🔄 卡若AI 同步 2026-02-22 12:42 | 更新：金仓、卡木、运营中枢工作台 | 排除 >20MB: 8 个

01_卡资（金）/金仓_存储备份/服务器管理/scripts/腾讯云_TAT_Node批量修复.py

@@ -6,6 +6,7 @@
 依赖：pip install tencentcloud-sdk-python-common tencentcloud-sdk-python-tat
 """
 import base64
+import json
 import os
 import re
 import sys
@@ -143,8 +144,16 @@ def main():
         resp2 = client.DescribeInvocationTasks(req2)
         for t in (resp2.InvocationTaskSet or []):
             print("  任务:", getattr(t, "InvocationTaskId", t), "状态:", getattr(t, "TaskStatus", "N/A"))
-            if hasattr(t, "Output") and t.Output:
-                print("  输出:", (t.Output or "")[:800])
+            tr = getattr(t, "TaskResult", None)
+            if tr:
+                try:
+                    j = json.loads(tr) if isinstance(tr, str) else {}
+                    out = j.get("Output", "")
+                    if out:
+                        out = base64.b64decode(out).decode("utf-8", errors="replace")
+                        print("  输出:", out[:1200])
+                except Exception:
+                    pass
     except Exception as e:
         print("  查询结果异常:", e)
     print("=" * 50)

01_卡资（金）/金仓_存储备份/服务器管理/scripts/腾讯云_TAT_kr宝塔_全量修复.py

@@ -7,6 +7,7 @@
 凭证：00_账号与API索引.md
 """
 import base64
+import json
 import os
 import re
 import sys
@@ -16,24 +17,16 @@ KR_INSTANCE_ID = "ins-aw0tnqjo"
 REGION = "ap-guangzhou"
 
 SHELL_SCRIPT = r'''#!/bin/bash
-set -e
 echo "=== kr宝塔 全量修复：Nginx(宝塔) + 全部 Node 项目 ==="
 
 # 1. Nginx：确认使用宝塔 Nginx，非系统 Nginx
 echo ""
 echo "【1】Nginx 检查与修复"
-NGX=$(ps aux | grep -E "nginx|nginx:" | grep -v grep | head -1 || true)
-if echo "$NGX" | grep -q "/usr/sbin/nginx"; then
-  echo "  检测到系统 Nginx，切换为宝塔 Nginx..."
-  killall nginx 2>/dev/null || true
-  sleep 2
-fi
-# 若无 nginx 或需确保宝塔 nginx
-if ! pgrep -f "/www/server/nginx" >/dev/null 2>&1; then
-  /www/server/nginx/sbin/nginx -c /www/server/nginx/conf/nginx.conf 2>/dev/null && echo "  宝塔 Nginx 已启动" || echo "  Nginx 可能已在运行"
-fi
-nginx -t 2>/dev/null && nginx -s reload 2>/dev/null && echo "  Nginx 重载完成"
-echo "  当前 Nginx: $(ps aux | grep nginx | grep -v grep | head -1 | awk '{print $11}')"
+NGX=$(ps aux | grep nginx | grep -v grep | head -1 || true)
+echo "$NGX" | grep -q "/usr/sbin/nginx" && { echo "  切换为宝塔 Nginx..."; killall nginx 2>/dev/null || true; sleep 2; }
+pgrep -f "/www/server/nginx" >/dev/null 2>&1 || /www/server/nginx/sbin/nginx -c /www/server/nginx/conf/nginx.conf 2>/dev/null || true
+nginx -t 2>/dev/null && nginx -s reload 2>/dev/null || true
+echo "  Nginx 检查完成"
 
