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

platform/backend/app/api/tasks.py

@@ -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",