性能优化: 任务列表瘦身+详情分页+auto状态独立存储+磁盘统计缓存+db异步同步; 统计数值未读取时显示-

- /api/tasks 响应 1.57MB -> 12KB (latest_run 只带摘要, 一次读 runs.json)
- /api/stats 磁盘占用加 30s 缓存, 去除重复全量读
- 详情接口分页返回 results (默认100条/页), 前端表格分页+页码跳转
- auto 任务 pending/visited 队列迁移到 data/auto_state/ 独立文件 (tasks.json 1.6MB -> 14KB)
- MySQL 同步改后台线程异步执行 (db.sync_run_async/upsert_task_async), 不再阻塞爬虫和 API
- 启动时自动迁移存量 auto 状态
- 统计/回收站/运行中数值在未读到真实数据前显示 '-'

README.md

@@ -49,12 +49,14 @@
 - 可启用/停用调度，自动计算下次执行时间；到点自动开跑，跑完自动计算下一次
 
 ### 4. 自动爬取模式
-给定一个起始网址，系统自动从页面里发现链接、按规则筛选后 BFS 爬取：
+给定一个起始网址，系统**持续递归**爬取：爬取页面 → 自动发现符合规则的链接 → 继续爬取 → 继续发现……直到**无新链接可爬**时自动结束。
 - **🧪 试爬取**：先用起始网址试跑一次，展示规则筛选后的链接清单（将爬取哪些、排除哪些及原因），确认规则符合预期后再正式开爬
+- **最大爬取深度**：0=无限制（默认），N=只爬 N 层
+- **最大页数**：0=无限制，默认 1000 作安全上限（防止动态无限链接的站点失控）
+- **🔄 缓存续爬**：提取但未爬取的链接自动存入缓存队列（连同已爬集合一并持久化），任务停止/中断后再次运行，从缓存队列**继续爬取**，已爬过的不会重复；起始网址每次运行都重新爬取（不去重）；规则变更后可点「🧹 清空缓存」重新开始
 - **包含规则**：只爬包含指定子串（或正则）的链接
 - **排除规则**：跳过匹配的链接（如 login、/tag/）
 - **仅同域名**：限制在起始网站内
-- **最大页数 / 最大深度**：控制爬取规模
 - 其余参数（间隔、重试、图片、通知）同批量模式
 
 ### 5. 资源操作信息（元数据）
@@ -64,7 +66,7 @@
 - 详情页结果表中点「📋 元数据」即可在线查看
 
 ### 6. 数据库记录（MySQL）
-任务、运行记录、爬取结果实时写入 MySQL（`121.40.164.32:16006`，库 `crawler`），**网页完整内容不入库**（存磁盘文件），库中只存标题、网址、状态、文件路径等元数据：
+任务、运行记录、爬取结果实时写入 MySQL（`121.40.164.32:16006`，账号 `uni_crawler`，库 `uni_crawler`），**网页完整内容不入库**（存磁盘文件），库中只存标题、网址、状态、文件路径等元数据：
 - `crawl_tasks` — 任务信息（含回收站标记 deleted_at）
 - `crawl_runs` — 每次运行记录（状态/进度/成功失败数/图片数/时间）
 - `crawl_results` — 每页一条（**status: OK/FAIL** 成功失败标记、标题、网址、来源链接、深度、错误信息、文件路径、图片数）

app.py

@@ -5,6 +5,7 @@
 """
 import os
 import threading
+import time
 from datetime import datetime
 
 from flask import Flask, jsonify, request, send_file, send_from_directory
@@ -95,7 +96,7 @@ def persist_cb(task_id, run):
     run["progress"]["percent"] = round(done * 100 / total) if total else 0
     store.save_run(task_id, run)
     try:
-        db.sync_run(run, lambda tid, r: store.save_run(tid, r))
+        db.sync_run_async(run)      # 后台异步同步, 不再阻塞爬虫线程
     except Exception as e:
         print(f"[db] 同步失败: {e}", flush=True)
 
@@ -111,7 +112,7 @@ def start_run(task):
         job = CrawlJob(task, run, persist_cb)
         JOBS[task["id"]] = job
         job.start()
-        db.upsert_task(task)              # 确保任务在库中
+        db.upsert_task_async(task)        # 确保任务在库中 (异步)
         return run, None
 
 
@@ -147,12 +148,37 @@ def api_status():
 
 # ---------------- API: 任务 ----------------
 
+def _strip_auto_state(task):
+    """剥离任务对象中的待爬/已爬队列 (已独立存储, 避免大 JSON 拖慢接口)"""
+    if task.get("mode") == "auto" and task.get("auto"):
+        task["auto"] = {k: v for k, v in task["auto"].items()
+                         if k not in ("pending", "visited")}
+    return task
+
+
+def _run_summary(run):
+    """运行记录摘要 (不含 results/logs), 供列表/详情页使用"""
+    return {
+        "id": run["id"],
+        "task_id": run.get("task_id", ""),
+        "mode": run.get("mode", ""),
+        "status": run.get("status", ""),
+        "progress": run.get("progress", {}),
+        "started_at": run.get("started_at", ""),
+        "finished_at": run.get("finished_at", ""),
+        "stats": run.get("stats", {}),
+        "out_dir": run.get("out_dir", ""),
+    }
+
+
 @app.route("/api/tasks", methods=["GET"])
 def api_tasks():
     tasks = [t for t in store.load_tasks() if not t.get("deleted_at")]
+    runs_map = store.load_runs_map()      # 只全量读一次
     for t in tasks:
-        runs = store.get_runs(t["id"])
-        t["latest_run"] = runs[-1] if runs else None
+        _strip_auto_state(t)
+        runs = runs_map.get(t["id"], [])
+        t["latest_run"] = _run_summary(runs[-1]) if runs else None
         with JOBS_LOCK:
             job = JOBS.get(t["id"])
             t["running"] = bool(job and job.is_running())
@@ -220,7 +246,7 @@ def api_create_task():
             return jsonify({"error": "请至少填写一个网址"}), 400
 
     store.upsert_task(task)
-    db.upsert_task(task)
+    db.upsert_task_async(task)
     return jsonify(task), 201
 
