hz4th_coder
a0b870ee98
- 搜索服务改用 namespace 隔离浏览器实例（参考 webtest-agent 方案）， 解决定时任务与手动处理并发调用 agent-browser 互相踢掉导致搜索结果为0的问题 - agent-browser 调用加瞬时错误自动重试，使用 /tmp/xdg-rt 目录 - 步骤4 提示词放宽：品牌/系列相关页面也纳入提取，无精确型号时兜底选最相关内容 - 步骤4 大模型返回空时自动兜底，不再直接跳过导致会话卡死 - 会话收尾修复：失败/无数据时会话状态正确标记，并自动移除待处理产品 - 处理入口统一：定时任务/单产品/批量处理均改为 process_monitor 大模型流程（防重） - paramhub_client 增加登录态自动恢复：401 或连接失败时自动重新登录重试 - 全流程实测通过：deepseek-v4-flash-0731 → review_id 19ed2daaabfe
996 lines
45 KiB
Python
996 lines
45 KiB
Python
"""
|
||
处理步骤监控服务 - 记录和监控产品处理流程
|
||
"""
|
||
import os
|
||
import time
|
||
import uuid
|
||
import json
|
||
import threading
|
||
import logging
|
||
from datetime import datetime
|
||
from models.database import db
|
||
from services.search_service import search_service
|
||
from services.paramhub_client import paramhub_client
|
||
from services.llm_client import llm_client
|
||
|
||
logger = logging.getLogger('process_monitor')
|
||
|
||
# 处理步骤定义（大模型版）
|
||
PROCESS_STEPS = [
|
||
{'num': 1, 'name': '搜索内容库', 'description': '从内容库搜索相关文章'},
|
||
{'num': 2, 'name': '搜索互联网', 'description': '从互联网搜索最新数据'},
|
||
{'num': 3, 'name': '抓取网页内容', 'description': '抓取搜索结果网页的详细内容'},
|
||
{'num': 4, 'name': '提取产品数据(大模型)', 'description': '调用大模型筛选并提取产品相关内容'},
|
||
{'num': 5, 'name': '填充字段(大模型)', 'description': '调用大模型生成产品数据并检查格式'},
|
||
{'num': 6, 'name': '提交审核', 'description': '提交产品数据到ParamHub审核系统'},
|
||
]
|
||
|
||
class ProcessMonitor:
|
||
"""处理步骤监控器"""
|
||
|
||
def __init__(self):
|
||
self.active_sessions = {}
|
||
self.step_timers = {}
|
||
|
||
def create_session_id(self):
|
||
"""生成会话ID"""
|
||
return f"proc_{datetime.now().strftime('%Y%m%d_%H%M%S')}_{uuid.uuid4().hex[:8]}"
|
||
|
||
def start_process(self, product_name, category=None, subcategory=None):
|
||
"""启动产品处理流程（同产品防重）"""
|
||
# 防重：检查该产品是否已有活跃会话
|
||
active = db.get_active_sessions()
|
||
for s in active:
|
||
if s['product_name'] == product_name and s['status'] in ('pending', 'running', 'paused'):
|
||
logger.warning(f"产品 {product_name} 已有活跃会话 {s['session_id']}，跳过重复启动")
|
||
return s['session_id'], False
|
||
|
||
session_id = self.create_session_id()
|
||
|
||
# 创建会话记录
|
||
db.create_process_session(session_id, product_name, category, subcategory)
|
||
|
||
# 初始化控制信息
|
||
self.active_sessions[session_id] = {
|
||
'paused': False,
|
||
'stop': False,
|
||
'current_step': 0
|
||
}
|
||
|
||
# 启动后台线程处理
|
||
thread = threading.Thread(
|
||
target=self._run_process,
|
||
args=(session_id, product_name, category, subcategory),
|
||
daemon=True
|
||
)
|
||
thread.start()
|
||
|
||
logger.info(f"启动处理会话: {session_id}, 产品: {product_name}")
|
||
return session_id, True
|
||
|
||
def _run_process(self, session_id, product_name, category, subcategory):
|
||
"""执行处理流程"""
|
||
try:
|
||
db.update_session_status(session_id, 'running')
|
||
|
||
result = {'success': False, 'message': '', 'review_id': None}
|
||
|
||
all_data = {
|
||
'library_results': [],
|
||
'internet_results': [],
|
||
'fetched_contents': [],
|
||
'extracted_data': None,
|
||
'filled_data': None
|
||
}
|
||
|
||
# 步骤1: 搜索内容库
|
||
if not self._check_pause(session_id):
|
||
self._start_step(session_id, product_name, 1, '搜索内容库')
|
||
try:
|
||
articles = db.search_articles(product_name, category)
|
||
all_data['library_results'] = articles
|
||
self._complete_step(session_id, 1, {'count': len(articles)})
|
||
logger.info(f"[{session_id}] 步骤1完成: 找到 {len(articles)} 篇文章")
|
||
except Exception as e:
|
||
self._fail_step(session_id, 1, str(e))
|
||
result['message'] = f'搜索内容库失败: {e}'
|
||
|
||
# 步骤2: 搜索互联网
|
||
if not self._check_pause(session_id) and not result.get('message'):
|
||
self._start_step(session_id, product_name, 2, '搜索互联网')
|
||
try:
|
||
internet_results = search_service.search_internet(product_name, max_results=10)
|
||
all_data['internet_results'] = internet_results
|
||
self._complete_step(session_id, 2, {'count': len(internet_results)})
|
||
logger.info(f"[{session_id}] 步骤2完成: 找到 {len(internet_results)} 条结果")
|
||
except Exception as e:
|
||
self._complete_step(session_id, 2, {'count': 0, 'error': str(e)})
|
||
|
||
# 步骤3: 抓取网页内容
|
||
if not self._check_pause(session_id) and all_data['internet_results']:
|
||
self._start_step(session_id, product_name, 3, '抓取网页内容')
|
||
try:
|
||
fetched = []
|
||
failed_count = 0
|
||
urls_to_fetch = [r['url'] for r in all_data['internet_results']]
|
||
total_urls = len(urls_to_fetch)
|
||
|
||
# 创建后台任务记录，这样 /search 页面能看到进度
|
||
bg_task_id = f"fetch_{session_id}"
|
||
db.create_task(bg_task_id, 'fetch_urls', {
|
||
'total': total_urls,
|
||
'auto_save': True,
|
||
'category': category,
|
||
'source': 'process_monitor',
|
||
'product_name': product_name
|
||
})
|
||
db.update_task_status(bg_task_id, 'running', total=total_urls)
|
||
|
||
for i, url in enumerate(urls_to_fetch):
|
||
if self._check_pause(session_id):
|
||
db.update_task_status(bg_task_id, 'stopped', progress=i)
|
||
break
|
||
|
||
# 获取当前URL对应的标题
|
||
result_item = next((r for r in all_data['internet_results'] if r.get('url') == url), {})
|
||
current_title = result_item.get('title', url[:50])
|
||
|
||
# 更新后台任务进度
|
||
db.update_task_status(
|
||
bg_task_id, 'running',
|
||
progress=i,
|
||
current_item=current_title
|
||
)
|
||
|
||
fetch_result = search_service.fetch_url_content(url)
|
||
if fetch_result.get('success'):
|
||
title = fetch_result.get('title', '')
|
||
content = fetch_result.get('content', '')
|
||
article_id = None
|
||
|
||
# 保存到内容库
|
||
try:
|
||
existing = db.search_articles(url)
|
||
if existing and len(existing) > 0:
|
||
# 已存在，使用现有ID
|
||
article_id = existing[0].get('id')
|
||
logger.info(f"[{session_id}] 内容库已存在: {title[:30]}, ID={article_id}")
|
||
else:
|
||
# 新增，获取返回的ID
|
||
article_id = db.add_article(
|
||
product_names=[],
|
||
category=category or '',
|
||
keywords=[],
|
||
summary=content[:200] if content else '',
|
||
content=content,
|
||
source=url,
|
||
url=url,
|
||
search_title=title
|
||
)
|
||
logger.info(f"[{session_id}] 已保存到内容库: {title[:30]}, ID={article_id}")
|
||
except Exception as save_error:
|
||
logger.warning(f"[{session_id}] 保存内容库失败: {save_error}")
|
||
|
||
fetched.append({
|
||
'id': article_id,
|
||
'url': url,
|
||
'title': title,
|
||
'content': content[:500]
|
||
})
|
||
else:
|
||
# 记录失败URL
|
||
failed_count += 1
|
||
error_msg = fetch_result.get('error', '抓取失败')
|
||
try:
|
||
db.add_failed_url(url, product_name, error_msg, source='process_monitor')
|
||
logger.warning(f"[{session_id}] 抓取失败，已记录: {url}")
|
||
except Exception as e:
|
||
logger.error(f"[{session_id}] 记录失败URL出错: {e}")
|
||
|
||
time.sleep(0.3)
|
||
|
||
# 更新后台任务状态为完成
|
||
db.update_task_status(
|
||
bg_task_id, 'completed',
|
||
progress=total_urls,
|
||
result={
|
||
'total': total_urls,
|
||
'success': len(fetched),
|
||
'failed': failed_count,
|
||
'saved': len(fetched)
|
||
}
|
||
)
|
||
|
||
all_data['fetched_contents'] = fetched
|
||
self._complete_step(session_id, 3, {'count': len(fetched), 'failed': failed_count})
|
||
logger.info(f"[{session_id}] 步骤3完成: 抓取 {len(fetched)} 个网页, 失败 {failed_count} 个")
|
||
except Exception as e:
|
||
# 更新后台任务状态为失败
|
||
if 'bg_task_id' in locals():
|
||
db.update_task_status(bg_task_id, 'failed', error_message=str(e))
|
||
self._fail_step(session_id, 3, str(e))
|
||
|
||
# 步骤4: 提取产品数据（调用大模型筛选相关内容）
|
||
if not self._check_pause(session_id):
|
||
self._start_step(session_id, product_name, 4, '提取产品数据(大模型)')
|
||
try:
|
||
# 构建任务文本
|
||
task_text = self._build_agent_task(
|
||
product_name, category, subcategory, all_data
|
||
)
|
||
|
||
# 直接调用大模型
|
||
agent_result = self._call_llm(task_text)
|
||
|
||
if agent_result.get('success'):
|
||
parsed = self._parse_agent_response(agent_result.get('output', ''))
|
||
|
||
if parsed and parsed.get('relevant_ids'):
|
||
# 根据ID从内容库获取实际内容
|
||
relevant_contents = []
|
||
for aid in parsed['relevant_ids']:
|
||
article = db.get_article_by_id(aid)
|
||
if article:
|
||
relevant_contents.append({
|
||
'id': aid,
|
||
'title': article.get('search_title', ''),
|
||
'url': article.get('url', ''),
|
||
'content': article.get('content', ''),
|
||
'summary': article.get('summary', ''),
|
||
'analysis': parsed.get('analysis', {}).get(str(aid), '')
|
||
})
|
||
|
||
all_data['extracted_data'] = {
|
||
'name': product_name,
|
||
'relevant_ids': parsed['relevant_ids'],
|
||
'relevant_contents': relevant_contents,
|
||
'confidence': parsed.get('confidence', 'unknown'),
|
||
'raw_output': agent_result.get('output', '')
|
||
}
|
||
|
||
self._complete_step(session_id, 4, {
