liuguancen/gangyan
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[运维] 接入内网 searxng + 清理启动脚本 + 修 log-trim 权限

Browse Source 
搜索接口：

- duckduckgo_search.py / ZhipuSearchAPI.py 切换到内网 searxng (原 43.251.225.121 / 134.122.191.214 已失效)

启动脚本清理：

- 删除废弃 backend/ 目录 (与 chat_web_backend/ 编译产物 jar MD5 相同，仅是改名副本)

- 删除 start_all.sh 与 langchain-chat/{start,stop,stop_quick,shutdown_all,restart}.sh (被 scripts/*-restart.sh 覆盖)

- 删除 chat_web_backend/{start,test_mysql}.sh

修复：

- scripts/backend-restart.sh 对齐当前实际在跑的 chat_web_backend.jar (profile=dev)

- scripts/log-trim-daemon.sh 把 LOCK 移到 /tmp 按用户命名，修复非首次用户跑时的 Permission denied

新增：

- scripts/start-all.sh：一键启动入口，串联 mysql/redis/milvus/langchain/backend/frontend，含端口自检

- chat_web_backend/application-local.yml.archived：原 backend/ 下 yj profile 覆盖配置的归档备份

其他：

- .gitignore 忽略 scripts/pptist-deploy/PPTist/ (323M 第三方源码树)
This commit is contained in:
liuguancen
2026-04-20 15:59:11 +08:00
parent 279b104434
commit 0c3a393d04
18 changed files with 222 additions and 585 deletions
Download Patch File Download Diff File Expand all files Collapse all files