 
@@ -229,11 +255,59 @@ def api_task_detail(tid):
     task = store.get_task(tid)
     if not task:
         return jsonify({"error": "任务不存在"}), 404
-    task["runs"] = list(reversed(store.get_runs(tid)))
+    # 分页参数: 指定 run + 页码 (默认最新 run 第 1 页)
+    rid = request.args.get("run", "")
+    try:
+        page = max(int(request.args.get("page", 1)), 1)
+    except ValueError:
+        page = 1
+    try:
+        page_size = min(max(int(request.args.get("page_size", 100)), 10), 500)
+    except ValueError:
+        page_size = 100
+
+    runs = store.get_runs(tid)
+    if rid:
+        cur = next((r for r in runs if r["id"] == rid), None)
+        if cur is None:
+            cur = runs[-1] if runs else None   # 指定的 run 不存在时回退最新
+    else:
+        cur = runs[-1] if runs else None
     with JOBS_LOCK:
         job = JOBS.get(tid)
         task["running"] = bool(job and job.is_running())
-    return jsonify(task)
+
+    result = _strip_auto_state(dict(task))
+    result["runs"] = [_run_summary(r) for r in reversed(runs)]
+    result["run"] = None
+    result["run_id"] = ""
+    result["run_total"] = 0
+    result["run_page"] = 1
+    result["run_pages"] = 1
+    if cur:
+        results = cur.get("results", [])
+        total = len(results)
+        pages = max((total + page_size - 1) // page_size, 1)
+        page = min(page, pages)
+        start = (page - 1) * page_size
+        snap = {k: v for k, v in cur.items() if k != "logs"}   # logs 走独立轮询接口
+        snap["results"] = results[start:start + page_size]
+        snap["results_total"] = total
+        snap["run_page"] = page
+        snap["run_pages"] = pages
+        snap["run_page_size"] = page_size
+        result["run"] = snap
+        result["run_id"] = cur["id"]
+        result["run_total"] = total
+        result["run_page"] = page
+        result["run_pages"] = pages
+        result["run_page_size"] = page_size
+    # auto 状态数量 (独立文件)
+    if task.get("mode") == "auto":
+        st = store.load_auto_state(tid, task)
+        result["auto_pending_count"] = len(st.get("pending", []))
+        result["auto_visited_count"] = len(st.get("visited", []))
+    return jsonify(result)
 
 
 @app.route("/api/tasks/<tid>", methods=["PUT"])
@@ -267,7 +341,7 @@ def api_update_task(tid):
             task["schedule"] = sch
         task["updated_at"] = now_str()
         store.upsert_task(task)
-        db.upsert_task(task)
+        db.upsert_task_async(task)
     return jsonify(task)
 
 
@@ -283,7 +357,7 @@ def api_delete_task(tid):
     with JOBS_LOCK:
         JOBS.pop(tid, None)
     store.soft_delete_task(tid, now_str())
-    db.upsert_task(store.get_task(tid))
+    db.upsert_task_async(store.get_task(tid))
     return jsonify({"ok": True, "msg": "已移入回收站"})
 
 
@@ -306,8 +380,9 @@ def _purge_out_dir(out_dir):
 @app.route("/api/trash", methods=["GET"])
 def api_trash_list():
     items = store.list_trash()
+    runs_map = store.load_runs_map()
     for t in items:
-        t["runs_count"] = len(store.get_runs(t["id"]))
+        t["runs_count"] = len(runs_map.get(t["id"], []))
         t["out_dir"] = resolve_out_dir(t)
     return jsonify(items)
 
@@ -318,7 +393,7 @@ def api_trash_restore(tid):
     if not task or not task.get("deleted_at"):
         return jsonify({"error": "任务不在回收站中"}), 404
     store.restore_task(tid)
-    db.upsert_task(store.get_task(tid))
+    db.upsert_task_async(store.get_task(tid))
     return jsonify({"ok": True, "msg": "已恢复"})
 
 
@@ -386,13 +461,20 @@ def api_resume(tid):
 
 # ---------------- API: 统计 ----------------
 
+# 磁盘占用统计缓存 (避免每次轮询都逐个 stat 文件)
+_DISK_CACHE = {"ts": 0.0, "mb": 0.0}
+DISK_CACHE_TTL = 30   # 秒
+
+
 @app.route("/api/stats")
 def api_stats():
-    tasks = [t for t in store.load_tasks() if not t.get("deleted_at")]
-    trash_count = sum(1 for t in store.load_tasks() if t.get("deleted_at"))
+    tasks = store.load_tasks()
+    active = [t for t in tasks if not t.get("deleted_at")]
+    trash_count = len(tasks) - len(active)
+    runs_map = store.load_runs_map()      # 只全量读一次
     total_runs = ok = fail = imgs = 0
-    for t in tasks:
-        for r in store.get_runs(t["id"]):
+    for t in active:
+        for r in runs_map.get(t["id"], []):
             total_runs += 1
             st = r.get("stats") or {}
             ok += st.get("ok", 0)
@@ -400,10 +482,11 @@ def api_stats():
             imgs += st.get("images", 0)
     with JOBS_LOCK:
         running = sum(1 for j in JOBS.values() if j.is_running())
-    # 统计各任务输出目录的磁盘占用
+    now = time.time()
+    if now - _DISK_CACHE["ts"] > DISK_CACHE_TTL:
         size = 0
         seen = set()
-    for t in tasks:
+        for t in active:
             d = os.path.realpath(resolve_out_dir(t))
             if d in seen or not os.path.isdir(d):
                 continue
@@ -414,14 +497,15 @@ def api_stats():
                         size += os.path.getsize(os.path.join(root, f))
                     except OSError:
                         pass
+        _DISK_CACHE.update(ts=now, mb=round(size / 1048576, 1))
     return jsonify({
-        "tasks": len(tasks),
+        "tasks": len(active),
         "running": running,
         "runs": total_runs,
         "ok": ok,
         "fail": fail,
         "images": imgs,
-        "disk_mb": round(size / 1048576, 1),
+        "disk_mb": _DISK_CACHE["mb"],
         "trash": trash_count,
     })
 