|
||
'has_data': True,
|
||
'agent': '大模型',
|
||
'model': self._get_active_model_name(),
|
||
'task_text': task_text,
|
||
'relevant_ids': parsed['relevant_ids'],
|
||
'relevant_count': len(relevant_contents),
|
||
'confidence': parsed.get('confidence', 'unknown'),
|
||
'agent_output': agent_result.get('output', '')[:2000]
|
||
})
|
||
logger.info(f"[{session_id}] 步骤4完成: 大模型返回 {len(parsed['relevant_ids'])} 个相关ID")
|
||
else:
|
||
# 大模型未返回相关ID：兜底处理——选取搜索内容中前1条作为候选
|
||
all_data['extracted_data'] = None
|
||
fallback_ids = []
|
||
if all_data.get('fetched_contents'):
|
||
fallback_ids = [all_data['fetched_contents'][0]['id']]
|
||
elif all_data.get('library_results'):
|
||
fallback_ids = [all_data['library_results'][0]['id']]
|
||
|
||
fallback_contents = []
|
||
for aid in fallback_ids:
|
||
article = db.get_article_by_id(aid)
|
||
if article:
|
||
fallback_contents.append({
|
||
'id': aid,
|
||
'title': article.get('search_title', ''),
|
||
'url': article.get('url', ''),
|
||
'content': article.get('content', ''),
|
||
'summary': article.get('summary', ''),
|
||
'analysis': '兜底候选：大模型未筛选出精确相关数据，选用该条作为参考'
|
||
})
|
||
|
||
if fallback_contents:
|
||
all_data['extracted_data'] = {
|
||
'name': product_name,
|
||
'relevant_ids': fallback_ids,
|
||
'relevant_contents': fallback_contents,
|
||
'confidence': 'low',
|
||
'raw_output': agent_result.get('output', ''),
|
||
'fallback': True
|
||
}
|
||
self._complete_step(session_id, 4, {
|
||
'has_data': True,
|
||
'agent': '大模型',
|
||
'model': self._get_active_model_name(),
|
||
'task_text': task_text,
|
||
'relevant_ids': fallback_ids,
|
||
'relevant_count': len(fallback_contents),
|
||
'confidence': 'low',
|
||
'fallback': True,
|
||
'agent_output': agent_result.get('output', '')[:2000]
|
||
})
|
||
logger.info(f"[{session_id}] 步骤4兜底: 大模型未筛选出相关ID，选用 {fallback_ids} 作为候选")
|
||
else:
|
||
self._complete_step(session_id, 4, {
|
||
'has_data': False,
|
||
'agent': '大模型',
|
||
'model': self._get_active_model_name(),
|
||
'task_text': task_text,
|
||
'agent_output': agent_result.get('output', '')[:2000]
|
||
}, status='skipped')
|
||
result['message'] = '大模型未找到相关数据且无兜底内容'
|
||
else:
|
||
self._fail_step(session_id, 4, f"大模型调用失败: {agent_result.get('error', '未知错误')}")
|
||
result['message'] = f'大模型调用失败: {agent_result.get("error")}'
|
||
except Exception as e:
|
||
self._fail_step(session_id, 4, str(e))
|
||
|
||
# 步骤5: 填充字段（调用大模型生成数据并检查格式）
|
||
if not self._check_pause(session_id) and all_data['extracted_data']:
|
||
self._start_step(session_id, product_name, 5, '填充字段(大模型)')
|
||
try:
|
||
# 构建任务文本
|
||
fill_task_text = self._build_fill_fields_task(
|
||
product_name, category, subcategory, all_data['extracted_data']
|
||
)
|
||
|
||
# 直接调用大模型
|
||
fill_agent_result = self._call_llm(fill_task_text)
|
||
|
||
if fill_agent_result.get('success'):
|
||
fill_parsed = self._parse_fill_agent_response(fill_agent_result.get('output', ''))
|
||
|
||
if fill_parsed and fill_parsed.get('success'):
|
||
product_data = fill_parsed.get('product_data', {})
|
||
format_check = fill_parsed.get('format_check', {})
|
||
|
||
# 本地格式验证
|
||
validation_result = self._validate_product_data(product_data, category)
|
||
|
||
if validation_result.get('valid'):
|
||
all_data['filled_data'] = product_data
|
||
|
||
self._complete_step(session_id, 5, {
|
||
'filled': True,
|
||
'agent': '大模型',
|
||
'model': self._get_active_model_name(),
|
||
'task_text': fill_task_text,
|
||
'product_data': product_data,
|
||
'format_check': format_check,
|
||
'validation': validation_result,
|
||
'agent_output': fill_agent_result.get('output', '')[:2000]
|
||
})
|
||
logger.info(f"[{session_id}] 步骤5完成: 数据生成成功，格式验证通过")
|
||
else:
|
||
# 格式验证失败，记录问题
|
||
self._fail_step(session_id, 5, f"数据格式验证失败: {validation_result.get('errors', [])}")
|
||
result['message'] = '数据格式验证失败'
|
||
else:
|
||
error_msg = fill_parsed.get('message', '未知错误') if fill_parsed else '解析失败'
|
||
self._fail_step(session_id, 5, f"大模型执行失败: {error_msg}")
|
||
result['message'] = f'大模型执行失败: {error_msg}'
|
||
else:
|
||
self._fail_step(session_id, 5, f"大模型调用失败: {fill_agent_result.get('error', '未知错误')}")
|
||
result['message'] = f'大模型调用失败: {fill_agent_result.get("error")}'
|
||
except Exception as e:
|
||
self._fail_step(session_id, 5, str(e))
|
||
|
||
# 步骤6: 提交审核（直接调用ParamHub API，不再依赖智能体）
|
||
if not self._check_pause(session_id) and all_data['filled_data']:
|
||
self._start_step(session_id, product_name, 6, '提交审核')
|
||
try:
|
||
category_type = self._get_category_type(category)
|
||
subcategory_id = subcategory
|
||
success, review_id_or_error = paramhub_client.submit_for_review(
|
||
category_type,
|
||
all_data['filled_data'],
|
||
subcategory_id
|
||
)
|
||
|
||
if success:
|
||
review_id = review_id_or_error
|
||
self._complete_step(session_id, 6, {
|
||
'submitted': True,
|
||
'agent': 'ParamHub API',
|
||
'review_id': review_id,
|
||
'product_data': all_data['filled_data']
|
||
})
|
||
|
||
result['success'] = True
|
||
result['review_id'] = review_id
|
||
|
||
db.update_session_status(session_id, 'completed',
|
||
review_id=review_id,
|
||
result=json.dumps(result, ensure_ascii=False))
|
||
|
||
db.add_process_history(
|
||
product_name=product_name,
|
||
category=category,
|
||
subcategory=subcategory,
|
||
status='submitted',
|
||
review_id=review_id,
|
||
details=all_data
|
||
)
|
||
logger.info(f"[{session_id}] 步骤6完成: 提交成功, review_id={review_id}")
|
||
else:
|
||
self._fail_step(session_id, 6, f"提交失败: {review_id_or_error}")
|
||
result['message'] = f'提交失败: {review_id_or_error}'
|
||
except Exception as e:
|
||
self._fail_step(session_id, 6, str(e))
|
||
|
||
# 兜底收尾：未成功提交的会话标记为 failed
|
||
if not result.get('success'):
|
||
if not result.get('message'):
|
||
result['message'] = '处理未完成（无有效数据或中途停止）'
|
||
# 会话已在步骤失败时标记 failed；这里确保未走到提交分支的会话也收尾
|
||
cur_status = db.get_session_status(session_id)
|
||
if cur_status and cur_status.get('status') in ('running', 'pending'):
|
||
db.update_session_status(session_id, 'failed')
|
||
# 记录处理历史（异常/失败情况）
|
||
db.add_process_history(
|
||
product_name=product_name,
|
||
category=category,
|
||
subcategory=subcategory,
|
||
status='failed',
|
||
details={'message': result['message'], 'all_data': {
|
||
'library_count': len(all_data.get('library_results', [])),
|
||
'internet_count': len(all_data.get('internet_results', [])),
|
||
'fetched_count': len(all_data.get('fetched_contents', [])),
|
||
'extracted': bool(all_data.get('extracted_data')),
|
||
'filled': bool(all_data.get('filled_data'))
|
||
}}
|
||
)
|
||
|
||
# 处理结束（无论成功失败），从待处理列表移除
|
||
try:
|
||
db.remove_pending_product(product_name)
|
||
except Exception as e:
|
||
logger.warning(f"[{session_id}] 移除待处理产品失败: {e}")
|
||
|
||
# 清理
|
||
if session_id in self.active_sessions:
|
||
del self.active_sessions[session_id]
|
||
|
||
return result
|
||
|
||
except Exception as e:
|
||
logger.error(f"处理会话异常: {session_id} - {e}")
|
||
db.update_session_status(session_id, 'failed')
|
||
# 确保清理
|
||
if session_id in self.active_sessions:
|
||
del self.active_sessions[session_id]
|
||
return {'success': False, 'message': str(e)}
|
||
|
||
def _start_step(self, session_id, product_name, step_num, step_name):
|
||
"""开始步骤"""
|
||
db.update_session_status(session_id, 'running', current_step=step_num)
|
||
db.add_process_step(session_id, product_name, step_num, step_name)
|
||
if session_id not in self.step_timers:
|
||
self.step_timers[session_id] = {}
|
||
self.step_timers[session_id][step_num] = time.time()
|
||
|
||
def _complete_step(self, session_id, step_num, step_data=None, status='completed'):
|
||
"""完成步骤"""
|
||
duration_ms = None
|
||
if session_id in self.step_timers and step_num in self.step_timers[session_id]:
|
||
duration_ms = int((time.time() - self.step_timers[session_id][step_num]) * 1000)
|
||
db.update_step_status(session_id, step_num, status, step_data=step_data, duration_ms=duration_ms)
|
||
|
||
def _fail_step(self, session_id, step_num, error_message):
|
||
"""步骤失败"""
|
||
duration_ms = None
|
||
if session_id in self.step_timers and step_num in self.step_timers[session_id]:
|
||
duration_ms = int((time.time() - self.step_timers[session_id][step_num]) * 1000)
|
||
db.update_step_status(session_id, step_num, 'failed', error_message=error_message, duration_ms=duration_ms)
|
||
db.update_session_status(session_id, 'failed')
|
||
|
||
def _check_pause(self, session_id):
|
||
"""检查是否暂停"""
|
||
if session_id not in self.active_sessions:
|
||
