.org/2005/Atom">
@@ -186,3 +208,39 @@ def test_theme_static_accumulation_bonus_can_upgrade_to_recommend():
     assert result["score"] >= 6
     assert result["decision"] == "recommend"
     assert any("生态主题+强静K蓄力升权" in s for s in result["signals"])
+
+
+def test_run_once_caps_processed_events(monkeypatch):
+    events = [
+        {
+            "source": "unit",
+            "symbol": f"T{i}/USDT",
+            "title": f"event {i}",
+            "url": "",
+            "published_at": datetime.now(),
+            "importance": "A",
+            "event_type": "news",
+            "event_hash": f"h{i}",
+            "raw": {},
+        }
+        for i in range(6)
+    ]
+    processed = []
+
+    monkeypatch.setattr(ed, "init_db", lambda: None)
+    monkeypatch.setattr(ed, "init_event_tables", lambda: None)
+    monkeypatch.setattr(ed, "collect_events", lambda: events)
+    monkeypatch.setattr(ed, "store_events", lambda collected: collected)
+    monkeypatch.setattr(ed, "process_event", lambda event: processed.append(event["symbol"]) or {
+        "event": event,
+        "result": {"decision": "ignore", "score": 0, "reason": ""},
+        "rec_id": 0,
+        "pushed": False,
+    })
+    monkeypatch.setattr(ed, "log_cron_run", lambda **kwargs: None)
+
+    result = ed.run_once(limit=2, max_seconds=30)
+
+    assert result["processed_count"] == 2
+    assert result["skipped_due_to_limit"] == 4
+    assert processed == ["T0/USDT", "T1/USDT"]

tests/test_screener_optimizations.py

@@ -3,6 +3,7 @@ import sys
 from datetime import datetime
 
 import pandas as pd
+import requests
 
 PROJECT_DIR = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
 if PROJECT_DIR not in sys.path:
@@ -119,6 +120,33 @@ def test_fetch_all_tickers_filters_inactive_and_stale_markets(monkeypatch):
     assert any(x["symbol"] == "DEAD/USDT" and x["reason_code"] == "inactive_market" for x in exclusions)
 
 
+def test_fetch_spot_24h_tickers_uses_cache_when_dns_fails(monkeypatch, tmp_path):
+    monkeypatch.setattr(altcoin_screener, "EXCHANGE_CACHE_DIR", tmp_path)
+    cached_data = [
+        {
+            "symbol": "AIUSDT",
+            "lastPrice": "1.23",
+            "priceChangePercent": "8.5",
+            "quoteVolume": "1234567",
+            "highPrice": "1.4",
+            "lowPrice": "1.0",
+            "closeTime": 1770000000000,
+        }
+    ]
+    altcoin_screener._write_spot_24h_ticker_cache(cached_data)
+
+    def fail_get(*args, **kwargs):
+        raise requests.exceptions.ConnectionError("dns failed")
+
+    monkeypatch.setattr(altcoin_screener.requests, "get", fail_get)
+
+    tickers = altcoin_screener._fetch_spot_24h_tickers()
+
+    assert tickers["AI/USDT"]["last"] == 1.23
+    assert tickers["AI/USDT"]["percentage"] == 8.5
+    assert tickers["AI/USDT"]["quoteVolume"] == 1234567
+
+
 def _mock_weights():
     return {
         "量价齐飞": 5,