@@ -448,6 +532,28 @@ def api_probe():
     return jsonify(result)
 
 
+@app.route("/api/tasks/<tid>/clear_cache", methods=["POST"])
+def api_clear_cache(tid):
+    """清空自动任务的待爬缓存队列与已爬集合 (规则变更后重新开始用)"""
+    task = store.get_task(tid)
+    if not task:
+        return jsonify({"error": "任务不存在"}), 404
+    if task.get("mode") != "auto":
+        return jsonify({"error": "仅自动爬取任务支持清空缓存"}), 400
+    with JOBS_LOCK:
+        job = JOBS.get(tid)
+        if job and job.is_running():
+            return jsonify({"error": "任务正在运行，无法清空缓存"}), 409
+    auto = task.setdefault("auto", {})
+    auto["pending"] = []
+    auto["visited"] = []
+    store.clear_auto_state(tid)          # 同时清空独立状态文件
+    task["updated_at"] = now_str()
+    store.upsert_task(task)
+    db.upsert_task_async(task)
+    return jsonify({"ok": True, "msg": "缓存队列已清空"})
+
+
 @app.route("/api/tasks/<tid>/probe", methods=["POST"])
 def api_task_probe(tid):
     """对已保存的自动任务执行试爬取 (使用保存的规则)"""
@@ -479,6 +585,7 @@ def api_search():
     ql = q.lower()
     results = []
     seen = set()
+    runs_map = store.load_runs_map()      # 只全量读一次
     for task in store.load_tasks():
         if task.get("deleted_at"):
             continue                       # 回收站任务不参与搜索
@@ -488,7 +595,7 @@ def api_search():
             or any(ql in u.lower() for u in task.get("urls", []))
             or ql in ((auto.get("seed_url") or "").lower())
         )
-        for run in store.get_runs(task["id"]):
+        for run in runs_map.get(task["id"], []):
             for r in run.get("results", []):
                 url = r.get("url") or ""
                 title = r.get("title") or ""
@@ -572,13 +679,35 @@ def api_file():
 scheduler = Scheduler(start_run)
 
 
+def migrate_auto_state():
+    """启动时把旧任务对象里的 auto pending/visited 迁移到独立文件, 给 tasks.json 瘦身"""
+    migrated = 0
+    for t in store.load_tasks():
+        if t.get("mode") != "auto":
+            continue
+        auto = t.get("auto") or {}
+        if auto.get("pending") or auto.get("visited"):
+            store.save_auto_state(t["id"], {
+                "pending": auto.get("pending", []),
+                "visited": auto.get("visited", []),
+            })
+            auto.pop("pending", None)
+            auto.pop("visited", None)
+            t["updated_at"] = now_str()
+            store.upsert_task(t)
+            migrated += 1
+    if migrated:
+        print(f"[universal-crawler] 已迁移 {migrated} 个 auto 任务的队列状态到独立文件", flush=True)
+
+
 def main():
     os.makedirs(os.path.join(HERE, "data"), exist_ok=True)
     os.makedirs(os.path.join(HERE, "out"), exist_ok=True)
     db.init_db()
-    # 全量同步存量任务到数据库
+    migrate_auto_state()
+    # 全量同步存量任务到数据库 (异步, 不阻塞启动)
     for t in store.load_tasks():
-        db.upsert_task(t)
+        db.upsert_task_async(t)
     scheduler.start()
     print(f"[universal-crawler] 启动完成, 管理界面: http://0.0.0.0:{PORT}/")
     app.run(host="0.0.0.0", port=PORT, threaded=True, debug=False)

backfill.py

@@ -0,0 +1,26 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+历史爬取记录回填数据库 (幂等, 可重复执行)
+
+用法:
+    python backfill.py                  # 回填所有任务的历史记录
+    python backfill.py <task_id> ...    # 只回填指定任务
+
+说明:
+    - 将本地 data/ 中的任务/运行/爬取结果(含成功与失败)全量写入 MySQL
+    - 网页内容不入库, 只写元数据; 重复执行不会产生重复记录
+"""
+import sys
+
+import db
+import store
+
+if __name__ == "__main__":
+    ids = [a for a in sys.argv[1:] if a.strip()] or None
+    db.init_db()
+    stats = db.sync_all_history(ids)
+    if ids:
+        print(f"[backfill] 已回填 {len(ids)} 个任务: {stats}")
+    else:
+        print(f"[backfill] 已回填全部任务: {stats}")

db.py

@@ -5,21 +5,23 @@ MySQL 记录层: 任务 / 运行 / 爬取结果 写入数据库
 - 爬取成功/失败均有 status 标记 (OK / FAIL), 运行记录有 run 状态标记
 - 所有操作容错: 数据库不可用时不影响爬取主流程 (仅打印日志)
 """
+import copy
 import json
 import os
 import time
+from concurrent.futures import ThreadPoolExecutor
 
 import pymysql
 
 DB_CONFIG = dict(
     host="121.40.164.32",
     port=16006,
-    user="root",
-    password="hz_123",
+    user="uni_crawler",
+    password="wleD6x2T",
     charset="utf8mb4",
 )
 
-DB_NAME = "crawler"
+DB_NAME = "uni_crawler"
 