 # 2. 全部 Node 项目批量启动（宝塔 API）
 echo ""
@@ -166,10 +159,23 @@ def main():
         req2.Filters = [f]
         r2 = client.DescribeInvocationTasks(req2)
         for t in (r2.InvocationTaskSet or []):
-            print("  状态:", getattr(t, "TaskStatus", ""))
-            out = getattr(t, "Output", None) or ""
-            if out:
-                print("  输出:\n", out[:4000])
+            st = getattr(t, "TaskStatus", "")
+            print("  状态:", st)
+            tr = getattr(t, "TaskResult", None)
+            if tr:
+                j = json.loads(tr) if isinstance(tr, str) else (vars(tr) if hasattr(tr, "__dict__") else {})
+                out = j.get("Output", "")
+                if out:
+                    try: out = base64.b64decode(out).decode("utf-8", errors="replace")
+                    except: pass
+                    print("  输出:\n", (out or "")[:4000])
+                err = j.get("Error", "")
+                if err:
+                    try: err = base64.b64decode(err).decode("utf-8", errors="replace")
+                    except: pass
+                    print("  错误:", (err or "")[:1000])
+            elif getattr(t, "Output", None):
+                print("  输出:", (t.Output or "")[:4000])
     except Exception as e:
         print("  查询:", e)
     return 0

