sync_data.sh

#!/bin/sh

# 检查必须的环境变量是否已设置
if [ -z "$HF_TOKEN" ] || [ -z "$DATASET_ID" ]; then
    echo "HF_TOKEN 和 DATASET_ID 环境变量未设置"
    exit 1
fi

# 激活 Python 虚拟环境
. /app/venv/bin/activate

# 生成 HuggingFace 同步脚本（Python 实现核心同步逻辑）
cat > hf_sync.py << 'EOL'
from huggingface_hub import HfApi
import sys
import os
import tarfile
import tempfile

# 管理备份文件，超过最大数量则自动删除最旧的备份
def manage_backups(api, repo_id, max_files=50):
    files = api.list_repo_files(repo_id=repo_id, repo_type="dataset")
    backup_files = [f for f in files if f.startswith('backup_') and f.endswith('.tar.gz')]
    backup_files.sort()  # 从旧到新
    if len(backup_files) >= max_files:
        files_to_delete = backup_files[:(len(backup_files) - max_files + 1)]
        for file_to_delete in files_to_delete:
            try:
                api.delete_file(path_in_repo=file_to_delete, repo_id=repo_id, repo_type="dataset")
                print(f'已删除旧备份: {file_to_delete}')
            except Exception as e:
                print(f'删除 {file_to_delete} 时出错: {str(e)}')

# 上传备份文件到 HuggingFace
def upload_backup(file_path, file_name, token, repo_id):
    api = HfApi(token=token)
    try:
        api.upload_file(
            path_or_fileobj=file_path,
            path_in_repo=file_name,
            repo_id=repo_id,
            repo_type="dataset"
        )
        print(f"成功上传备份: {file_name}")
        manage_backups(api, repo_id)
    except Exception as e:
        print(f"上传文件出错: {str(e)}")

# 下载最新的备份并解压到指定目录
def download_latest_backup(token, repo_id, extract_path):
    try:
        api = HfApi(token=token)
        files = api.list_repo_files(repo_id=repo_id, repo_type="dataset")
        backup_files = [f for f in files if f.startswith('backup_') and f.endswith('.tar.gz')]
        if not backup_files:
            print("未找到任何备份文件")
            return
        latest_backup = sorted(backup_files)[-1]
        with tempfile.TemporaryDirectory() as temp_dir:
            filepath = api.hf_hub_download(
                repo_id=repo_id,
                filename=latest_backup,
                repo_type="dataset",
                local_dir=temp_dir
            )
            if filepath and os.path.exists(filepath):
                with tarfile.open(filepath, 'r:gz') as tar:
                    tar.extractall(extract_path)
                print(f"成功恢复备份: {latest_backup}")
    except Exception as e:
        print(f"下载备份出错: {str(e)}")

# 清理所有 LFS（大文件存储）备份文件，只保留最新的 keep_latest 个
def clear_lfs_files(token, repo_id, keep_latest=3):
    try:
        api = HfApi(token=token)
        files = api.list_repo_files(repo_id=repo_id, repo_type="dataset")
        lfs_files = [f for f in files if f.endswith('.tar.gz')]
        lfs_files.sort()  # 从旧到新
        files_to_delete = lfs_files[:-keep_latest] if len(lfs_files) > keep_latest else []
        if not files_to_delete:
            print("没有需要删除的旧 LFS 文件")
            return
        for lfs_file in files_to_delete:
            try:
                api.delete_file(path_in_repo=lfs_file, repo_id=repo_id, repo_type="dataset")
                print(f"已删除旧 LFS 文件: {lfs_file}")
            except Exception as e:
                print(f"删除 LFS 文件出错: {str(e)}")
    except Exception as e:
        print(f"清理 LFS 文件出错: {str(e)}")

# 合并所有历史提交为一个（Super Squash），可选保留最新的几个提交
def super_squash_history(token, repo_id, keep_latest=1):
    try:
        api = HfApi(token=token)
        api.super_squash_history(repo_id=repo_id, repo_type="dataset", keep_last_commits=keep_latest)
        print(f"成功执行 super_squash_history，将历史合并，仅保留最新的 {keep_latest} 个提交。")
    except Exception as e:
        print(f"super_squash_history 执行出错: {str(e)}")

# 主函数，根据命令行参数执行不同操作
if __name__ == "__main__":
    if len(sys.argv) < 2:
        print("用法: python hf_sync.py <操作> [参数]")
        sys.exit(1)
    action = sys.argv[1]
    token = os.getenv('HF_TOKEN')
    repo_id = os.getenv('DATASET_ID')
    if action == "upload":
        if len(sys.argv) < 4:
            print("用法: python hf_sync.py upload <file_path> <file_name>")
            sys.exit(1)
        file_path = sys.argv[2]
        file_name = sys.argv[3]
        upload_backup(file_path, file_name, token, repo_id)
    elif action == "download":
        extract_path = sys.argv[2] if len(sys.argv) > 2 else '.'
        download_latest_backup(token, repo_id, extract_path)
    elif action == "clear_lfs":
        keep_latest = int(sys.argv[2]) if len(sys.argv) > 2 else 3
        clear_lfs_files(token, repo_id, keep_latest)
    elif action == "super_squash":
        keep_latest = int(sys.argv[2]) if len(sys.argv) > 2 else 1
        super_squash_history(token, repo_id, keep_latest)
    else:
        print("无效操作。可用操作: upload, download, clear_lfs, super_squash")
        sys.exit(1)
EOL

# 首次启动时自动下载最新备份
echo "正在从 HuggingFace 下载最新备份..."
python hf_sync.py download "./"

# 定时同步函数
sync_data() {
    while true; do
        echo "开始同步流程，当前时间: $(date)"
        STORAGE_PATH="./storage"
        if [ -d "${STORAGE_PATH}" ]; then
            BACKUP_FILE="backup_$(date +%Y%m%d_%H%M%S).tar.gz"
            tar -czf "${BACKUP_FILE}" -C "${STORAGE_PATH}" .
            echo "已创建备份: ${BACKUP_FILE}"
            python hf_sync.py upload "${BACKUP_FILE}" "${BACKUP_FILE}"
            rm -f "${BACKUP_FILE}"
            python hf_sync.py clear_lfs 3
        else
            echo "数据目录不存在: ${STORAGE_PATH}"
        fi
        SYNC_INTERVAL=${SYNC_INTERVAL:-7200}
        echo "等待 ${SYNC_INTERVAL} 秒后进行下一次同步..."
        sleep $SYNC_INTERVAL
    done
}

# 后台启动同步进程
sync_data &

# 启动主应用（请根据实际路径调整）
exec bash install_reader.sh