chore: opencode冗余清理 + LLM任务级模型选择 + systemd服务化

- 删除 opencode_search.py / mcp_search_server.py 及所有 MCP 引用 - 移除搜索缓存定时任务（scheduled_refresh_search_cache） - 清理前后端所有 opencode/MCP 代码和注释 - LLM 提供商量换：opencode-go→nvidia(默认)+sensenova(合规审查) - llm_configs 新增 is_default 字段，API 层互斥逻辑 - 所有定时任务支持独立 LLM 模型选择（LLM_TASK_PROVIDER env） - compliance_optimizer.py 修复：import os / 解硬编码 / 关键词过滤 - Scheduler 日志修复：始终 INSERT，避免僵尸 running 行 - Systemd 服务化：Restart=always / 单 worker / Type=exec - 搜索提供商：替换 opencode→360/搜狗/微信（免 Key） - 更新 AGENTS.md / PROGRESS.md
2026-06-02 15:38:16 +08:00
parent abbf1468bc
commit 499c511140
25 changed files with 451 additions and 678 deletions
@@ -9,6 +9,14 @@ from .auth import get_current_admin

 router = APIRouter(prefix="/api/admin/llmconfigs", tags=["admin"])

+def _apply_default_exclusive(config: LLMConfig, db: Session):
+    """当 config.is_default=True 时，将其他所有配置的 is_default 置为 False"""
+    if config.is_default:
+        db.query(LLMConfig).filter(LLMConfig.id != config.id).update(
+            {"is_default": False}, synchronize_session=False
+        )
+        db.flush()
+
@router.get("", response_model=List[LLMConfigResponse])
 def list_llm_configs(
    request: Request,
@@ -42,6 +50,8 @@ def create_llm_config(
    """创建 LLM 配置"""
    config = LLMConfig(**config_data.model_dump())
    db.add(config)
+    db.flush()
+    _apply_default_exclusive(config, db)
    db.commit()
    db.refresh(config)
    return config
@@ -61,6 +71,7 @@ def update_llm_config(
    update_data = config_update.model_dump(exclude_unset=True)
    for field, value in update_data.items():
        setattr(config, field, value)
+    _apply_default_exclusive(config, db)
    db.commit()
    db.refresh(config)
    return config
@@ -119,16 +119,6 @@ def test_provider(provider_id: int, data: dict = {}, db: Session = Depends(get_d
            if resp.status_code != 200:
                return {"ok": False, "error": f"HTTP {resp.status_code}: {resp.text[:200]}"}
            return {"ok": True, "results": resp.json().get("webPages", {}).get("value", [])[:3]}
-        elif p.provider_type == "mcp":
-            import subprocess, json as _json
-            mcp_script = Path(__file__).resolve().parent.parent.parent.parent.parent / "scripts" / "mcp_search_server.py"
-            r = subprocess.run(
-                [sys.executable, str(mcp_script), "--query", query],
-                capture_output=True, text=True, timeout=90,
-            )
-            if r.returncode != 0:
-                return {"ok": False, "error": f"子进程失败: {r.stderr[:200]}"}
-            return {"ok": True, "results": _json.loads(r.stdout)[:3]}
        return {"ok": False, "error": f"Unknown provider_type: {p.provider_type}"}
    except Exception as e:
        return {"ok": False, "error": str(e)}
@@ -280,28 +280,6 @@ def trigger_metrics_sync():
    except Exception as e:
        raise HTTPException(status_code=500, detail=str(e))

-@router.post("/refresh-search-cache/run")
-def trigger_refresh_search_cache(db: Session = Depends(get_db), current_user=Depends(get_current_user)):
-    try:
-        import sys as sys_mod
-        scripts_dir = PROJECT_ROOT / "scripts"
-        from ..database import SessionLocal as _ss
-        proc = subprocess.Popen(
-            [sys_mod.executable, str(scripts_dir / "opencode_search.py"), "--refresh-cache"],
-            stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True,
-            cwd=str(PROJECT_ROOT)
-        )
-        logger.info("Search cache refresh started (pid=%s)", proc.pid)
-        log = TaskLog(module_id="scheduled_refresh_search_cache", task_name="🔍 搜索缓存", status="running", message="搜索缓存刷新已启动", triggered_by="manual", started_at=datetime.now(timezone.utc), result_data={"pid": proc.pid})
-        db.add(log)
-        db.commit()
-        log_id = log.id
-        t = threading.Thread(target=_monitor_subprocess, args=(log_id, proc, "scheduled_refresh_search_cache", "🔍 搜索缓存", _ss), daemon=True)
-        t.start()
-        return {"message": "搜索缓存刷新已后台启动", "pid": proc.pid, "log_id": log_id}
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=str(e))
-
@router.post("/trends/run")
 def trigger_trends_refresh(db: Session = Depends(get_db), current_user=Depends(get_current_user)):
    try:
@@ -370,7 +348,6 @@ def get_modules_status(db: Session = Depends(get_db)):
    config_map = {c.module_id: c for c in configs}

    MODULE_META = {
-        "scheduled_refresh_search_cache": {"name": "🔍 搜索缓存", "cron": "01:00", "params_desc": {"refresh_queries": "搜索关键词列表"}},
        "scheduled_fetch_trends": {"name": "🔥 热点趋势", "cron": "01:10", "params_desc": {}},
        "scheduled_collect": {"name": "📡 内容采集", "cron": "01:30", "params_desc": {"max_topics": "最大选题数", "categories": "采集类别"}},
        "scheduled_generate": {"name": "🤖 内容创作", "cron": "02:00", "params_desc": {"auto_review": "自动合规审查"}},
@@ -11,14 +11,13 @@ from .auth import get_current_admin
 router = APIRouter(prefix="/api/admin/task-configs", tags=["admin"])