03_卡木（木）/木叶_视频内容/视频切片/脚本/soul_slice_by_image_themes.py

@@ -0,0 +1,356 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Soul 按图片7主题切片
+========================
+
+严格使用「视频剪辑方案」图片中的 7 个主题，不自行创作。
+时间节点：通过分析转录找出每个完整主题的起止时间（非固定）。
+
+图片7主题（高峰时刻）：
+  1. 引出问题  2. 解决方案  3. 案例分享  4. 未来展望
+  5. 痛点强调  6. 福利展示  7. 权威背书
+
+用法：
+  python3 soul_slice_by_image_themes.py --video "soul 派对 106场.mp4" --transcript transcript.srt
+  python3 soul_slice_by_image_themes.py -v "xxx.mp4" --output-dir ./output --vertical
+"""
+
+import argparse
+import json
+import re
+import subprocess
+import sys
+from pathlib import Path
+
+SCRIPT_DIR = Path(__file__).resolve().parent
+OLLAMA_URL = "http://localhost:11434"
+DEFAULT_CTA = "关注我，每天学一招私域干货"
+
+# 热点切片 7 主题（来自「关于soul派对视频切片内容规划及数据评估报告」）
+# 不可自行创作，严格使用图片中的主题名
+IMAGE_THEMES = [
+    "情绪共鸣",    # 引出问题、营造氛围、启发思考
+    "深度洞察",    # 提出观点、数据支持、解析原理
+    "痛点聚焦",    # 痛点强调、情景代入、加深理解
+    "价值输出",    # 解决方案、干货分享、技能传授
+    "案例解析",    # 实际案例、成功经验、经验复盘
+    "未来趋势",    # 行业展望、发展预测、创新思路
+    "行动召唤",    # 福利展示、即刻行动、引导转化
+]
+
+# 每个主题的检索关键词（用于在转录中定位完整主题）
+THEME_KEYWORDS = {
+    "情绪共鸣": ["问题", "共鸣", "氛围", "思考", "你遇到过", "大家有没有", "感受"],
+    "深度洞察": ["观点", "数据", "原理", "分析", "洞察", "核心", "本质"],
+    "痛点聚焦": ["痛点", "坑", "踩雷", "千万别", "要注意", "难点", "困扰"],
+    "价值输出": ["方案", "方法", "干货", "技能", "怎么做", "核心", "关键"],
+    "案例解析": ["案例", "实际", "真实", "经验", "复盘", "成功", "数据"],
+    "未来趋势": ["未来", "趋势", "展望", "预测", "创新", "发展", "接下来"],
+    "行动召唤": ["福利", "行动", "转化", "免费", "领取", "送", "关注", "分享"],
+}
+
+
+def parse_srt_segments(srt_path: Path) -> list:
+    """解析 SRT 为 [{start_sec, end_sec, text, start_time, end_time}, ...]"""
+    content = srt_path.read_text(encoding="utf-8")
+    segments = []
+    pattern = r"(\d+)\n(\d{2}):(\d{2}):(\d{2}),(\d{3}) --> (\d{2}):(\d{2}):(\d{2}),(\d{3})\n(.*?)(?=\n\n|\Z)"
+    for m in re.findall(pattern, content, re.DOTALL):
+        sh, sm, ss = int(m[1]), int(m[2]), int(m[3])
+        eh, em, es = int(m[5]), int(m[6]), int(m[7])
+        start_sec = sh * 3600 + sm * 60 + ss
+        end_sec = eh * 3600 + em * 60 + es
+        text = m[9].strip().replace("\n", " ")
+        if len(text) > 2:
+            segments.append({
+                "start_sec": start_sec,
+                "end_sec": end_sec,
+                "start_time": f"{sh:02d}:{sm:02d}:{ss:02d}",
+                "end_time": f"{eh:02d}:{em:02d}:{es:02d}",
+                "text": text,
+            })
+    return segments
+
+
+def find_theme_segment(segments: list, theme: str, duration_sec: float) -> dict | None:
+    """
+    在转录中找该主题的完整段落。
+    策略：找包含关键词的连续段落，取前后扩展成 60–120 秒。
+    """
+    keywords = THEME_KEYWORDS.get(theme, [theme])
+    candidates = []
+    for i, s in enumerate(segments):
+        t = s["text"]
+        for kw in keywords:
+            if kw in t:
+                candidates.append((i, s, kw))
+                break
+
+    if not candidates:
+        return None
+
+    # 取第一个匹配，扩展为完整段落（前后各扩展若干句）
+    idx, seg, _ = candidates[0]
+    start_sec = seg["start_sec"]
+    end_sec = seg["end_sec"]
+    # 向前扩展 30 秒或到前一个主题
+    extend_before = 30
+    extend_after = 90
+    start_sec = max(0, start_sec - extend_before)
+    end_sec = min(duration_sec, end_sec + extend_after)
+    # 限制单段 60–150 秒
+    if end_sec - start_sec > 150:
+        end_sec = start_sec + 120
+    if end_sec - start_sec < 45:
+        end_sec = min(duration_sec, start_sec + 90)
+
+    h1, m1, s1 = int(start_sec // 3600), int((start_sec % 3600) // 60), int(start_sec % 60)
+    h2, m2, s2 = int(end_sec // 3600), int((end_sec % 3600) // 60), int(end_sec % 60)
+    excerpt = seg["text"][:60] + "..." if len(seg["text"]) > 60 else seg["text"]
+
+    return {
+        "theme": theme,
+        "title": theme,
+        "start_time": f"{h1:02d}:{m1:02d}:{s1:02d}",
+        "end_time": f"{h2:02d}:{m2:02d}:{s2:02d}",
+        "hook_3sec": theme[:12] if len(theme) > 4 else theme,
+        "cta_ending": DEFAULT_CTA,
+        "transcript_excerpt": excerpt,
+    }
+
+
+def analyze_by_ollama(transcript_path: Path, duration_sec: float) -> list:
+    """用 Ollama 分析转录，输出7主题的时间节点"""
+    try:
+        import requests
+    except ImportError:
+        return []
+
+    segments = parse_srt_segments(transcript_path)
+    with open(transcript_path, "r", encoding="utf-8") as f:
+        raw = f.read()
+    txt = raw[:15000] if len(raw) > 15000 else raw
+
+    themes_str = "、".join(IMAGE_THEMES)
+    prompt = f"""你分析视频文字稿，为以下7个主题各找出【一段完整表述】的起止时间。只输出JSON数组，不要其他文字。
+7个主题（必须全部输出）：{themes_str}
+
+每个元素格式：
+{{"theme":"主题名","title":"主题名","start_time":"HH:MM:SS","end_time":"HH:MM:SS","hook_3sec":"前3秒文案","cta_ending":"关注我，每天学一招私域干货"}}
+
+要求：
+- 每个主题对应视频中实际讲该内容的完整段落
+- 时长 60–120 秒
+- 相邻段落间隔至少20秒
+- 时间从文字稿中提取，必须真实存在
+
+文字稿：
+---
+{txt[:12000]}
+---"""
+
+    try:
+        r = requests.post(
+            f"{OLLAMA_URL}/api/generate",
+            json={"model": "qwen2.5:7b", "prompt": prompt, "stream": False},
+            timeout=120,
+        )
+        if r.status_code != 200:
+            return []
+        out = r.json().get("response", "")
+        m = re.search(r"\[[\s\S]*?\]", out)
+        if m:
+            arr = json.loads(m.group())
+            out_list = []
+            for h in arr:
+                if h.get("theme") in IMAGE_THEMES or h.get("title") in IMAGE_THEMES:
+                    theme = h.get("theme") or h.get("title")
+                    if theme not in [x.get("theme") for x in out_list]:
+                        out_list.append({
+                            "theme": theme,
+                            "title": theme,
+                            "start_time": h.get("start_time", "00:00:00"),
+                            "end_time": h.get("end_time", "00:01:00"),
+                            "hook_3sec": h.get("hook_3sec", f"精彩{theme}"),
+                            "cta_ending": h.get("cta_ending", DEFAULT_CTA),
+                        })
+            if len(out_list) >= 5:
+                return out_list
+    except Exception as e:
+        print(f"  ⚠ Ollama 分析失败: {e}", flush=True)
+    return []
+
+
+def analyze_by_keywords(transcript_path: Path, duration_sec: float) -> list:
+    """按关键词在转录中定位7主题（规则兜底）"""
+    segments = parse_srt_segments(transcript_path)
+    result = []
+    used_ranges = []
+    for theme in IMAGE_THEMES:
+        h = find_theme_segment(segments, theme, duration_sec)
+        if h:
+            start_sec = int(h["start_time"][:2]) * 3600 + int(h["start_time"][3:5]) * 60 + int(h["start_time"][6:8])
+            end_sec = int(h["end_time"][:2]) * 3600 + int(h["end_time"][3:5]) * 60 + int(h["end_time"][6:8])
+            # 避免与前一段重叠
+            overlap = any(s <= start_sec < e or s < end_sec <= e for s, e in used_ranges)
+            if not overlap and end_sec - start_sec >= 40:
+                used_ranges.append((start_sec, end_sec))
+                result.append(h)
+    return result
+
+
+def get_video_duration(path: Path) -> float:
+    cmd = ["ffprobe", "-v", "error", "-show_entries", "format=duration",
+           "-of", "default=noprint_wrappers=1:nokey=1", str(path)]
+    r = subprocess.run(cmd, capture_output=True, text=True)
+    return float(r.stdout.strip()) if r.returncode == 0 else 0
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Soul 按图片7主题切片")
+    parser.add_argument("--video", "-v", required=True, help="输入视频")
+    parser.add_argument("--transcript", "-t", help="transcript.srt（默认：同目录或 output 目录）")
+    parser.add_argument("--output-dir", "-o", help="输出目录")
+    parser.add_argument("--vertical", action="store_true", help="输出竖屏 9:16")
+    parser.add_argument("--skip-clips", action="store_true", help="跳过切片（仅重新增强）")
+    args = parser.parse_args()
+
+    video_path = Path(args.video).resolve()
+    if not video_path.exists():
+        print(f"❌ 视频不存在: {video_path}")
+        sys.exit(1)
+
+    duration = get_video_duration(video_path)
+    print(f"📹 视频: {video_path.name} ({duration/60:.1f} 分钟)")
+
+    base = video_path.parent / (video_path.stem + "_output")
+    if args.output_dir:
+        base = Path(args.output_dir).resolve()
+    base.mkdir(parents=True, exist_ok=True)
+
+    transcript_path = None
+    if args.transcript:
+        transcript_path = Path(args.transcript).resolve()
+    else:
+        for p in [base / "transcript.srt", video_path.parent / (video_path.stem + "_output") / "transcript.srt"]:
+            if p.exists():
+                transcript_path = p
+                break
+    if not transcript_path or not transcript_path.exists():
+        print("❌ 未找到 transcript.srt，请先完成转录")
+        sys.exit(1)
+
+    # 分析：优先 Ollama，失败则用关键词
