extractor

Extractor - 结构化字段提取器

�?HTML 页面中提取结构化字段，输�?JSON 格式。支持四种提取方式：CSS / XPath / Regex / LLM�?

快速开�?

from scripts.extractor import PageExtractor, load_html

# �?URL 加载
html = load_html("https://example.com")

# 或从文件加载
html = load_html("page.html")

# 创建提取�?ext = PageExtractor(html)

# CSS 提取
result = ext.extract_by_css({
    "title": "h1",
    "price": ".price",
    "links": {"selector": "a", "attr": "href", "all": True}
})

# XPath 提取
result = ext.extract_by_xpath({
    "titles": "//h1/text()",
    "emails": "//a[contains(@href,'mailto')]/@href"
})

# 正则提取
result = ext.extract_by_regex({
    "phone": {"pattern": r"1[3-9]\d{9}", "all": False},
    "emails": r"[\w.]+@[\w.]+"
})

# 混合提取
result = ext.extract_all(
    css={"title": "h1"},
    xpath={"specs": "//li/text()"},
    regex={"phone": r"1\d{10}"}
)

CLI 用法

# �?URL 提取
python scripts/extractor.py --url "https://example.com" --css '{"title": "h1"}' --pretty

# 从文件提�?python scripts/extractor.py --file page.html --xpath '{"links": "//a/@href"}' --pretty

# 混合提取
python scripts/extractor.py --file page.html \
    --css '{"title": "h1"}' \
    --regex '{"phone": "1[3-9]\\d{9}"}' \
    --pretty -o result.json

配置格式

CSS 选择�?

简单格式（字符串）�?```json {"title": "h1.product-title"}

完整格式（dict）：
```json
{
    "title": {"selector": "h1", "text": true},
    "link": {"selector": "a.btn", "attr": "href"},
    "items": {"selector": "li.item", "all": true}
}

XPath

简单格式（字符串）�?```json {"titles": "//h1/text()"}

完整格式（dict）：
```json
{
    "first_title": {"xpath": "//h1/text()", "all": false},
    "all_links": {"xpath": "//a/@href", "all": true}
}

正则表达�?

简单格式（字符串）�?```json {"emails": "[\w.]+@[\w.]+"}

完整格式（dict）：
```json
{
    "phone": {"pattern": "1[3-9]\\d{9}", "all": false},
    "price": {"pattern": "¥(\\d+\\.\\d{2})", "all": true},
    "email_ci": {"pattern": "admin@site", "ignorecase": true}
}

输出格式

所有方法返�?{字段�? 值} �?JSON 映射�?- 单个值：字符�?- 多个值：字符串列�?- 无匹配：null

依赖

pip install beautifulsoup4 lxml playwright
playwright install chromium  # �?URL 模式需�?```

## 限制

- LLM 提取尚未实现（标记为 TODO�?- URL 模式需�?Playwright + Chromium
- 大页面建议先裁剪 HTML 再提�?
## 详细文档

参见 [references/usage.md](references/usage.md)