242
langchain-chat/server/agent/tools/duckduckgo_search.py
View File

@@ -1,186 +1,148 @@
import asyncio
import re
import aiohttp
import json
import logging
import requests
from pydantic import BaseModel, Field
from server.chat import utils
# 配置日志记录器
logging.basicConfig(level=logging.INFO, format='%(asctime)s [%(levelname)s] %(message)s')
logger = logging.getLogger(__name__)
async def duckduckgo_search_iter(query: str, uuid: str = "",time: str = "", resource_type: str = None, limit: int = 3):
# 定义三个API的URL
text_url = 'http://43.251.225.121/inspur/search_text'
video_url = 'http://43.251.225.121/inspur/search_video'
news_url = 'http://43.251.225.121/inspur/search_new'
# 新接口：内网 searxng 服务（原 43.251.225.121 已下线）
# aiohttp 与该 searxng 配合会 30s 超时（疑似 header/UA 被拦），所以改用 requests。
SEARXNG_URL = 'http://118.196.92.255/searxng/search'
SEARXNG_HEADERS = {'User-Agent': 'Mozilla/5.0 (X11; Linux x86_64) gangyan-langchain'}
payload = {
"query": query,
"time": time
}
async def fetch(session, url, json_payload,limit):
logger.info(f"从 {url} 获取数据，请求参数: {json_payload}")
try:
json_payload["limit"] = limit
async with session.post(url, json=json_payload) as response:
if response.status != 200:
logger.error(f"向 {url} 请求失败，状态码 {response.status}")
data = await response.json()
logger.info(f"从 {url} 获取的资料数: {len(data) if isinstance(data, list) else '未知'}")
return data
except Exception as e:
logger.error(f"获取 {url} 数据时发生错误: {e}")
def _searxng_results_to_items(results, mapping, limit):
"""把 searxng 统一的 {url,title,content} 映射成老接口期望的字段格式"""
out = []
for r in results[:limit]:
title = r.get('title', '') or ''
url = r.get('url', '') or ''
content = r.get('content', '') or ''
item = {}
for dst_key, src in mapping.items():
if src == 'title':
item[dst_key] = title
elif src == 'url':
item[dst_key] = url
elif src == 'content':
item[dst_key] = content
out.append(item)
return out
def _sync_fetch(params, limit_n, kind):
logger.info(f"searxng 请求 {kind}: params={params}")
try:
r = requests.get(SEARXNG_URL, params=params, headers=SEARXNG_HEADERS, timeout=15)
if r.status_code != 200:
logger.error(f"searxng {kind} HTTP {r.status_code}")
return []
data = r.json()
results = data.get('results', []) if isinstance(data, dict) else []
logger.info(f"searxng {kind} 条数: {len(results)}")
if kind == 'text':
return _searxng_results_to_items(results, {'title': 'title', 'href': 'url', 'body': 'content'}, limit_n)
if kind == 'video':
return _searxng_results_to_items(results, {'title': 'title', 'content': 'url', 'description': 'content'}, limit_n)
if kind == 'news':
return _searxng_results_to_items(results, {'title': 'title', 'url': 'url', 'body': 'content'}, limit_n)
return []
except Exception as e:
logger.error(f"searxng {kind} 请求异常: {type(e).__name__}: {e}")
return []
# 根据 resource_type 确定要请求的 API
# 默认并发请求三个API
# 视频只请求 video_url
# 新闻只请求 news_url
# 其他类型只请求 text_url
async with aiohttp.ClientSession() as session:
logger.info("发起请求duckduckgo...")
n = limit % 3
limit1 = 0
limit2 = 0
limit3 = 0
match n:
case 0:
limit1 = limit//3
limit2 = limit1
limit3 = limit1
case 1:
limit1 = limit//3 +1
limit2 = limit//3
limit3 = limit2
case 2:
limit1 = limit//3 +1
limit2 = limit1
limit2 = limit
async def duckduckgo_search_iter(query: str, uuid: str = "", time: str = "", resource_type: str = None, limit: int = 3):
logger.info("发起 searxng 搜索请求...")
if resource_type is None or not resource_type == 'video':
text_task = asyncio.create_task(fetch(session, text_url, payload,limit1))
video_task = asyncio.create_task(fetch(session, video_url, payload, limit3))
news_task = asyncio.create_task(fetch(session, news_url, payload, limit2))
text_result, video_result, news_result = await asyncio.gather(text_task, video_task, news_task)
logger.info("合并结果...")
logger.info("合并结果完成")
combined_result = {
"text": text_result,
"video": video_result,
"news": news_result
}
# 三类按 limit 平均分配
n = limit % 3
if n == 0:
limit1 = limit2 = limit3 = limit // 3
elif n == 1:
limit1 = limit // 3 + 1
limit2 = limit3 = limit // 3
else:
limit1 = limit2 = limit // 3 + 1
limit3 = limit // 3
else:
video_result = await fetch(session, video_url, payload, limit)
combined_result = {
"video": video_result
}
del limit1,limit2,limit3
# elif resource_type == 'news':
# news_result = await fetch(session, news_url, payload)
# combined_result = {
# "news": news_result
# }
if resource_type is None or resource_type != 'video':
text_task = asyncio.to_thread(_sync_fetch, {'q': query, 'format': 'json', 'categories': 'general'}, limit1, 'text')
news_task = asyncio.to_thread(_sync_fetch, {'q': query, 'format': 'json', 'categories': 'news'}, limit2, 'news')
video_task = asyncio.to_thread(_sync_fetch, {'q': query, 'format': 'json', 'categories': 'videos'}, limit3, 'video')