return False
|
||
|
||
session = self.active_sessions[session_id]
|
||
|
||
if session.get('stop'):
|
||
return True
|
||
|
||
while session.get('paused'):
|
||
time.sleep(0.5)
|
||
if session.get('stop'):
|
||
return True
|
||
|
||
return False
|
||
|
||
def pause_session(self, session_id):
|
||
"""暂停会话"""
|
||
if session_id in self.active_sessions:
|
||
self.active_sessions[session_id]['paused'] = True
|
||
db.pause_session(session_id, '用户暂停')
|
||
return True
|
||
return False
|
||
|
||
def resume_session(self, session_id):
|
||
"""继续会话"""
|
||
if session_id in self.active_sessions:
|
||
self.active_sessions[session_id]['paused'] = False
|
||
db.resume_session(session_id)
|
||
return True
|
||
return False
|
||
|
||
def stop_session(self, session_id):
|
||
"""停止会话"""
|
||
# 先尝试从内存中停止
|
||
if session_id in self.active_sessions:
|
||
self.active_sessions[session_id]['stop'] = True
|
||
self.active_sessions[session_id]['paused'] = False
|
||
db.update_session_status(session_id, 'stopped')
|
||
logger.info(f"停止会话(内存): {session_id}")
|
||
return True
|
||
|
||
# 如果不在内存中，检查数据库并直接更新状态
|
||
session = db.get_process_session(session_id)
|
||
if session:
|
||
# 只有运行中或暂停状态的会话才能停止
|
||
if session.get('status') in ('running', 'paused', 'pending'):
|
||
db.update_session_status(session_id, 'stopped')
|
||
logger.info(f"停止会话(数据库): {session_id}")
|
||
return True
|
||
else:
|
||
logger.warning(f"会话状态为 {session.get('status')}，无法停止")
|
||
return False
|
||
|
||
logger.warning(f"会话不存在: {session_id}")
|
||
return False
|
||
|
||
def get_session_status(self, session_id):
|
||
"""获取会话状态"""
|
||
session = db.get_process_session(session_id)
|
||
if session:
|
||
steps = db.get_process_steps(session_id)
|
||
return {'session': session, 'steps': steps}
|
||
return None
|
||
|
||
def _build_agent_task(self, product_name, category, subcategory, all_data):
|
||
"""构建智能体任务文本"""
|
||
# 读取模板
|
||
template_file = os.path.join(
|
||
os.path.dirname(os.path.dirname(__file__)),
|
||
'config', 'agent_task_template.txt'
|
||
)
|
||
|
||
if os.path.exists(template_file):
|
||
with open(template_file, 'r', encoding='utf-8') as f:
|
||
template = f.read()
|
||
else:
|
||
# 默认模板
|
||
template = (
|
||
"请分析以下数据ID是否与产品「{{product_name}}」相关且对提取参数有用。\n"
|
||
"类别: {{category}} / {{subcategory}}\n\n"
|
||
"内容库结果ID: {{library_results}}\n\n"
|
||
"互联网已入库ID: {{internet_results}}\n\n"
|
||
"要求：输出相关且有用的ID列表，以JSON格式输出。"
|
||
)
|
||
|
||
# 构建内容库搜索结果ID列表
|
||
library_ids = []
|
||
for article in all_data.get('library_results', []):
|
||
aid = article.get('id')
|
||
if aid:
|
||
title = article.get('search_title', article.get('title', ''))
|
||
library_ids.append(f"ID {aid}: {title}")
|
||
library_text = '\n'.join(library_ids) if library_ids else '（无内容库搜索结果）'
|
||
|
||
# 构建互联网已入库数据ID列表
|
||
internet_ids = []
|
||
for item in all_data.get('fetched_contents', []):
|
||
aid = item.get('id')
|
||
if aid:
|
||
title = item.get('title', '')
|
||
internet_ids.append(f"ID {aid}: {title}")
|
||
internet_text = '\n'.join(internet_ids) if internet_ids else '（无互联网已入库数据）'
|
||
|
||
# 填充模板
|
||
task = template.replace('{{product_name}}', product_name or '未知')
|
||
task = task.replace('{{category}}', category or '未分类')
|
||
task = task.replace('{{subcategory}}', subcategory or '无')
|
||
task = task.replace('{{library_results}}', library_text)
|
||
task = task.replace('{{internet_results}}', internet_text)
|
||
|
||
return task
|
||
|
||
def _get_active_model_name(self):
|
||
"""获取当前激活的模型名称（用于日志/步骤展示）"""
|
||
try:
|
||
cfg = llm_client.get_active_config()
|
||
return cfg.get('model_name', '未知模型')
|
||
except Exception:
|
||
return '未知模型'
|
||
|
||
def _call_llm(self, task_text):
|
||
"""直接调用大模型执行任务（替代原来的 openclaw 智能体）"""
|
||
try:
|
||
logger.info(f"调用大模型执行任务，任务文本 [{len(task_text)} 字符]")
|
||
|
||
# 构建消息
|
||
messages = [
|
||
{
|
||
'role': 'system',
|
||
'content': (
|
||
'你是一个专业的产品数据提取与整理助手。'
|
||
'严格按照用户要求输出结果，遵循任务文本中的输出格式要求。'
|
||
'对于要求JSON输出的任务，必须只输出合法JSON，不要添加多余解释。'
|
||
)
|
||
},
|
||
{'role': 'user', 'content': task_text}
|
||
]
|
||
|
||
# 直接调用大模型
|
||
ok, result = llm_client.chat(
|
||
messages,
|
||
temperature=0.2,
|
||
max_tokens=8192,
|
||
timeout=600
|
||
)
|
||
|
||
if ok:
|
||
logger.info(f"大模型返回: {result[:500]}...")
