feat: 优化搜索广度、深度、效率及可靠性

- 新增百度千帆 AI 搜索 API 引擎 (scripts/engines/baidu_qianfan.py)
- CLI 新增 baidu 子命令,支持 --site 站点限定和 --recency 时间过滤
- UnifiedSearch: 引擎并发控制 (Semaphore) + 超时熔断 (wait_for)
- UnifiedSearch: 搜索结果按 URL 去重
- CLI 新增 health 健康检查命令
- sources 命令新增 API 引擎类别展示
- 支持 .env 文件读取 BAIDU_API_KEY
This commit is contained in:
vex1023 committed 2026-05-14 17:20:44 +08:00
1 parent 87935b59ce
commit 7eb1467e07
5 files changed
+322 -17

No files matched your search

+65 -5
View File
@@ -8,11 +8,13 @@ Usage:
Commands:
search Search all sources (default)
web Search web search engines (Google, Baidu, DuckDuckGo, etc.)
baidu Search via Baidu Qianfan AI Search API (需配置 BAIDU_API_KEY)
news Search news (财联社, 华尔街见闻, RSS)
wechat Search WeChat public account articles
wechat Search WeChat public account articles (via Sougou)
academic Search academic papers (ArXiv)
social Search social media (Twitter/X)
rss Search RSS news feeds
health Test engine connectivity
sources List all available sources
urls Get search URLs without executing
@@ -20,16 +22,21 @@ Options:
-n, --max-results N Max results per source (default: 10)
-j, --json Output as JSON
-o, --output FILE Save results to file
-s, --source NAME Specific source name
-s, --source NAME Specific source name(s)
-r, --region REGION Region filter (cn/global)
--site SITE Restrict to specific site (baidu command only, e.g. mp.weixin.qq.com)
--recency RECENCY Time filter: day, week, month, semiyear, year (baidu command only)
-v, --verbose Show debug logs
-h, --help Show this help
Examples:
web-search wechat 蒙面财经
web-search baidu 今日新闻 --recency week
web-search baidu "央视新闻 中美元首会晤" --site mp.weixin.qq.com --recency week
web-search news "人工智能" -n 5
web-search web "Python async" -r global -j
web-search search "量子计算" -o results.json
web-search health
web-search sources
web-search urls "机器学习"
"""
@@ -226,6 +233,15 @@ async def _search_social(query: str, max_results: int, **kwargs) -> list[Any]:
await us.close()
async def _search_baidu(query: str, max_results: int, **kwargs) -> list[Any]:
from scripts.engines.baidu_qianfan import BaiduQianFanEngine
engine = BaiduQianFanEngine({"api_key": os.getenv("BAIDU_API_KEY", "")})
try:
return await engine.search(query, max_results=max_results, **kwargs)
finally:
await engine.close()
async def _search_rss(query: str, max_results: int, **kwargs) -> list[Any]:
us = _get_unified()()
try:
@@ -258,6 +274,7 @@ def cmd_sources(args):
("academic_sources", "学术源"),
("wechat_sources", "微信源"),
("social_sources", "社交源"),
("api_engines", "API 引擎"),
("special_engines", "特殊源"),
]:
count = categories.get(cat_key, 0)
@@ -280,6 +297,26 @@ def cmd_urls(args):
print()
async def cmd_health(args):
"""Test all engine connectivity."""
us = _get_unified()()
try:
print("正在检测各引擎连通性...\n")
statuses = await us.search_health()
ok_count = sum(1 for s in statuses if s["status"] == "ok")
for s in statuses:
if s["status"] == "ok":
print(f" ✅ {s['name']}")
elif s["status"] == "timeout":
print(f" ⏱️ {s['name']} 超时")
else:
detail = s.get("detail", "")
print(f" ❌ {s['name']} {detail}")
print(f"\n总计: {len(statuses)} 个引擎, {ok_count} 个可用")
finally:
await us.close()
async def _run_search(args):
"""Route search to the right engine(s)."""