 DDL = [
     """
@@ -67,9 +69,7 @@ DDL = [
       source_url VARCHAR(2000),
       depth INT,
       attempts INT DEFAULT 1,
-      html_file VARCHAR(500),
-      txt_file VARCHAR(500),
-      meta_file VARCHAR(500),
+      base_file VARCHAR(500),
       image_count INT DEFAULT 0,
       image_files TEXT,
       UNIQUE KEY uk_run_url (run_id, url(500))
@@ -78,6 +78,17 @@ DDL = [
 ]
 
 
+def _migrate(conn):
+    """存量表结构迁移: 合并 html_file/txt_file/meta_file 为 base_file"""
+    with conn.cursor() as cur:
+        cur.execute("SHOW COLUMNS FROM crawl_results LIKE 'html_file'")
+        if cur.fetchone():
+            cur.execute("ALTER TABLE crawl_results ADD COLUMN base_file VARCHAR(500) NULL AFTER meta_file")
+            cur.execute("UPDATE crawl_results SET base_file = REPLACE(html_file, '.html', '') WHERE base_file IS NULL")
+            cur.execute("ALTER TABLE crawl_results DROP COLUMN html_file, DROP COLUMN txt_file, DROP COLUMN meta_file")
+            print("[db] 表结构迁移完成: html_file/txt_file/meta_file -> base_file")
+
+
 def _conn():
     cfg = dict(DB_CONFIG)
     cfg["database"] = DB_NAME
@@ -92,6 +103,39 @@ def _safe(fn, *args, **kwargs):
         print(f"[db] 操作失败: {e}", flush=True)
 
 
+# ---------------- 异步同步 (避免远程 MySQL 阻塞爬虫/API 主流程) ----------------
+
+_db_executor = ThreadPoolExecutor(max_workers=2, thread_name_prefix="db-sync")
+
+
+def upsert_task_async(task):
+    """异步写入/更新任务 (不阻塞调用方)"""
+    try:
+        snap = copy.deepcopy(task)
+        _db_executor.submit(upsert_task, snap)
+    except Exception as e:
+        print(f"[db] 异步任务同步失败: {e}", flush=True)
+
+
+def sync_run_async(run):
+    """异步同步运行记录+爬取结果; 完成后把 _db_count 回写到主线程 run 对象"""
+    try:
+        snap = copy.deepcopy(run)
+
+        def _work():
+            upsert_run(snap)
+            results = snap.get("results", [])
+            synced = snap.get("_db_count", 0)
+            if len(results) > synced:
+                insert_results(snap, results[synced:])
+                snap["_db_count"] = len(results)
+                run["_db_count"] = snap["_db_count"]
+
+        _db_executor.submit(_work)
+    except Exception as e:
+        print(f"[db] 异步运行同步失败: {e}", flush=True)
+
+
 def init_db():
     """建库建表 (幂等)"""
     try:
@@ -100,6 +144,7 @@ def init_db():
             cur.execute(f"CREATE DATABASE IF NOT EXISTS `{DB_NAME}` DEFAULT CHARACTER SET utf8mb4")
         conn.close()
         conn = _conn()
+        _migrate(conn)
         with conn.cursor() as cur:
             for ddl in DDL:
                 cur.execute(ddl)
@@ -219,8 +264,21 @@ def upsert_run(run):
 
 # ---------------- 爬取结果 (每页一条, 成功失败均记录) ----------------
 
+def _base_of(entry):
+    """从结果条目提取基础文件名 (html/txt/meta 三个后缀共用同一前缀)"""
+    for f in (entry.get("meta_file"), entry.get("html_file"), entry.get("txt_file")):
+        f = f or ""
+        if f.endswith(".meta.json"):
+            return f[:-10]
+        if f.endswith(".html"):
+            return f[:-5]
+        if f.endswith(".txt"):
+            return f[:-4]
+    return ""
+
+
 def insert_results(run, results):
-    """批量插入爬取结果 (增量)"""
+    """批量插入爬取结果 (增量); 文件只记基础名 base_file"""
     if not results:
         return
 
@@ -234,7 +292,7 @@ def insert_results(run, results):
                 r.get("status", "FAIL"), r.get("error"),
                 _dt(r.get("crawl_time")), (r.get("source_url") or "")[:2000],
                 r.get("depth"), r.get("attempts", 1),
-                r.get("html_file", ""), r.get("txt_file", ""), r.get("meta_file", ""),
+                _base_of(r),
                 len(r.get("images", []) or []),
                 _j([im.get("file") for im in (r.get("images") or [])]),
             ))
@@ -243,8 +301,8 @@ def insert_results(run, results):
                 """INSERT IGNORE INTO crawl_results
                    (run_id, task_id, mode, url, title, status, error,
                     crawl_time, source_url, depth, attempts,
-                    html_file, txt_file, meta_file, image_count, image_files)
-                   VALUES (%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s)""",
+                    base_file, image_count, image_files)
+                   VALUES (%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s,%s)""",
                 rows,
             )
         conn.close()
@@ -262,3 +320,30 @@ def sync_run(run, persist):
         insert_results(run, results[synced:])
         run["_db_count"] = len(results)
         persist(run.get("task_id"), run)
+
+
+# ---------------- 历史数据回填 (幂等) ----------------
+
+def sync_all_history(task_ids=None):
+    """把本地 JSON 中的历史任务/运行/结果全量回填数据库
+    可重复执行 (INSERT IGNORE + 唯一键去重)
+    task_ids: 指定只回填的任务ID列表, 默认全部
+    返回统计 dict
+    """
+    import store as _store
+    stats = {"tasks": 0, "runs": 0, "results": 0}
+    for task in _store.load_tasks():
+        if task_ids and task["id"] not in task_ids:
+            continue
+        upsert_task(task)
+        stats["tasks"] += 1
+        for run in _store.get_runs(task["id"]):
+            upsert_run(run)
+            stats["runs"] += 1
+            results = run.get("results", [])
+            if results:
+                insert_results(run, results)
+                stats["results"] += len(results)
+                run["_db_count"] = len(results)
+                _store.save_run(task["id"], run)   # 记录已同步标记, 避免运行中重复插入
+    return stats