|
||
return {'success': True, 'output': result}
|
||
else:
|
||
logger.error(f"大模型调用失败: {result}")
|
||
return {'success': False, 'error': result}
|
||
|
||
except Exception as e:
|
||
logger.error(f"大模型调用异常: {e}")
|
||
return {'success': False, 'error': str(e)}
|
||
|
||
def _parse_agent_response(self, output):
|
||
"""解析智能体返回的结果，提取relevant_ids"""
|
||
if not output:
|
||
return None
|
||
|
||
# 尝试从输出中提取JSON
|
||
import re
|
||
|
||
parsed_data = None
|
||
|
||
# 查找JSON块
|
||
json_match = re.search(r'```(?:json)?\s*(\{.*?\})\s*```', output, re.DOTALL)
|
||
if json_match:
|
||
try:
|
||
parsed_data = json.loads(json_match.group(1))
|
||
except json.JSONDecodeError:
|
||
pass
|
||
|
||
# 尝试直接解析整个输出为JSON
|
||
if not parsed_data:
|
||
try:
|
||
parsed_data = json.loads(output)
|
||
except json.JSONDecodeError:
|
||
pass
|
||
|
||
if parsed_data:
|
||
relevant_ids = parsed_data.get('relevant_ids', [])
|
||
# 确保都是整数
|
||
relevant_ids = [int(x) for x in relevant_ids if str(x).isdigit()]
|
||
|
||
return {
|
||
'relevant_ids': relevant_ids,
|
||
'analysis': parsed_data.get('analysis', {}),
|
||
'excluded_ids': parsed_data.get('excluded_ids', []),
|
||
'exclusion_reasons': parsed_data.get('exclusion_reasons', {}),
|
||
'confidence': parsed_data.get('confidence', 'unknown'),
|
||
'raw_output': output
|
||
}
|
||
|
||
# 无法解析为JSON，尝试从文本中提取ID
|
||
id_matches = re.findall(r'(?:ID|id)[\s:]*(\d+)', output)
|
||
if id_matches:
|
||
return {
|
||
'relevant_ids': [int(x) for x in id_matches],
|
||
'analysis': {},
|
||
'confidence': 'low',
|
||
'raw_output': output
|
||
}
|
||
|
||
return None
|
||
|
||
def _build_fill_fields_task(self, product_name, category, subcategory, extracted_data):
|
||
"""构建步骤5填充字段的智能体任务文本"""
|
||
# 读取模板
|
||
template_file = os.path.join(
|
||
os.path.dirname(os.path.dirname(__file__)),
|
||
'config', 'agent_fill_fields_template.txt'
|
||
)
|
||
|
||
if os.path.exists(template_file):
|
||
with open(template_file, 'r', encoding='utf-8') as f:
|
||
template = f.read()
|
||
else:
|
||
# 默认模板
|
||
template = (
|
||
"请根据内容库数据ID {{relevant_content_ids}} 整理产品「{{product_name}}」的参数并提交审核。\n"
|
||
"类别: {{category}} / {{subcategory}}\n"
|
||
"参考API文档: http://192.168.2.8:12007/hz4th_coder/param-hub-python/src/branch/master/API.md"
|
||
)
|
||
|
||
# 构建相关内容ID列表
|
||
relevant_ids = extracted_data.get('relevant_ids', [])
|
||
relevant_contents = extracted_data.get('relevant_contents', [])
|
||
|
||
if relevant_contents:
|
||
content_lines = []
|
||
for item in relevant_contents:
|
||
aid = item.get('id', '')
|
||
title = item.get('title', '')
|
||
content_lines.append(f"ID {aid}: {title}")
|
||
relevant_text = '\n'.join(content_lines)
|
||
elif relevant_ids:
|
||
relevant_text = '\n'.join([f"ID {aid}" for aid in relevant_ids])
|
||
else:
|
||
relevant_text = '（无相关内容ID）'
|
||
|
||
# 填充模板
|
||
task = template.replace('{{product_name}}', product_name or '未知')
|
||
task = task.replace('{{category}}', category or '未分类')
|
||
task = task.replace('{{subcategory}}', subcategory or '无')
|
||
task = task.replace('{{relevant_content_ids}}', relevant_text)
|
||
|
||
return task
|
||
|
||
def _parse_fill_agent_response(self, output):
|
||
"""解析步骤5智能体返回的结果"""
|
||
if not output:
|
||
return None
|
||
|
||
import re
|
||
|
||
parsed_data = None
|
||
|
||
# 查找JSON块
|
||
json_match = re.search(r'```(?:json)?\s*(\{.*?\})\s*```', output, re.DOTALL)
|
||
if json_match:
|
||
try:
|
||
parsed_data = json.loads(json_match.group(1))
|
||
except json.JSONDecodeError:
|
||
pass
|
||
|
||
# 尝试直接解析整个输出为JSON
|
||
if not parsed_data:
|
||
try:
|
||
parsed_data = json.loads(output)
|
||
except json.JSONDecodeError:
|
||