if args.verbose:
@@ -288,6 +325,7 @@ async def _run_search(args):
method_map = {
"search": _search_all,
"web": _search_web,
"baidu": _search_baidu,
"news": _search_news,
"wechat": _search_wechat,
"academic": _search_academic,
@@ -300,6 +338,11 @@ async def _run_search(args):
kwargs = {}
if args.command == "web" and args.region:
kwargs["region"] = args.region
if args.command == "baidu":
if args.site:
kwargs["site"] = args.site
if args.recency:
kwargs["recency"] = args.recency
try:
results = await search_fn(
@@ -318,6 +361,7 @@ async def _run_search(args):
cmd_names = {
"search": "全源搜索",
"web": "网页搜索",
"baidu": "百度千帆 AI 搜索",
"news": "新闻搜索",
"wechat": "微信公众号搜索",
"academic": "学术搜索",
@@ -376,8 +420,8 @@ def build_parser() -> argparse.ArgumentParser:
"command",
nargs="?",
default="search",
choices=["search", "web", "news", "wechat", "academic", "social",
"rss", "sources", "urls"],
choices=["search", "web", "baidu", "news", "wechat", "academic",
"social", "rss", "health", "sources", "urls"],
help="Search type (default: search)",
)
parser.add_argument(
@@ -414,6 +458,18 @@ def build_parser() -> argparse.ArgumentParser:
choices=["cn", "global"],
help="Region filter (web search only)",
)
parser.add_argument(
"--site",
type=str,
help="Restrict to specific site (baidu command only, e.g. mp.weixin.qq.com)",
)
parser.add_argument(
"--recency",
type=str,
choices=["day", "week", "month", "semiyear", "year"],
default=None,
help="Time filter (baidu command only, default: year)",
)
parser.add_argument(
"-v", "--verbose",
action="store_true",
@@ -434,11 +490,15 @@ def main():
if args.verbose:
logging.getLogger().setLevel(logging.DEBUG)
# sources and urls commands don't need a query
# sources, urls, health don't need a query
if args.command == "sources":
cmd_sources(args)
return
if args.command == "health":
asyncio.run(cmd_health(args))
return
if args.command == "urls":
if not args.query:
print("错误: urls 命令需要提供搜索关键词", file=sys.stderr)
+11 -3
View File
@@ -38,6 +38,11 @@ def get_social_sources(path: Optional[str] = None) -> list[dict[str, Any]]:
return _load_config(path).get("social_sources", [])
def get_api_engines(path: Optional[str] = None) -> list[dict[str, Any]]:
"""Get API-based search engines (Baidu Qianfan, etc.)."""
return _load_config(path).get("api_engines", [])
def get_special_engines(path: Optional[str] = None) -> list[dict[str, Any]]:
return _load_config(path).get("special_engines", [])
@@ -45,7 +50,8 @@ def get_special_engines(path: Optional[str] = None) -> list[dict[str, Any]]:
def get_engine(name: str, path: Optional[str] = None):
config = _load_config(path)
for category in ["search_engines", "news_sources", "academic_sources",
"wechat_sources", "social_sources", "special_engines"]:
"wechat_sources", "social_sources", "api_engines",
"special_engines"]:
for engine in config.get(category, []):
if engine["name"].lower() == name.lower():
return engine
@@ -56,7 +62,8 @@ def get_all_sources(path: Optional[str] = None) -> list[dict[str, Any]]:
config = _load_config(path)
all_sources = []
for category in ["search_engines", "news_sources", "academic_sources",
"wechat_sources", "social_sources", "special_engines"]:
"wechat_sources", "social_sources", "api_engines",
"special_engines"]:
all_sources.extend(config.get(category, []))
return all_sources
@@ -65,6 +72,7 @@ def list_source_categories(path: Optional[str] = None) -> dict[str, int]:
config = _load_config(path)
categories = [
"search_engines", "news_sources", "academic_sources",
"wechat_sources", "social_sources", "special_engines",
"wechat_sources", "social_sources", "api_engines",
"special_engines",
]
return {cat: len(config.get(cat, [])) for cat in categories}
+172
View File
@@ -0,0 +1,172 @@
"""
Baidu Qianfan AI Search API engine.
Uses the official Baidu Qianfan (千帆) AI Search API to get
structured search results. Requires BAIDU_API_KEY environment variable
or .env file in the project root.