 DEFAULT_CONFIGS = {
-    "scheduled_refresh_search_cache": {"name": "🔍 搜索缓存", "cron": "01:00", "params": {}},
-    "scheduled_fetch_trends": {"name": "🔥 热点趋势", "cron": "01:10", "params": {}},
-    "scheduled_collect": {"name": "📡 内容采集", "cron": "01:30", "params": {"max_topics": 20}},
-    "scheduled_generate": {"name": "🤖 内容创作", "cron": "02:00", "params": {"auto_review": True}},
-    "scheduled_optimize": {"name": "🔍 合规审查", "cron": "03:00", "params": {"auto_pass_threshold": 80}},
-    "scheduled_optimize_sources": {"name": "📡 信息源优化", "cron": "05:00", "params": {}},
-    "scheduled_metrics_sync": {"name": "📊 指标同步", "cron": "06:00", "params": {}},
-    "scheduled_task_monitor": {"name": "⏰ 任务监控", "cron": "*", "params": {}},
+    "scheduled_fetch_trends": {"name": "🔥 热点趋势", "cron": "01:10", "params": {"llm_provider": "nvidia"}},
+    "scheduled_collect": {"name": "📡 内容采集", "cron": "01:30", "params": {"max_topics": 20, "llm_provider": "nvidia"}},
+    "scheduled_generate": {"name": "🤖 内容创作", "cron": "02:00", "params": {"auto_review": True, "llm_provider": "nvidia"}},
+    "scheduled_optimize": {"name": "🔍 合规审查", "cron": "03:00", "params": {"auto_pass_threshold": 80, "llm_provider": "sensenova"}},
+    "scheduled_optimize_sources": {"name": "📡 信息源优化", "cron": "05:00", "params": {"llm_provider": "nvidia"}},
+    "scheduled_metrics_sync": {"name": "📊 指标同步", "cron": "06:00", "params": {"llm_provider": "nvidia"}},
+    "scheduled_task_monitor": {"name": "⏰ 任务监控", "cron": "*", "params": {"llm_provider": "nvidia"}},
 }

 def _attach_last_log(resp: TaskConfigResponse, db: Session, module_id: str) -> TaskConfigResponse:
@@ -12,7 +12,6 @@ from .auth import get_current_admin
 router = APIRouter(prefix="/api/admin/task-logs", tags=["admin"])

 MODULES = {
-    "scheduled_refresh_search_cache": "🔍 搜索缓存",
    "scheduled_fetch_trends": "🔥 热点趋势",
    "scheduled_collect": "📡 内容采集",
    "scheduled_generate": "🤖 内容创作",
@@ -79,7 +78,6 @@ def list_log_types(db: Session = Depends(get_db), admin_user=Depends(get_current
    for mid, name in MODULES.items():
        if mid in used_ids or True:
            log_file_map = {
-                "scheduled_refresh_search_cache": "opencode_search",
                "scheduled_fetch_trends": "trends",
                "scheduled_collect": "collector",
                "scheduled_generate": "creator",
@@ -236,7 +236,6 @@ def _get_module_detail_data(module_id: str, db, ROOT, DATA_DIR, LOGS_DIR, today_
    import json as json_mod
    import re as re_mod
    MODULE_META = {
-        "scheduled_refresh_search_cache": {"name": "🔍 搜索缓存", "description": "通过 opencode webfetch 联网搜索，刷新 8 个分类的搜索缓存，供内容采集器使用"},
        "scheduled_fetch_trends": {"name": "🔥 热点趋势", "description": "从百度、微博、知乎实时热搜 API 抓取当天热点，LLM 补充，存入 trends.json"},
        "scheduled_collect": {"name": "📡 内容采集", "description": "读取搜索缓存 + 热点趋势 + 历史表现 + AI 建议，经 LLM 分析后生成选题"},
        "scheduled_generate": {"name": "🤖 内容创作", "description": "基于选题，LLM 生成三平台文章（知乎、微信、小红书），存入 articles 表"},
@@ -253,23 +252,7 @@ def _get_module_detail_data(module_id: str, db, ROOT, DATA_DIR, LOGS_DIR, today_
    outputs = {}
    history = []

-    if module_id == "scheduled_refresh_search_cache":
-        cache_file = DATA_DIR / "search_cache.json"
-        if cache_file.exists():
-            try:
-                cache = json_mod.loads(cache_file.read_text(encoding="utf-8"))
-                meta_ = cache.pop("_metadata", {})
-                for q, results in cache.items():
-                    inputs.setdefault("搜索词", []).append(q)
-                    outputs.setdefault("各分类结果", []).append({
-                        "query": q, "count": len(results),
-                        "samples": [r.get("title","")[:50] for r in results[:3]]
-                    })
-                outputs["更新时间"] = meta_.get("updated_at", "")
-                outputs["结果总数"] = sum(len(v) for v in cache.values())
-            except Exception:
-                pass
-        # try reading queries from yaml
+    # try reading queries from yaml
        try:
            import yaml
            cfg_path = ROOT / "config" / "sources.yaml"
@@ -390,7 +373,6 @@ def _get_module_detail_data(module_id: str, db, ROOT, DATA_DIR, LOGS_DIR, today_

    # History from log files
    log_map = {
-        "scheduled_refresh_search_cache": LOGS_DIR / f"opencode_search_{today_str}.log",
        "scheduled_fetch_trends": LOGS_DIR / f"trends_{today_str}.log",
        "scheduled_collect": LOGS_DIR / f"collector_{today_str}.log",
        "scheduled_generate": LOGS_DIR / f"creator_{today_str}.log",