pass
|
||
|
||
if parsed_data:
|
||
return {
|
||
'success': parsed_data.get('success', False),
|
||
'product_data': parsed_data.get('product_data', {}),
|
||
'data_sources': parsed_data.get('data_sources', []),
|
||
'format_check': parsed_data.get('format_check', {}),
|
||
'message': parsed_data.get('message', ''),
|
||
'raw_output': output
|
||
}
|
||
|
||
return None
|
||
|
||
def _validate_product_data(self, product_data, category):
|
||
"""本地验证产品数据格式"""
|
||
errors = []
|
||
warnings = []
|
||
|
||
if not product_data:
|
||
return {'valid': False, 'errors': ['数据为空'], 'warnings': []}
|
||
|
||
# 检查必填字段
|
||
if not product_data.get('name'):
|
||
errors.append('缺少必填字段: name')
|
||
|
||
# 检查字段类型
|
||
category_type = self._get_category_type(category)
|
||
|
||
if category_type == 'model':
|
||
# AI模型字段验证
|
||
if 'parameters' in product_data and product_data['parameters']:
|
||
params = product_data['parameters']
|
||
if not isinstance(params, str) or not params.endswith('B'):
|
||
warnings.append('parameters应为字符串格式如"70B"')
|
||
|
||
if 'context_length' in product_data and product_data['context_length']:
|
||
ctx = product_data['context_length']
|
||
if not isinstance(ctx, int) or ctx <= 0:
|
||
errors.append('context_length应为正整数')
|
||
|
||
if 'mmlu' in product_data and product_data['mmlu']:
|
||
mmlu = product_data['mmlu']
|
||
if not isinstance(mmlu, (int, float)) or mmlu < 0 or mmlu > 100:
|
||
warnings.append('mmlu应为0-100之间的数值')
|
||
|
||
elif category_type == 'gpu':
|
||
# GPU字段验证
|
||
if 'memory_gb' in product_data and product_data['memory_gb']:
|
||
mem = product_data['memory_gb']
|
||
if not isinstance(mem, (int, float)) or mem <= 0:
|
||
errors.append('memory_gb应为正数')
|
||
|
||
if 'cuda_cores' in product_data and product_data['cuda_cores']:
|
||
cores = product_data['cuda_cores']
|
||
if not isinstance(cores, int) or cores <= 0:
|
||
errors.append('cuda_cores应为正整数')
|
||
|
||
if 'price_usd' in product_data and product_data['price_usd']:
|
||
price = product_data['price_usd']
|
||
if not isinstance(price, (int, float)) or price <= 0:
|
||
warnings.append('price_usd应为正数')
|
||
|
||
elif category_type == 'cpu':
|
||
# CPU字段验证
|
||
if 'cores' in product_data and product_data['cores']:
|
||
cores = product_data['cores']
|
||
if not isinstance(cores, int) or cores <= 0:
|
||
errors.append('cores应为正整数')
|
||
|
||
if 'threads' in product_data and product_data['threads']:
|
||
threads = product_data['threads']
|
||
if not isinstance(threads, int) or threads <= 0:
|
||
errors.append('threads应为正整数')
|
||
|
||
if 'base_clock' in product_data and product_data['base_clock']:
|
||
clock = product_data['base_clock']
|
||
if not isinstance(clock, (int, float)) or clock <= 0:
|
||
errors.append('base_clock应为正数')
|
||
|
||
# 检查布尔字段
|
||
for bool_field in ['visible', 'is_pinned']:
|
||
if bool_field in product_data:
|
||
if not isinstance(product_data[bool_field], bool):
|
||
warnings.append(f'{bool_field}应为布尔值')
|
||
|
||
return {
|
||
'valid': len(errors) == 0,
|
||
'errors': errors,
|
||
'warnings': warnings
|
||
}
|
||
|
||
def _build_submit_task(self, product_name, category, subcategory, product_data):
|
||
"""构建步骤6提交审核的智能体任务文本"""
|
||
# 读取模板
|
||
template_file = os.path.join(
|
||
os.path.dirname(os.path.dirname(__file__)),
|
||
'config', 'agent_submit_template.txt'
|
||
)
|
||
|
||
if os.path.exists(template_file):
|
||
with open(template_file, 'r', encoding='utf-8') as f:
|
||
template = f.read()
|
||
else:
|
||
# 默认模板
|
||
template = (
|
||
"请将以下产品数据提交到ParamHub审核系统。\n"
|
||
"产品名称: {{product_name}}\n"
|
||
"类别: {{category}} / {{subcategory}}\n\n"
|
||
"产品数据:\n{{product_data}}\n\n"
|
||