API docs: https://console.bce.baidu.com/qianfan/ais/console/apiKey
"""
from __future__ import annotations
import json
import logging
import os
from pathlib import Path
from typing import Any, Optional
import httpx
from scripts.core.base import SearchEngine, SearchResult
logger = logging.getLogger(__name__)
API_URL = "https://qianfan.baidubce.com/v2/ai_search/web_search"
ENV_KEY = "BAIDU_API_KEY"
def _load_env_file() -> dict[str, str]:
"""Load .env file from project root (simple parser, no external deps)."""
env_path = Path(__file__).resolve().parent.parent.parent / ".env"
if not env_path.exists():
return {}
env_vars = {}
with open(env_path, "r", encoding="utf-8") as f:
for line in f:
line = line.strip()
if not line or line.startswith("#") or "=" not in line:
continue
key, _, val = line.partition("=")
env_vars[key.strip()] = val.strip().strip("\"'")
return env_vars
def _get_api_key() -> str:
"""Get API key: env var > .env file > config."""
key = os.getenv(ENV_KEY)
if key:
return key
env_from_file = _load_env_file()
return env_from_file.get(ENV_KEY, "")
class BaiduQianFanEngine(SearchEngine):
"""百度千帆 AI 搜索 API 引擎."""
def __init__(self, config: Optional[dict] = None):
super().__init__("Baidu Qianfan", config)
self.api_key = (config or {}).get("api_key", "") or _get_api_key()
if not self.api_key:
logger.warning(
f"BAIDU_API_KEY not set! Set it in .env file or as environment variable."
)
timeout = (config or {}).get("timeout", 30)
self._client = httpx.AsyncClient(
timeout=timeout,
headers={
"Authorization": f"Bearer {self.api_key}",
"X-Appbuilder-From": "openclaw",
"Content-Type": "application/json",
},
follow_redirects=True,
)
async def search(
self, query: str, max_results: int = 10, **kwargs
) -> list[SearchResult]:
"""Search via Baidu Qianfan AI Search API.
Args:
query: Search keywords.
max_results: Max results to return.
**kwargs:
site: Restrict search to a specific site (e.g. "mp.weixin.qq.com").
recency: Time filter: "week", "month", "semiyear", "year" (default: "year").
edition: "standard" (default) or "lite".
safe_search: bool, default False.
"""
if not self.api_key:
return [
SearchResult(
title="错误: BAIDU_API_KEY 未配置",
url="https://console.bce.baidu.com/qianfan/ais/console/apiKey",
snippet="请先设置 BAIDU_API_KEY 环境变量",
source=self.name,
rank=1,
category="web",
extra={"error": "missing_api_key"},
)
]
# Build resource_type_filter based on max_results
top_k = min(max_results, 50)
resource_type_filter = [
{"type": "web", "top_k": top_k},
]
# Build search_filter if site is specified
search_filter = {}
site = kwargs.get("site")
if site:
search_filter["match"] = {"site": [site]}
request_body = {
"messages": [{"content": query, "role": "user"}],
"edition": kwargs.get("edition", "standard"),
"search_source": "baidu_search_v2",
"resource_type_filter": resource_type_filter,
"search_recency_filter": kwargs.get("recency", "year"),
"safe_search": kwargs.get("safe_search", False),
}
if search_filter:
request_body["search_filter"] = search_filter
if kwargs.get("block_websites"):
request_body["block_websites"] = kwargs["block_websites"]
try:
response = await self._client.post(API_URL, json=request_body)
response.raise_for_status()
data = response.json()
if "code" in data:
logger.error(f"Baidu Qianfan API error: {data.get('message', '')}")
return []
references = data.get("references", [])
results = []
for i, ref in enumerate(references[:max_results]):
results.append(
SearchResult(
title=ref.get("title", ""),
url=ref.get("url", ""),
snippet=ref.get("snippet", ""),
source=self.name,
rank=i + 1,
category="web",
extra={
k: ref[k]
for k in ("site", "date", "type", "source")
if k in ref
},
)
)
return results
except httpx.HTTPStatusError as e:
logger.error(f"Baidu Qianfan API HTTP error: {e.response.status_code}")
if e.response.status_code == 401:
logger.error("API Key 无效或未授权,请检查 BAIDU_API_KEY")
elif e.response.status_code == 429:
logger.error("API 配额不足")
return []
except httpx.RequestError as e:
logger.error(f"Baidu Qianfan API request failed: {e}")
return []
except json.JSONDecodeError as e:
logger.error(f"Baidu Qianfan API response parse failed: {e}")
return []
async def close(self):
await self._client.aclose()
+71 -9
View File
@@ -13,8 +13,8 @@ from typing import Any, Optional
from scripts.core.base import SearchResult
from scripts.core.config_loader import (