engine.py

@@ -21,6 +21,7 @@ from playwright_stealth import Stealth
 
 import notify
 import store
+import db
 
 HERE = os.path.dirname(os.path.abspath(__file__))
 DATA_DIR = os.path.join(HERE, "data")
@@ -567,33 +568,49 @@ class CrawlJob:
         return included
 
     def _crawl_auto(self):
+        """自动爬取: 持续递归 爬取->发现链接->爬取... 直到无新链接可爬或达到上限
+        - max_depth: 0=无限制, N=只爬 N 层
+        - max_pages: 0=无限制, N=安全上限
+        - 提取但未爬取的链接持久化到 auto.pending, 已爬集合持久化到 auto.visited
+        - 停止后再次运行从缓存队列继续爬; 起始网址每次运行都重新爬(不去重)
+        """
         run = self.run
         auto = self.task.get("auto", {})
         seed = auto.get("seed_url", "")
-        max_pages = int(auto.get("max_pages", 50) or 50)
-        max_depth = int(auto.get("max_depth", 2) or 2)
-        run["progress"]["total"] = max_pages
+        max_pages = int(auto.get("max_pages", 1000) or 0)     # 0 = 无限制
+        max_depth = int(auto.get("max_depth", 0) or 0)        # 0 = 无限制
         out_dir = self._resolve_out_dir()
         run["out_dir"] = out_dir
         os.makedirs(out_dir, exist_ok=True)
+
+        # 恢复持久化状态: 已爬集合 + 上次未爬完的缓存队列 (独立状态文件, 不撑大 tasks.json)
+        state = store.load_auto_state(self.task["id"], self.task)
+        visited = set(state.get("visited", []) or [])
+        pending = state.get("pending", []) or []
+        # 起始网址每次运行都爬(不做去重), 缓存队列继续消费
+        queue = [(seed, 0, "")]
+        if pending:
+            queue.extend((p["url"], p.get("depth", 0), p.get("source", "")) for p in pending)
+        queued = set(visited)
+        for u, _d, _s in queue:
+            queued.add(normalize_url(u))
+
+        run["progress"]["total"] = len(queue)
         self._persist()
 
         p, browser, ctx, page, cookie_file = self._open_browser()
-        queue = [(seed, 0, "")]               # (url, depth, 来源链接)
-        visited = set()                        # 规范化 URL 去重
-        queued = set([normalize_url(seed)])
-        idx = 0
         try:
             while queue and not self._stop.is_set():
                 self._wait_if_paused()
+                if max_pages > 0 and len(visited) >= max_pages:
+                    self._log("info", f"达到最大页数上限 {max_pages}, 停止")
+                    break
                 url, depth, src = queue.pop(0)
                 key = normalize_url(url)
-                if key in visited:
+                if key in visited and url != seed:     # 起始网址不去重, 其余已爬跳过
                     continue
-                if len(visited) >= max_pages:
-                    break
                 visited.add(key)
-                idx += 1
+                idx = len(visited)
                 run["progress"]["current_url"] = url
                 run["progress"]["done"] = len(visited)
                 self._persist()
@@ -602,7 +619,8 @@ class CrawlJob:
                 run["results"].append(entry)
                 self._bump_stats(entry)
                 self._persist()
-                if entry["status"] == "OK" and depth < max_depth:
+                # 无深度限制或未达深度限制时持续发现链接
+                if entry["status"] == "OK" and (max_depth == 0 or depth < max_depth):
                     for link in self._discover_links(page):
                         lk = normalize_url(link)
                         if lk not in visited and lk not in queued:
@@ -610,7 +628,19 @@ class CrawlJob:
                             queue.append((link, depth + 1, url))
                 if entry["status"] == "OK":
                     self._delay()
+            run["progress"]["total"] = len(visited)
+            if not self._stop.is_set() and len(queue) == 0:
+                self._log("info", f"无新链接可爬, 任务结束 (共 {len(visited)} 页)")
         finally:
             self._close_browser(p, browser, ctx, cookie_file)
-        run["progress"]["total"] = len(visited)
+            # 无论完成/停止/异常, 都保存缓存队列与已爬集合, 便于下次继续
+            try:
+                store.save_auto_state(self.task["id"], {
+                    "pending": [
+                        {"url": u, "depth": d, "source": s} for u, d, s in queue],
+                    "visited": list(visited),
+                })
+                db.upsert_task_async(self.task)
+            except Exception:
+                pass
         self._persist()

static/app.js

@@ -45,26 +45,32 @@ async function loadTasks() {
     pill.textContent = `运行中任务: ${running}`;
     pill.classList.toggle("active", running > 0);
     const st = await api("/api/status");
-    $("version").textContent = `v${st.version}`;
+    $("version").textContent = st && st.version ? `v${st.version}` : "v-";
   } catch (e) {
     toast("加载任务失败: " + e.message, true);
   }
   loadStats();
 }
 