"使用curl命令提交，并记录返回的review_id。"
|
||
)
|
||
|
||
# 填充模板
|
||
task = template.replace('{{product_name}}', product_name or '未知')
|
||
task = task.replace('{{category}}', category or '未分类')
|
||
task = task.replace('{{subcategory}}', subcategory or '无')
|
||
task = task.replace('{{product_data}}', json.dumps(product_data, ensure_ascii=False, indent=2))
|
||
|
||
return task
|
||
|
||
def _parse_submit_agent_response(self, output):
|
||
"""解析步骤6智能体返回的结果"""
|
||
if not output:
|
||
return None
|
||
|
||
import re
|
||
|
||
parsed_data = None
|
||
|
||
# 查找JSON块
|
||
json_match = re.search(r'```(?:json)?\s*(\{.*?\})\s*```', output, re.DOTALL)
|
||
if json_match:
|
||
try:
|
||
parsed_data = json.loads(json_match.group(1))
|
||
except json.JSONDecodeError:
|
||
pass
|
||
|
||
# 尝试直接解析整个输出为JSON
|
||
if not parsed_data:
|
||
try:
|
||
parsed_data = json.loads(output)
|
||
except json.JSONDecodeError:
|
||
pass
|
||
|
||
if parsed_data:
|
||
return {
|
||
'success': parsed_data.get('success', False),
|
||
'review_id': parsed_data.get('review_id'),
|
||
'message': parsed_data.get('message', ''),
|
||
'submitted_data': parsed_data.get('submitted_data', {}),
|
||
'raw_output': output
|
||
}
|
||
|
||
# 尝试从文本中提取review_id
|
||
review_match = re.search(r'review[_-]?id[\s:]*([\w-]+)', output, re.I)
|
||
if review_match:
|
||
return {
|
||
'success': True,
|
||
'review_id': review_match.group(1),
|
||
'message': '从输出中提取到review_id',
|
||
'raw_output': output
|
||
}
|
||
|
||
return None
|
||
|
||
def _extract_data(self, product_name, all_data):
|
||
"""提取产品数据（备用，已被智能体替代）"""
|
||
all_content = []
|
||
|
||
for article in all_data.get('library_results', []):
|
||
content = article.get('content', '')
|
||
if content:
|
||
all_content.append(content)
|
||
|
||
for item in all_data.get('fetched_contents', []):
|
||
content = item.get('content', '')
|
||
if content:
|
||
all_content.append(content)
|
||
|
||
if not all_content:
|
||
return None
|
||
|
||
return {
|
||
'name': product_name,
|
||
'raw_content': '\n---\n'.join(all_content[:3])
|
||
}
|
||
|
||
def _fill_fields(self, extracted_data, category, subcategory):
|
||
"""填充字段"""
|
||
if not extracted_data:
|
||
return None
|
||
|
||
import re
|
||
|
||
filled = {
|
||
'name': extracted_data.get('name', ''),
|
||
'visible': True,
|
||
'is_pinned': False
|
||
}
|
||
|
||
# 从relevant_contents中拼接所有内容
|
||
relevant_contents = extracted_data.get('relevant_contents', [])
|
||
all_content = '\n---\n'.join([
|
||
c.get('content', '') or c.get('summary', '')
|
||
for c in relevant_contents
|
||
if c.get('content') or c.get('summary')
|
||
])
|
||
|
||
# 兼容旧格式
|
||
if not all_content:
|
||
all_content = extracted_data.get('raw_content', '')
|
||
|
||
params_match = re.search(r'(\d+(?:\.\d+)?)\s*[Bb]', all_content)
|
||
if params_match:
|
||
filled['parameters'] = f"{params_match.group(1)}B"
|
||
|
||
date_match = re.search(r'(\d{4}[-/]\d{1,2}[-/]\d{1,2})', all_content)
|
||
if date_match:
|
||
filled['publish_date'] = date_match.group(1).replace('/', '-')
|
||
|
||
filled['_source'] = 'auto_manager'
|
||
filled['_extracted_at'] = datetime.now().isoformat()
|
||
filled['_relevant_ids'] = extracted_data.get('relevant_ids', [])
|
||
|
||
return filled
|
||
|
||
def _get_category_type(self, category):
|
||
"""获取分类类型"""
|
||
if not category:
|
||
return 'dynamic'
|
||
|
||
category_lower = category.lower()
|
||
if 'model' in category_lower or 'ai' in category_lower:
|
||
return 'model'
|
||
elif 'gpu' in category_lower:
|
||
return 'gpu'
|
||
elif 'cpu' in category_lower:
|
||
return 'cpu'
|
||
return 'dynamic'
|
||
|
||
# 全局处理监控实例
|
||
process_monitor = ProcessMonitor() |