get_search_engines, get_news_sources, get_academic_sources,
get_wechat_sources, get_special_engines, get_social_sources,
get_all_sources, get_engine,
get_wechat_sources, get_api_engines, get_special_engines,
get_social_sources, get_all_sources, get_engine,
)
from scripts.engines.url_engine import UrlSearchEngine
from scripts.engines.parser_engines import DuckDuckGoEngine
@@ -24,15 +24,22 @@ from scripts.news.rss import RSSNewsEngine
from scripts.academic.arxiv import ArXivEngine
from scripts.wechat.sogou_weixin import SogouWeChatEngine
from scripts.social.twitter import TwitterSearchEngine
from scripts.engines.baidu_qianfan import BaiduQianFanEngine
logger = logging.getLogger(__name__)
# Default timeout per engine (seconds)
_DEFAULT_ENGINE_TIMEOUT = 15
# Max concurrent engine calls
_DEFAULT_MAX_CONCURRENT = 5
SOURCE_CATEGORIES = {
"web": get_search_engines,
"news": get_news_sources,
"academic": get_academic_sources,
"wechat": get_wechat_sources,
"social": get_social_sources,
"api": get_api_engines,
"special": get_special_engines,
}
@@ -44,6 +51,10 @@ class UnifiedSearch:
self.config = config or {}
self._cache: dict[str, list[SearchResult]] = {}
self._engines: dict[str, Any] = {}
self._timeout = self.config.get("engine_timeout", _DEFAULT_ENGINE_TIMEOUT)
self._semaphore = asyncio.Semaphore(
self.config.get("max_concurrent", _DEFAULT_MAX_CONCURRENT)
)
async def __aenter__(self):
return self
@@ -80,6 +91,7 @@ class UnifiedSearch:
if not tasks:
return []
# Run all engines with semaphore-based concurrency control
results_lists = await asyncio.gather(*tasks, return_exceptions=True)
all_results: list[SearchResult] = []
for results in results_lists:
@@ -88,9 +100,64 @@ class UnifiedSearch:
continue
all_results.extend(results)
# Dedup by URL, keep first occurrence
all_results = self._dedup_results(all_results)
self._cache[cache_key] = all_results
return all_results
def _dedup_results(self, results: list[SearchResult]) -> list[SearchResult]:
"""Remove duplicates by URL, keeping first occurrence."""
seen: set[str] = set()
deduped = []
for r in results:
key = r.url.strip().rstrip("/")
if key and key not in seen:
seen.add(key)
deduped.append(r)
return deduped
async def _safe_search(self, engine, query, max_results, **kwargs):
async with self._semaphore:
try:
return await asyncio.wait_for(
engine.search(query, max_results=max_results, **kwargs),
timeout=self._timeout,
)
except asyncio.TimeoutError:
logger.warning(f"Engine {engine.name} timed out after {self._timeout}s")
return []
except Exception as e:
logger.error(f"Engine {engine.name} failed: {e}")
return []
async def search_health(self) -> list[dict[str, Any]]:
"""Check health of all configured engines."""
statuses = []
seen_names = set()
for cat_key, cat_fn in SOURCE_CATEGORIES.items():
for engine_def in cat_fn():
name = engine_def.get("name", "?")
if name in seen_names:
continue
seen_names.add(name)
engine = self._get_engine(engine_def)
if engine is None:
statuses.append({"name": name, "status": "error", "detail": "engine_init_failed"})
continue
try:
async with self._semaphore:
await asyncio.wait_for(
engine.search("health check", max_results=1),
timeout=8,
)
statuses.append({"name": name, "status": "ok"})
except asyncio.TimeoutError:
statuses.append({"name": name, "status": "timeout"})
except Exception as e:
statuses.append({"name": name, "status": "error", "detail": str(e)[:80]})
return statuses
def _get_engine(self, engine_def: dict[str, Any]):
"""Get or create an engine instance by definition."""
name = engine_def.get("name", "")
@@ -101,6 +168,8 @@ class UnifiedSearch:
if name == "DuckDuckGo":
engine = DuckDuckGoEngine(self.config)
elif engine_type == "baidu_qianfan" or "baidu qianfan" in name.lower():
engine = BaiduQianFanEngine(self.config)
elif "cls" in name.lower():
engine = CLSNewsEngine(self.config)
elif "wallstreet" in name.lower() or "华尔街" in name:
@@ -121,13 +190,6 @@ class UnifiedSearch:
self._engines[name] = engine
return engine
async def _safe_search(self, engine, query, max_results, **kwargs):
try:
return await engine.search(query, max_results=max_results, **kwargs)
except Exception as e:
logger.error(f"Engine {engine.name} failed: {e}")
return []
async def search_engines(self, query, region=None, max_results=10):
"""Quick search across web search engines."""
engines = get_search_engines()