+    print("🔍 分析转录，提取图片7主题的时间节点...")
+    highlights = analyze_by_ollama(transcript_path, duration)
+    if len(highlights) < 5:
+        print("  使用关键词规则兜底...")
+        highlights = analyze_by_keywords(transcript_path, duration)
+
+    # 确保7个主题都有（缺失的用均匀补位）
+    have_themes = {h["theme"] for h in highlights}
+    for theme in IMAGE_THEMES:
+        if theme not in have_themes:
+            interval = duration / (len(IMAGE_THEMES) + 1)
+            idx = IMAGE_THEMES.index(theme) + 1
+            start_sec = int(interval * idx)
+            end_sec = min(int(start_sec + 90), int(duration) - 5)
+            h = int(start_sec // 3600)
+            m = int((start_sec % 3600) // 60)
+            s = int(start_sec % 60)
+            eh = int(end_sec // 3600)
+            em = int((end_sec % 3600) // 60)
+            es = int(end_sec % 60)
+            highlights.append({
+                "theme": theme,
+                "title": theme,
+                "start_time": f"{h:02d}:{m:02d}:{s:02d}",
+                "end_time": f"{eh:02d}:{em:02d}:{es:02d}",
+                "hook_3sec": f"精彩{theme}",
+                "cta_ending": DEFAULT_CTA,
+            })
+    highlights = sorted(highlights, key=lambda x: IMAGE_THEMES.index(x["theme"]) if x["theme"] in IMAGE_THEMES else 99)
+
+    # 去重，保证顺序
+    seen = set()
+    unique = []
+    for h in highlights:
+        if h["theme"] not in seen and h["theme"] in IMAGE_THEMES:
+            seen.add(h["theme"])
+            unique.append(h)
+    highlights = unique[:7]
+
+    highlights_path = base / "highlights_image7.json"
+    with open(highlights_path, "w", encoding="utf-8") as f:
+        json.dump(highlights, f, ensure_ascii=False, indent=2)
+    print(f"  ✓ 已保存: {highlights_path}")
+    for i, h in enumerate(highlights, 1):
+        print(f"    {i}. {h['title']} {h['start_time']} → {h['end_time']}")
+
+    if args.skip_clips:
+        print("⏭ 跳过切片（--skip-clips）")
+        clips_dir = base / "clips"
+        if clips_dir.exists():
+            enhanced_dir = base / "clips_enhanced_vertical" if args.vertical else base / "clips_enhanced"
+            soul_enhance = SCRIPT_DIR / "soul_enhance.py"
+            subprocess.run([
+                sys.executable, str(soul_enhance),
+                "--clips", str(clips_dir),
+                "--highlights", str(highlights_path),
+                "--transcript", str(transcript_path),
+                "--output", str(enhanced_dir),
+            ], check=True)
+        sys.exit(0)
+
+    clips_dir = base / "clips"
+    clips_dir.mkdir(parents=True, exist_ok=True)
+
+    batch_clip = SCRIPT_DIR / "batch_clip.py"
+    prefix = "soul106" if "106" in video_path.name else "soul"
+    subprocess.run([
+        sys.executable, str(batch_clip),
+        "--input", str(video_path),
+        "--highlights", str(highlights_path),
+        "--output", str(clips_dir),
+        "--prefix", prefix,
+    ], check=True)
+
+    if args.vertical:
+        # 转为竖屏 9:16 (1080x1920)
+        vertical_dir = base / "clips_vertical"
+        vertical_dir.mkdir(parents=True, exist_ok=True)
+        for f in sorted(clips_dir.glob("*.mp4")):
+            out_f = vertical_dir / f.name
+            cmd = [
+                "ffmpeg", "-y", "-i", str(f),
+                "-vf", "scale=1080:1920:force_original_aspect_ratio=increase,crop=1080:1920",
+                "-c:v", "libx264", "-preset", "fast", "-crf", "23",
+                "-c:a", "aac", "-b:a", "128k", str(out_f)
+            ]
+            subprocess.run(cmd, capture_output=True)
+        print(f"  ✓ 竖屏切片: {vertical_dir}")
+        clips_dir = vertical_dir
+
+    enhanced_dir = base / "clips_enhanced"
+    if args.vertical:
+        enhanced_dir = base / "clips_enhanced_vertical"
+    enhanced_dir.mkdir(parents=True, exist_ok=True)
+
+    soul_enhance = SCRIPT_DIR / "soul_enhance.py"
+    subprocess.run([
+        sys.executable, str(soul_enhance),
+        "--clips", str(clips_dir),
+        "--highlights", str(highlights_path),
+        "--transcript", str(transcript_path),
+        "--output", str(enhanced_dir),
+    ], check=True)
+
+    print("\n✅ 完成")
+    print(f"📂 切片: {clips_dir}")
+    print(f"📂 增强: {enhanced_dir}")
+    print(f"📋 方案: {highlights_path}")
+
+
+if __name__ == "__main__":
+    main()

