别让 AI 替你咀嚼信息：搜索引擎摘要化的隐忧与自救之道

docker run -d --name searxng \
  -p 8888:8080 \
  -e SEARXNG_BASE_URL=http://localhost:8888/ \
  searxng/searxng:latest

#!/usr/bin/env python3
"""
search_raw.py —— 拒绝摘要，保留信源
用法: python search_raw.py "python asyncio coroutine deprecated"
输出: 按信源可靠性分级排列的原始搜索结果，附带摘要片段（非AI生成，而是原文excerpt）
"""

import requests
import json
import sys
from collections import defaultdict

# 信源可靠性分级（可根据自己的偏好调整）
TIER_HIGH = ["docs.python.org", "stackoverflow.com", "github.com", "official"]
TIER_MID  = ["medium.com", "dev.to", "realpython.com", "martinfowler.com"]
TIER_LOW  = ["quora.com", "reddit.com"]  # reddit 有好内容但噪音多，放中间偏下

def classify_tier(url: str) -> str:
    for domain in TIER_HIGH:
        if domain in url:
            return "🟢 高可信"
    for domain in TIER_MID:
        if domain in url:
            return "🟡 中等可信"
    for domain in TIER_LOW:
        if domain in url:
            return "🟠 需交叉验证"
    return "⚪ 未分级"

def search_searxng(query: str, instance: str = "http://localhost:8888") -> list:
    params = {
        "q": query,
        "format": "json",
        "categories": "general,it",  # 限定通用+IT类结果
    }
    resp = requests.get(f"{instance}/search", params=params, timeout=10)
    resp.raise_for_status()
    return resp.json().get("results", [])

def present_results(results: list):
    grouped = defaultdict(list)
    for r in results:
        url = r.get("url", "")
        tier = classify_tier(url)
        grouped[tier].append({
            "title": r.get("title", ""),
            "url": url,
            "snippet": r.get("content", ""),  # 这是搜索引擎原始excerpt，非AI生成
        })

    # 按分级顺序输出
    order = ["🟢 高可信", "🟡 中等可信", "🟠 需交叉验证", "⚪ 未分级"]
    for tier in order:
        items = grouped[tier]
        if not items:
            continue
        print(f"\n{'='*60}")
        print(f"  {tier}  ({len(items)} 条结果)")
        print(f"{'='*60}")
        for i, item in enumerate(items, 1):
            print(f"  {i}. {item['title']}")
            print(f"     {item['url']}")
            # 只展示前150字的原文片段，鼓励你点击进去读全文
            snippet = item['snippet'][:150].strip()
            if snippet:
                print(f"     原文片段: {snippet}...")
            print()

if __name__ == "__main__":
    if len(sys.argv) < 2:
        print("用法: python search_raw.py <搜索词>")
        sys.exit(1)

    query = " ".join(sys.argv[1:])
    print(f"🔍 搜索: {query}")
    print(f"   (返回原始信源，不做AI摘要，判断权在你)\n")

    results = search_searxng(query)
    if not results:
        print("未找到结果。")
        sys.exit(0)

    present_results(results)
    print("💡 提示: 以上片段来自搜索引擎原始excerpt，不是AI生成。")
    print("   请点击链接阅读原文，获取完整上下文和论证过程。")

$ python search_raw.py "python 3.12 asyncio coroutine removed"

🔍 搜索: python 3.12 asyncio coroutine removed
   (返回原始信源，不做AI摘要，判断权在你)

============================================================
  🟢 高可信  (2 条结果)
============================================================
  1. What's New In Python 3.12
     https://docs.python.org/3/whatsnew/3.12.html
     原文片段: The asyncio.coroutine() decorator has been removed...
     # 这里你会看到官方文档的原文表述，可以确认精确的移除范围和替代方案

  2. asyncio — Asynchronous I/O — Python 3.12 documentation
     https://docs.python.org/3/library/asyncio.html
     原文片段: Deprecated since version 3.11, removed in version 3.12...

============================================================
  🟡 中等可信  (3 条结果)
============================================================
  ...

💡 提示: 以上片段来自搜索引擎原始excerpt，不是AI生成。
   请点击链接阅读原文，获取完整上下文和论证过程。