+const STAT_IDS = ["stTasks", "stRunning", "stRuns", "stOk", "stFail", "stImgs", "stDisk"];
+
 async function loadStats() {
+  // 未读到真实数据前, 一律显示 "-"
+  STAT_IDS.forEach((id) => { $(id).textContent = "-"; });
+  $("trashCount").textContent = "-";
   try {
     const s = await api("/api/stats");
-    $("stTasks").textContent = s.tasks;
-    $("stRunning").textContent = s.running;
-    $("stRuns").textContent = s.runs;
-    $("stOk").textContent = s.ok;
-    $("stFail").textContent = s.fail;
-    $("stImgs").textContent = s.images;
-    $("stDisk").textContent = s.disk_mb >= 1024
-      ? (s.disk_mb / 1024).toFixed(1) + " GB" : s.disk_mb + " MB";
-    $("trashCount").textContent = s.trash || 0;
-  } catch (e) { /* 统计失败忽略 */ }
+    $("stTasks").textContent = s.tasks ?? "-";
+    $("stRunning").textContent = s.running ?? "-";
+    $("stRuns").textContent = s.runs ?? "-";
+    $("stOk").textContent = s.ok ?? "-";
+    $("stFail").textContent = s.fail ?? "-";
+    $("stImgs").textContent = s.images ?? "-";
+    const mb = Number(s.disk_mb);
+    $("stDisk").textContent = Number.isFinite(mb) && s.disk_mb != null
+      ? (mb >= 1024 ? (mb / 1024).toFixed(1) + " GB" : mb + " MB") : "-";
+    $("trashCount").textContent = s.trash ?? "-";
+  } catch (e) { /* 统计失败保持 "-" */ }
 }
 
 function taskStatusBadge(t) {
@@ -82,7 +88,7 @@ function taskStatusBadge(t) {
 const GROUPS = [
   { key: "batch", label: "📄 批量爬取", desc: "一次性爬取指定网址列表" },
   { key: "scheduled", label: "⏰ 定时爬取", desc: "按间隔或 cron 表达式定时执行" },
-  { key: "auto", label: "🤖 自动爬取", desc: "从起始网址自动发现链接并爬取" },
+  { key: "auto", label: "🤖 自动爬取", desc: "从起始网址自动发现链接，持续递归爬取到无新链接为止" },
 ];
 
 function renderTasks() {
@@ -434,8 +440,8 @@ async function openEdit(tid) {
     f.elements["exclude"].value = (t.auto.exclude || []).join("\n");
     f.elements["same_domain"].checked = t.auto.same_domain !== false;
     f.elements["use_regex"].checked = !!t.auto.use_regex;
-    f.elements["max_pages"].value = t.auto.max_pages ?? 50;
-    f.elements["max_depth"].value = t.auto.max_depth ?? 2;
+    f.elements["max_pages"].value = t.auto.max_pages ?? 1000;
+    f.elements["max_depth"].value = t.auto.max_depth ?? 0;
   }
   syncScheduleUI();
   $("formHint").textContent = t.running
@@ -481,8 +487,8 @@ async function submitForm(e) {
       exclude: splitLines(f.elements["exclude"].value),
       same_domain: f.elements["same_domain"].checked,
       use_regex: f.elements["use_regex"].checked,
-      max_pages: parseInt(f.elements["max_pages"].value) || 50,
-      max_depth: parseInt(f.elements["max_depth"].value) || 2,
+      max_pages: parseInt(f.elements["max_pages"].value) || 0,
+      max_depth: parseInt(f.elements["max_depth"].value) || 0,
     };
   }
   if (mode === "scheduled") {
@@ -516,13 +522,15 @@ async function submitForm(e) {
 }
 
 /* ---------------- 详情 ---------------- */
+const DETAIL_PAGE_SIZE = 100;   // 详情结果每页条数
+
 async function openDetail(tid) {
   try {
-    const t = await api(`/api/tasks/${tid}`);
+    const t = await api(`/api/tasks/${tid}?page_size=${DETAIL_PAGE_SIZE}`);
     state.detail.task = t;
     const runs = t.runs || [];
-    state.detail.runId = runs.find((r) => r.status === "running")
-      ? runs[0].id : (runs[0] ? runs[0].id : null);
+    const cur = t.run;
+    state.detail.runId = cur ? cur.id : (runs[0] ? runs[0].id : null);
     state.detail.logOffset = 0;
     $("detailTitle").textContent = `任务详情 · ${t.name}`;
     renderDetail();
@@ -531,11 +539,38 @@ async function openDetail(tid) {
   } catch (e) { toast(e.message, true); }
 }
 
+async function loadRunPage(tid, rid, page) {
+  return api(`/api/tasks/${tid}?run=${rid}&page=${page}&page_size=${DETAIL_PAGE_SIZE}`);
+}
+
+async function selectRun(rid) {
+  const t = state.detail.task;
+  if (!t) return;
+  try {
+    const d = await loadRunPage(t.id, rid, 1);
+    state.detail.task = d;
+    state.detail.runId = d.run ? d.run.id : null;
+    state.detail.logOffset = 0;
+    renderDetail();
+    startLogPoll();
+  } catch (e) { toast(e.message, true); }
+}
+
+async function goRunPage(p) {
+  const t = state.detail.task;
+  if (!t || !state.detail.runId) return;
+  try {
+    const d = await loadRunPage(t.id, state.detail.runId, p);
+    state.detail.task = d;
+    renderDetail();
+  } catch (e) { toast(e.message, true); }
+}
+
 function renderDetail() {
   const t = state.detail.task;
   const runs = t.runs || [];
-  const run = runs.find((r) => r.id === state.detail.runId) || runs[0] || null;
-  state.detail.runId = run ? run.id : null;
+  const run = t.run || null;
+  state.detail.runId = run ? run.id : (runs[0] ? runs[0].id : null);
   const cfg = t.config || {};
 
   const meta = `
@@ -544,7 +579,10 @@ function renderDetail() {
       <span>任务ID: <b>${t.id}</b></span>
       <span>创建: <b>${fmtTime(t.created_at)}</b></span>
       <span>输出: <b>${esc(cfg.out_dir || "out/" + t.id)}</b></span>
-      ${t.mode === "auto" ? `<span>起始: <b>${esc((t.auto && t.auto.seed_url) || "")}</b></span>` : ""}
+      ${t.mode === "auto" ? `
+        <span>起始: <b>${esc((t.auto && t.auto.seed_url) || "")}</b></span>
+        <span>待爬缓存: <b>${t.auto_pending_count ?? "-"}</b> 条（停止后可继续爬取）</span>
+        <span>已爬: <b>${t.auto_visited_count ?? "-"}</b> 页</span>` : ""}
     </div>`;
 
