AI整理订阅源清单：按主题生成阅读器可导入的OPML

{
  "feeds": [
    {
      "id": "F1",
      "title": "轻量开发日志",
      "xmlUrl": "https://feeds.example.test/dev.xml",
      "htmlUrl": "https://example.test/dev",
      "description": "开发、API与代码工具"
    },
    {
      "id": "F2",
      "title": "原生设计札记",
      "xmlUrl": "https://feeds.example.test/design.xml",
      "htmlUrl": "https://example.test/design",
      "description": "排版、图像与设计"
    },
    {
      "id": "F3",
      "title": "接口速记",
      "xmlUrl": "https://feeds.example.test/api.xml",
      "htmlUrl": "https://example.test/api",
      "description": "接口与开发经验"
    }
  ]
}

python3 -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt
python run.py

{
  "model": "gemini-2.5-flash-lite",
  "messages": [
    {
      "role": "system",
      "content": "严格只输出单个JSON对象，不要Markdown代码围栏，不要解释。只处理给定原创虚构材料，只输出JSON。按简介把feeds分类为开发或设计；返回groups数组，每项title和feed_ids，保留每个feed，不虚构地址。 输出结构：{\"groups\": [{\"title\": \"string\", \"feed_ids\": [\"string\"]}]}"
    },
    {
      "role": "user",
      "content": "任务约束（只输出JSON对象，不用代码围栏）：\n只处理给定原创虚构材料，只输出JSON。按简介把feeds分类为开发或设计；返回groups数组，每项title和feed_ids，保留每个feed，不虚构地址。 输出结构：{\"groups\": [{\"title\": \"string\", \"feed_ids\": [\"string\"]}]}\n待处理合成材料：\n{\"feeds\": [{\"id\": \"F1\", \"title\": \"轻量开发日志\", \"xmlUrl\": \"https://feeds.example.test/dev.xml\", \"htmlUrl\": \"https://example.test/dev\", \"description\": \"开发、API与代码工具\"}, {\"id\": \"F2\", \"title\": \"原生设计札记\", \"xmlUrl\": \"https://feeds.example.test/design.xml\", \"htmlUrl\": \"https://example.test/design\", \"description\": \"排版、图像与设计\"}, {\"id\": \"F3\", \"title\": \"接口速记\", \"xmlUrl\": \"https://feeds.example.test/api.xml\", \"htmlUrl\": \"https://example.test/api\", \"description\": \"接口与开发经验\"}]}"
    }
  ],
  "temperature": 0,
  "max_tokens": 1024,
  "provider": {
    "sort": "success_rate"
  },
  "response_format": {
    "type": "json_object"
  }
}

byid = {r['id']: r for r in data['feeds']}
root = ET.Element('opml', version='2.0')
head = ET.SubElement(root, 'head')
ET.SubElement(head, 'title').text = '原创合成订阅清单'
body = ET.SubElement(root, 'body')
seen = []
for group in output['groups']:
    node = ET.SubElement(body, 'outline', text=group['title'], title=group['title'])
    for feedid in group['feed_ids']:
        r = byid[feedid]
        seen.append(feedid)
        ET.SubElement(node, 'outline', type='rss', text=r['title'], title=r['title'], xmlUrl=r['xmlUrl'], htmlUrl=r['htmlUrl'])
ET.ElementTree(root).write(folder / 'subscriptions.opml', encoding='utf-8', xml_declaration=True)
artifacts.append('subscriptions.opml')
r = ET.parse(folder / 'subscriptions.opml').getroot()
rows = [{'group': group.attrib['text'], 'title': node.attrib['text'], 'xmlUrl': node.attrib['xmlUrl'], 'htmlUrl': node.attrib['htmlUrl']} for group in r.findall('./body/outline') for node in group.findall('outline')]
check('OPML原生层级2.0可回读', r.tag == 'opml' and r.attrib['version'] == '2.0')
check('所有feed恰好一次', sorted(seen) == sorted(byid) and len(seen) == len(set(seen)))
check('xmlUrl与htmlUrl都保留', {x['xmlUrl'] for x in rows} == {x['xmlUrl'] for x in data['feeds']} and {x['htmlUrl'] for x in rows} == {x['htmlUrl'] for x in data['feeds']})
result = {'groups': output['groups'], 'roundtrip': rows, 'network_fetches': 0}

{
  "groups": [
    {
      "title": "开发",
      "feed_ids": [
        "F1",
        "F3"
      ]
    },
    {
      "title": "设计",
      "feed_ids": [
        "F2"
      ]
    }
  ],
  "roundtrip": [
    {
      "group": "开发",
      "title": "轻量开发日志",
      "xmlUrl": "https://feeds.example.test/dev.xml",
      "htmlUrl": "https://example.test/dev"
    },
    {
      "group": "开发",
      "title": "接口速记",
      "xmlUrl": "https://feeds.example.test/api.xml",
      "htmlUrl": "https://example.test/api"
    },
    {
      "group": "设计",
      "title": "原生设计札记",
      "xmlUrl": "https://feeds.example.test/design.xml",
      "htmlUrl": "https://example.test/design"
    }
  ],
  "network_fetches": 0
}