text_result, news_result, video_result = await asyncio.gather(text_task, news_task, video_task)
combined_result = {
"text": text_result,
"video": video_result,
"news": news_result,
}
else:
video_result = await asyncio.to_thread(_sync_fetch, {'q': query, 'format': 'json', 'categories': 'videos'}, limit, 'video')
combined_result = {"video": video_result}
# else: # 其他类型
# text_result = await fetch(session, text_url, payload)
# combined_result = {
# "text": text_result
# }
logger.info("searxng 请求已完成")
logger.info("请求已完成")
res = []
source = []
info = utils.get_shared_variable(uuid)
index = info["num"]
if "text" in combined_result:
for item in combined_result["text"]:
index += 1
res.append(f'资料[{index}] 资料标题{item["title"]}({item["href"]}) 资料内容为: {item["body"]}')
source.append(f'资料[{index}] [{item["title"]}]({item["href"]})')
if "video" in combined_result:
for item in combined_result["video"]:
index += 1
res.append(f'资料[{index}] 视频标题[{item["title"]}]({item["content"]}) 视频内容为: {item["description"]}')
source.append(f'视频资料[{index}] [{item["title"]}]({item["content"]})')
if "news" in combined_result:
for item in combined_result["news"]:
index += 1
res.append(f'资料[{index}] 新闻标题[{item["title"]}]({item["url"]}) 新闻内容为: {item["body"]}')
source.append(f'资料[{index}] [{item["title"]}]({item["url"]})')
info["source_docs"].extend(source)
utils.set_shared_variable(uuid, info)
return res,source
res = []
source = []
info = utils.get_shared_variable(uuid)
index = info["num"]
if "text" in combined_result:
for item in combined_result["text"]:
index += 1
res.append(f'资料[{index}] 资料标题{item["title"]}({item["href"]}) 资料内容为: {item["body"]}')
source.append(f'资料[{index}] [{item["title"]}]({item["href"]})')
if "video" in combined_result:
for item in combined_result["video"]:
index += 1
res.append(f'资料[{index}] 视频标题[{item["title"]}]({item["content"]}) 视频内容为: {item["description"]}')
source.append(f'视频资料[{index}] [{item["title"]}]({item["content"]})')
if "news" in combined_result:
for item in combined_result["news"]:
index += 1
res.append(f'资料[{index}] 新闻标题[{item["title"]}]({item["url"]}) 新闻内容为: {item["body"]}')
source.append(f'资料[{index}] [{item["title"]}]({item["url"]})')
info["source_docs"].extend(source)
utils.set_shared_variable(uuid, info)
return res, source
def duckduckgo_search(query: str, time: str = "", resource_type: str = None):
logger.info(f"模型输入: {query}")
# 对传入的 query 字段进行解析
# 判断 query 是否包含 "}{"
# if "}{" in query:
# # 将 query 分割为两个JSON字符串
# split_index = query.find("}{")
# json_part1 = query[:split_index+1]
# json_part2 = query[split_index+1:]
# try:
# obj1 = json.loads(json_part1)
# obj2 = json.loads(json_part2)
# # 提取 query, resource_type, time, uuid
# parsed_query = obj1.get("query", "")
# parsed_resource_type = obj1.get("resource_type", None)
# parsed_time = obj1.get("time", time) # 如obj1未包含time则使用传入的默认值
# parsed_uuid = obj2.get("uuid", "")
logger.info(f"模型输入: {query}")
matches = re.findall(r'\{.*?\}', query)
if len(matches)>=2:
if len(matches) >= 2:
query = matches[0]
else:
return "<关键指令>不需要再调用该工具了</关键指令>"
parsed_uuid = ""
parsed_limit = 3
try:
obj1= json.loads(query)
obj1 = json.loads(query)
parsed_query = obj1.get("query", "")
parsed_limit = obj1.get("limit", 3)
parsed_resource_type = obj1.get("resource_type", None)
parsed_time = obj1.get("time", time) # 如obj1未包含time则使用传入的默认值
parsed_time = obj1.get("time", time)
parsed_uuid = json.loads(matches[1])["uuid"]
# 将解析到的值覆盖原有的参数
query = parsed_query if parsed_query else query
resource_type = parsed_resource_type if parsed_resource_type else resource_type
time = parsed_time if parsed_time else time
logger.info(f"解析完成，query: {query}, uuid: {parsed_uuid}, time: {time}, resource_type: {resource_type}, parsed_limit: {parsed_limit}")
except json.JSONDecodeError as e:
logger.error(f"解析JSON出错: {e}")
# 在同步环境中运行异步函数
combined_result = asyncio.run(duckduckgo_search_iter(query, parsed_uuid, time, resource_type, parsed_limit))
# 以标准json格式输出
logger.info("返回JSON格式的结果给到模型...")
return combined_result
class DuckduckgoInput(BaseModel):
location: str = Field(description="网络搜索查询")
if __name__ == "__main__":
# 测试调用
# 1. 默认请求三个API
# result_default = duckduckgo_search("粉末冶金", "m", "default")
# print("duckduckgo输出(默认):\n", result_default)
# # 2. 只请求视频
# result_video = duckduckgo_search("粉末冶金", "m", "video")
# print("duckduckgo输出(视频):\n", result_video)
# # 3. 只请求新闻
# result_news = duckduckgo_search("粉末冶金", "m", "news")
# print("duckduckgo输出(新闻):\n", result_news)
# 4. 其它类型只请求文本
result_other = duckduckgo_search("粉末冶金", "m", "other")
print("duckduckgo输出(其他):\n", result_other)
result_other = duckduckgo_search('{"query":"粉末冶金","limit":3}{"uuid":"test-uuid"}', "m", "other")
print("searxng输出(其他):\n", result_other)