start_system.py

#!/usr/bin/env python3

import asyncio
import argparse
import logging
import os
import sys
from integration_orchestrator import RepairOrchestrator
from repair_loop_engine import RepairLoopEngine, LoopConfig
from auto_repair_executor import AutoRepairExecutor
from rollback_manager import RollbackManager
from data_models import SpaceInfo, SpaceStatus
from datetime import datetime


def setup_logging(level: str = "INFO"):
    logging.basicConfig(
        level=getattr(logging, level.upper()),
        format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
        handlers=[
            logging.StreamHandler(sys.stdout),
            logging.FileHandler('hf_repair_system.log')
        ]
    )


async def start_monitoring(spaces: list, config: LoopConfig):
    logger = logging.getLogger(__name__)
    
    hf_client = None
    repair_executor = AutoRepairExecutor(hf_client, repo_path=".")
    rollback_manager = RollbackManager()
    loop_engine = RepairLoopEngine(repair_executor, config)
    orchestrator = RepairOrchestrator(hf_client)
    orchestrator.set_components(repair_executor, loop_engine, rollback_manager)
    
    for space_id in spaces:
        space_info = SpaceInfo(
            space_id=space_id,
            name=space_id.split('/')[-1],
            repository_url=f"https://huggingface.co/spaces/{space_id}",
            current_status=SpaceStatus.RUNNING,
            last_updated=datetime.now()
        )
        loop_engine.add_space(space_info)
        logger.info(f"Added monitoring for Space: {space_id}")
    
    await orchestrator.start_monitoring()
    logger.info("Repair system monitoring started")
    
    try:
        while True:
            await asyncio.sleep(60)
            stats = orchestrator.get_orchestrator_stats()
            logger.info(f"System stats: {stats}")
    except KeyboardInterrupt:
        logger.info("Received interrupt signal")
    finally:
        await orchestrator.stop_monitoring()
        logger.info("Repair system monitoring stopped")


async def repair_single_space(space_id: str, dry_run: bool = False):
    logger = logging.getLogger(__name__)
    hf_client = None
    orchestrator = RepairOrchestrator(hf_client)
    
    logger.info(f"Starting repair for Space: {space_id}")
    
    if dry_run:
        logger.info("DRY RUN: Would analyze and plan repair, but not execute")
        return
    
    logger.info("Repair functionality requires full system setup")


def validate_spaces(spaces: list) -> bool:
    for space_id in spaces:
        if '/' not in space_id or len(space_id.split('/')) != 2:
            print(f"Invalid Space ID format: {space_id}")
            print("Expected format: username/space-name")
            return False
    return True


def main():
    parser = argparse.ArgumentParser(
        description="HuggingFace Spaces 自动修复系统",
        formatter_class=argparse.RawDescriptionHelpFormatter,
        epilog="示例:\n  # 监控模式\n  python start_system.py --monitor user/space1 user/space2\n  \n  # 修复单个 Space\n  python start_system.py --repair user/space1\n  \n  # 试运行\n  python start_system.py --repair user/space1 --dry-run"
    )
    
    mode_group = parser.add_mutually_exclusive_group(required=True)
    mode_group.add_argument('--monitor', nargs='+', metavar='SPACE', help='监控指定的 Spaces')
    mode_group.add_argument('--repair', metavar='SPACE', help='修复指定的 Space')
    
    parser.add_argument('--max-attempts', type=int, default=5, help='最大修复尝试次数')
    parser.add_argument('--timeout', type=int, default=60, help='超时时间（分钟）')
    parser.add_argument('--check-interval', type=int, default=60, help='检查间隔（秒）')
    parser.add_argument('--max-concurrent', type=int, default=3, help='最大并发修复数')
    parser.add_argument('--log-level', choices=['DEBUG', 'INFO', 'WARNING', 'ERROR'], default='INFO', help='日志级别')
    parser.add_argument('--dry-run', action='store_true', help='试运行模式')
    parser.add_argument('--version', action='version', version='%(prog)s 1.0.0')
    
    args = parser.parse_args()
    
    setup_logging(args.log_level)
    logger = logging.getLogger(__name__)
    
    if not os.getenv('HF_TOKEN') and not args.dry_run:
        logger.error("HF_TOKEN environment variable is required")
        sys.exit(1)
    
    try:
        if args.monitor:
            if not validate_spaces(args.monitor):
                sys.exit(1)
            
            config = LoopConfig(
                max_iterations=args.max_attempts,
                timeout_minutes=args.timeout,
                check_interval_seconds=args.check_interval,
                max_concurrent_repairs=args.max_concurrent
            )
            
            logger.info(f"Starting monitoring for {len(args.monitor)} spaces")
            asyncio.run(start_monitoring(args.monitor, config))
            
        elif args.repair:
            if not validate_spaces([args.repair]):
                sys.exit(1)
            
            asyncio.run(repair_single_space(args.repair, args.dry_run))
    
    except KeyboardInterrupt:
        sys.exit(0)
    except Exception as e:
        logger.error(f"Fatal error: {e}")
        sys.exit(1)


if __name__ == "__main__":
    main()