   let sched = "";
@@ -565,7 +603,8 @@ function renderDetail() {
         <button class="btn" onclick="actTask('${t.id}','pause')">⏸ 暂停</button>
         <button class="btn danger" onclick="actTask('${t.id}','stop')">⏹ 终止</button>`
       : `<button class="btn primary" onclick="actTask('${t.id}','start')">▶ 立即执行</button>`}
-      ${t.mode === "auto" ? `<button class="btn" onclick="probeTask('${t.id}')">🧪 试爬取</button>` : ""}
+      ${t.mode === "auto" ? `<button class="btn" onclick="probeTask('${t.id}')">🧪 试爬取</button>
+        <button class="btn" ${t.running ? "disabled" : ""} onclick="clearCache('${t.id}')">🧹 清空缓存</button>` : ""}
       <button class="btn" onclick="openEdit('${t.id}')">✏️ 编辑配置</button>
     </div>`;
 
@@ -600,7 +639,7 @@ function renderRunPanel(t, run) {
     const meta = r.meta_file
       ? `<span class="file-link" onclick="showMeta('${t.id}','${esc(r.meta_file)}')">📋 元数据</span>` : "—";
     return `<tr>
-      <td>${i + 1}</td>
+      <td>${(run.results_total ? (run.run_page - 1) * DETAIL_PAGE_SIZE : 0) + i + 1}</td>
       <td class="${statusCls}">${r.status}</td>
       <td class="title-cell" title="${esc(r.title)}">${esc(r.title)}</td>
       <td class="url-cell" title="${esc(r.url)}">${esc(r.url)}</td>
@@ -617,6 +656,18 @@ function renderRunPanel(t, run) {
         onclick="previewFile('${t.id}','${esc(im.file)}','图片预览')">`).join("")}
     </div>` : "";
 
+  const total = run.results_total || results.length;
+  const pages = run.run_pages || 1;
+  const page = run.run_page || 1;
+  const pager = total > 0 ? `
+    <div class="pager">
+      ${total > results.length ? `
+        <button class="btn sm" ${page <= 1 ? "disabled" : ""} onclick="goRunPage(${page - 1})">◀ 上一页</button>
+        <span class="pager-info">第 <b>${page}</b> / ${pages} 页 · 共 ${total} 条</span>
+        <button class="btn sm" ${page >= pages ? "disabled" : ""} onclick="goRunPage(${page + 1})">下一页 ▶</button>`
+      : `<span class="pager-info">共 ${total} 条</span>`}
+    </div>` : "";
+
   return `
   <div class="run-panel">
     <div class="run-stats">
@@ -636,7 +687,8 @@ function renderRunPanel(t, run) {
           <thead><tr><th>#</th><th>状态</th><th>标题</th><th>网址</th><th>爬取时间</th><th>HTML</th><th>TXT</th><th>图片</th><th>元数据</th><th>错误</th></tr></thead>
           <tbody>${rows}</tbody>
         </table>
-      </div>` : '<div class="card-line">暂无结果</div>'}
+      </div>
+      ${pager}` : '<div class="card-line">暂无结果</div>'}
     ${thumbHtml}
     <div class="logs" id="logBox"></div>
   </div>`;
@@ -647,13 +699,6 @@ function scrollThumbs() {
   if (box) box.scrollIntoView({ behavior: "smooth", block: "center" });
 }
 
-async function selectRun(rid) {
-  state.detail.runId = rid;
-  state.detail.logOffset = 0;
-  renderDetail();
-  startLogPoll();
-}
-
 async function startLogPoll() {
   stopLogPoll();
   const t = state.detail.task;
@@ -721,6 +766,15 @@ async function probeTask(tid) {
   } catch (e) { toast(e.message, true); }
 }
 
+async function clearCache(tid) {
+  if (!confirm("清空待爬缓存与已爬记录？\n下次运行将从起始网址重新开始爬取。")) return;
+  try {
+    const r = await api(`/api/tasks/${tid}/clear_cache`, { method: "POST" });
+    toast(r.msg || "已清空");
+    openDetail(tid);
+  } catch (e) { toast(e.message, true); }
+}
+
 function renderProbe(r) {
   const body = $("probeBody");
   if (!r.ok) {

static/index.html

@@ -14,9 +14,9 @@
       <input id="searchInput" placeholder="搜索网址 / 标题 / 任务名..." maxlength="100">
       <button id="btnSearch" class="btn sm primary" title="搜索">🔍</button>
     </div>
-    <span class="status-pill" id="statusPill">运行中任务: 0</span>
+    <span class="status-pill" id="statusPill">运行中任务: -</span>
     <button id="btnTheme" class="btn ghost" title="切换日间/夜间主题">☀️</button>
-    <button id="btnTrash" class="btn ghost" title="回收站">🗑️ 回收站 <span id="trashCount" class="trash-count">0</span></button>
+    <button id="btnTrash" class="btn ghost" title="回收站">🗑️ 回收站 <span id="trashCount" class="trash-count">-</span></button>
     <button id="btnRefresh" class="btn ghost">⟳ 刷新</button>
     <button id="btnNew" class="btn primary">＋ 新建任务</button>
   </div>
@@ -24,13 +24,13 @@
 