运营中枢/工作台/gitea_push_log.md

@@ -85,3 +85,4 @@
 | 2026-02-22 11:40:59 | 🔄 卡若AI 同步 2026-02-22 11:40 | 更新：水桥平台对接、卡木、运营中枢工作台 | 排除 >20MB: 8 个 |
 | 2026-02-22 11:44:40 | 🔄 卡若AI 同步 2026-02-22 11:44 | 更新：水桥平台对接、卡木、运营中枢工作台 | 排除 >20MB: 8 个 |
 | 2026-02-22 11:47:38 | 🔄 卡若AI 同步 2026-02-22 11:47 | 更新：水桥平台对接、运营中枢工作台 | 排除 >20MB: 8 个 |
+| 2026-02-22 11:58:17 | 🔄 卡若AI 同步 2026-02-22 11:58 | 更新：金仓、水桥平台对接、卡木、运营中枢工作台 | 排除 >20MB: 8 个 |

运营中枢/工作台/代码管理.md

@@ -88,3 +88,4 @@
 | 2026-02-22 11:40:59 | 成功 | 成功 | 🔄 卡若AI 同步 2026-02-22 11:40 | 更新：水桥平台对接、卡木、运营中枢工作台 | 排除 >20MB: 8 个 | [仓库](http://open.quwanzhi.com:3000/fnvtk/karuo-ai) [百科](http://open.quwanzhi.com:3000/fnvtk/karuo-ai/wiki) |
 | 2026-02-22 11:44:40 | 成功 | 成功 | 🔄 卡若AI 同步 2026-02-22 11:44 | 更新：水桥平台对接、卡木、运营中枢工作台 | 排除 >20MB: 8 个 | [仓库](http://open.quwanzhi.com:3000/fnvtk/karuo-ai) [百科](http://open.quwanzhi.com:3000/fnvtk/karuo-ai/wiki) |
 | 2026-02-22 11:47:38 | 成功 | 成功 | 🔄 卡若AI 同步 2026-02-22 11:47 | 更新：水桥平台对接、运营中枢工作台 | 排除 >20MB: 8 个 | [仓库](http://open.quwanzhi.com:3000/fnvtk/karuo-ai) [百科](http://open.quwanzhi.com:3000/fnvtk/karuo-ai/wiki) |
+| 2026-02-22 11:58:17 | 成功 | 成功 | 🔄 卡若AI 同步 2026-02-22 11:58 | 更新：金仓、水桥平台对接、卡木、运营中枢工作台 | 排除 >20MB: 8 个 | [仓库](http://open.quwanzhi.com:3000/fnvtk/karuo-ai) [百科](http://open.quwanzhi.com:3000/fnvtk/karuo-ai/wiki) |