 <main>
   <div class="stats-bar" id="statsBar">
-    <div class="stat-card"><div class="stat-num" id="stTasks">0</div><div class="stat-label">📋 任务总数</div></div>
-    <div class="stat-card"><div class="stat-num num-run" id="stRunning">0</div><div class="stat-label">🔄 正在运行</div></div>
-    <div class="stat-card"><div class="stat-num" id="stRuns">0</div><div class="stat-label">📊 累计运行次数</div></div>
-    <div class="stat-card"><div class="stat-num num-ok" id="stOk">0</div><div class="stat-label">✅ 成功页面</div></div>
-    <div class="stat-card"><div class="stat-num num-fail" id="stFail">0</div><div class="stat-label">❌ 失败页面</div></div>
-    <div class="stat-card"><div class="stat-num" id="stImgs">0</div><div class="stat-label">🖼️ 已爬图片</div></div>
-    <div class="stat-card"><div class="stat-num" id="stDisk">0</div><div class="stat-label">💾 磁盘占用</div></div>
+    <div class="stat-card"><div class="stat-num" id="stTasks">-</div><div class="stat-label">📋 任务总数</div></div>
+    <div class="stat-card"><div class="stat-num num-run" id="stRunning">-</div><div class="stat-label">🔄 正在运行</div></div>
+    <div class="stat-card"><div class="stat-num" id="stRuns">-</div><div class="stat-label">📊 累计运行次数</div></div>
+    <div class="stat-card"><div class="stat-num num-ok" id="stOk">-</div><div class="stat-label">✅ 成功页面</div></div>
+    <div class="stat-card"><div class="stat-num num-fail" id="stFail">-</div><div class="stat-label">❌ 失败页面</div></div>
+    <div class="stat-card"><div class="stat-num" id="stImgs">-</div><div class="stat-label">🖼️ 已爬图片</div></div>
+    <div class="stat-card"><div class="stat-num" id="stDisk">-</div><div class="stat-label">💾 磁盘占用</div></div>
   </div>
   <div id="taskList" class="task-grid"></div>
   <div id="emptyState" class="empty hidden">
@@ -116,7 +116,7 @@
         </div>
 
         <div id="autoBox" class="hidden box">
-          <div class="field"><label>起始网址 *（系统将自动发现符合规则的链接并爬取）</label>
+          <div class="field"><label>起始网址 *（自动发现符合规则的链接，持续递归爬取直到无新链接可爬）</label>
             <div class="row2">
               <input name="seed_url" placeholder="https://example.com/news" style="flex:1">
               <button type="button" class="btn sm" id="btnProbe">🧪 试爬取</button>
@@ -133,8 +133,10 @@
             <div class="field check"><label><input name="use_regex" type="checkbox"> 规则按正则匹配</label></div>
           </div>
           <div class="row2">
-            <div class="field"><label>最大页数</label><input name="max_pages" type="number" value="50"></div>
-            <div class="field"><label>最大爬取深度</label><input name="max_depth" type="number" value="2"></div>
+            <div class="field"><label>最大爬取深度（0=无限制，N=只爬 N 层）</label>
+              <input name="max_depth" type="number" min="0" value="0"></div>
+            <div class="field"><label>最大页数（0=无限制，默认 1000 作安全上限）</label>
+              <input name="max_pages" type="number" min="0" value="1000"></div>
           </div>
         </div>
       </div>

static/style.css

@@ -275,3 +275,15 @@ td.title-cell { max-width: 220px; overflow: hidden; text-overflow: ellipsis; }
 @keyframes fadein { from { opacity: 0; transform: translate(-50%, -8px); } }
 ::-webkit-scrollbar { width: 8px; height: 8px; }
 ::-webkit-scrollbar-thumb { background: var(--thumb); border-radius: 4px; }
+
+/* 结果分页控件 */
+.pager {
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  gap: 12px;
+  padding: 10px 0;
+  flex-wrap: wrap;
+}
+.pager-info { color: var(--text-dim, #999); font-size: 13px; }
+.pager .btn:disabled { opacity: 0.4; cursor: not-allowed; }

store.py

@@ -10,6 +10,7 @@ HERE = os.path.dirname(os.path.abspath(__file__))
 DATA_DIR = os.path.join(HERE, "data")
 TASKS_FILE = os.path.join(DATA_DIR, "tasks.json")
 RUNS_FILE = os.path.join(DATA_DIR, "runs.json")   # {task_id: [run, ...]}
+AUTO_STATE_DIR = os.path.join(DATA_DIR, "auto_state")  # auto 任务待爬/已爬队列独立存储
 
 _lock = threading.RLock()
 
@@ -140,6 +141,46 @@ def purge_trash():
         return purged
 
 
+# ---------------- auto 任务状态 (待爬队列/已爬集合) ----------------
+# 独立于 tasks.json 存储: 避免上万条 URL 撑大任务文件导致全量读写变慢
+
+def auto_state_path(task_id):
+    return os.path.join(AUTO_STATE_DIR, f"{task_id}.json")
+
+
+def load_auto_state(task_id, fallback_task=None):
+    """读取 auto 任务的 pending/visited 状态; 无独立文件时从旧任务对象迁移"""
+    with _lock:
+        path = auto_state_path(task_id)
+        if os.path.exists(path):
+            st = _load(path, {}) or {}
+            return {"pending": st.get("pending", []), "visited": st.get("visited", [])}
+        if fallback_task:
+            auto = fallback_task.get("auto") or {}
+            st = {"pending": auto.get("pending", []), "visited": auto.get("visited", [])}
+            if st["pending"] or st["visited"]:
+                _save(path, st)   # 自动迁移旧数据
+                return st
+    return {"pending": [], "visited": []}
+
+
+def save_auto_state(task_id, state):
+    with _lock:
+        _save(auto_state_path(task_id), {
+            "pending": state.get("pending", []),
+            "visited": state.get("visited", []),
+        })
+
+
+def clear_auto_state(task_id):
+    """删除 auto 状态文件"""
+    with _lock:
+        try:
+            os.remove(auto_state_path(task_id))
+        except OSError:
+            pass
+
+
 # ---------------- 运行记录 ----------------
 
 def load_runs_map():