HFWeeklyPaper / app.py
HowardZhangdqs's picture
Create app.py
c65f6a0 verified
raw
history blame
13.2 kB
import json
import hashlib
import datetime
import requests
import os
import gradio as gr
from datetime import datetime, timedelta
from dataclasses import dataclass
from datetime import datetime
from typing import List, Optional, Any, Dict
# 修改后的数据类(添加 Optional 和默认值)
@dataclass
class Author:
_id: Optional[str] = None
name: Optional[str] = None
hidden: Optional[bool] = None
@dataclass
class Paper:
id: Optional[str] = None
authors: List[Author] = None
publishedAt: Optional[datetime] = None
title: Optional[str] = None
summary: Optional[str] = None
upvotes: Optional[int] = None
discussionId: Optional[str] = None
@dataclass
class SubmittedBy:
_id: Optional[str] = None
avatarUrl: Optional[str] = None
fullname: Optional[str] = None
name: Optional[str] = None
type: Optional[str] = None
isPro: Optional[bool] = None
isHf: Optional[bool] = None
isMod: Optional[bool] = None
followerCount: Optional[int] = None
@dataclass
class Article:
paper: Optional[Paper] = None
publishedAt: Optional[datetime] = None
title: Optional[str] = None
thumbnail: Optional[str] = None
numComments: Optional[int] = None
submittedBy: Optional[SubmittedBy] = None
isAuthorParticipating: Optional[bool] = None
def safe_get(data: Dict, *keys: str) -> Any:
"""安全获取嵌套字典值"""
for key in keys:
data = data.get(key, {}) if isinstance(data, dict) else None
return data if data != {} else None
def parse_article(data: Dict[str, Any]) -> Article:
"""容错式解析函数"""
def parse_datetime(dt_str: Optional[str]) -> Optional[datetime]:
"""安全解析时间"""
if not dt_str:
return None
try:
if dt_str.endswith('Z'):
dt_str = dt_str[:-1] + '+00:00'
return datetime.fromisoformat(dt_str)
except ValueError:
return None
# 解析作者列表
authors = []
for author_data in safe_get(data, "paper", "authors") or []:
authors.append(Author(
_id=author_data.get("_id"),
name=author_data.get("name"),
hidden=author_data.get("hidden")
))
# 解析论文
paper = Paper(
id=safe_get(data, "paper", "id"),
authors=authors,
publishedAt=parse_datetime(safe_get(data, "paper", "publishedAt")),
title=safe_get(data, "paper", "title"),
summary=safe_get(data, "paper", "summary"),
upvotes=safe_get(data, "paper", "upvotes"),
discussionId=safe_get(data, "paper", "discussionId")
) if safe_get(data, "paper") else None
# 解析提交者
submitted_by_data = safe_get(data, "submittedBy")
submitted_by = SubmittedBy(
_id=submitted_by_data.get("_id") if submitted_by_data else None,
avatarUrl=submitted_by_data.get("avatarUrl") if submitted_by_data else None,
fullname=submitted_by_data.get("fullname") if submitted_by_data else None,
name=submitted_by_data.get("name") if submitted_by_data else None,
type=submitted_by_data.get("type") if submitted_by_data else None,
isPro=submitted_by_data.get("isPro") if submitted_by_data else None,
isHf=submitted_by_data.get("isHf") if submitted_by_data else None,
isMod=submitted_by_data.get("isMod") if submitted_by_data else None,
followerCount=submitted_by_data.get("followerCount") if submitted_by_data else None
) if submitted_by_data else None
# 构建最终对象
return Article(
paper=paper,
publishedAt=parse_datetime(data.get("publishedAt")),
title=data.get("title"),
thumbnail=data.get("thumbnail"),
numComments=data.get("numComments"),
submittedBy=submitted_by,
isAuthorParticipating=data.get("isAuthorParticipating")
)
API_URL = "https://huggingface.co/api/daily_papers"
cache = {}
def make_request(url: str):
# Create a hash of the URL to use as the cache key
url_hash = hashlib.md5(url.encode()).hexdigest()
# Check if the response is already cached
if url_hash in cache:
print(f"Cache hit for URL: {url}")
return cache[url_hash]
http_proxy = os.getenv("HF_HTTP_PROXY")
https_proxy = os.getenv("HF_HTTPS_PROXY")
proxies = {
"http": http_proxy,
"https": https_proxy
} if http_proxy or https_proxy else None
attempts = 0
while attempts < 3:
try:
response = requests.get(url, proxies=proxies)
response.raise_for_status()
data = response.json()
# Cache the response
cache[url_hash] = data
return data
except requests.RequestException as e:
attempts += 1
print(f"Attempt {attempts} failed: {e}")
if attempts == 3:
return []
def fetch_papers():
data = make_request(API_URL)
return [parse_article(item) for item in data]
def fetch_papers_with_date(date: datetime):
formatted_date = date.strftime("%Y-%m-%d")
data = make_request(API_URL + "?date=" + formatted_date)
return [parse_article(item) for item in data]
def fetch_papers_with_daterange(start_date: datetime, end_date: datetime):
# return []
# 每天的数据都是独立的,所以只需要遍历日期范围即可
articles = []
current_date = start_date
while current_date <= end_date:
print(current_date)
articles.extend(fetch_papers_with_date(current_date))
print(f"Total articles: {len(articles)}")
current_date += datetime.timedelta(days=1)
# 根据每个文章的.paper.id去重
unique_articles = {}
for article in articles:
if article.paper.id not in unique_articles:
unique_articles[article.paper.id] = article
return list(unique_articles.values())
def sort_by_date(articles):
return sorted(articles, key=lambda x: x.publishedAt, reverse=True)
def sort_by_upvotes(articles):
return sorted(articles, key=lambda x: x.paper.upvotes, reverse=True)
def sort_by_comments(articles):
return sorted(articles, key=lambda x: x.numComments, reverse=True)
def format_author(author):
"""格式化作者信息"""
if not author:
return ""
hidden_status = "(隐藏)" if author.hidden else ""
if author.name:
return f"<a href='https://scholar.google.com/citations?view_op=search_authors&mauthors={author.name.replace(' ', '+')}'>{author.name}</a>{hidden_status}"
return f"匿名作者{hidden_status}"
def format_paper_info(article):
"""生成论文展示的 HTML 内容"""
if not article.paper:
return "论文信息缺失"
info = []
# 标题部分
info.append(f"<h2>{article.title or '无标题论文'}</h2>")
# 缩略图
if article.thumbnail:
info.append(f"<p><img src='{article.thumbnail}' style='max-width: 30em; width: 100%; margin: auto'/></p>")
# 基本信息
info.append(f"<p><strong>论文 ID</strong>:<a href='https://huggingface.co/papers/{article.paper.id}'>{article.paper.id or '未知'}</a></p>")
info.append(f"<p><strong>发布时间</strong>:{article.paper.publishedAt.strftime('%Y-%m-%d %H:%M') if article.paper.publishedAt else '未知'}</p>")
# 作者信息
authors = "、".join([format_author(a) for a in article.paper.authors]) if article.paper.authors else "作者信息暂缺"
info.append(f"<p><strong>作者</strong>:{authors}</p>")
# 摘要
if article.paper.summary:
summary = article.paper.summary.replace('{{', '{').replace('}}', '}').replace('\n', ' ')
info.append(f"<h3>摘要</h3><p>{summary}</p>")
# 讨论信息
info.append(f"<p><strong>点赞数</strong>:{article.paper.upvotes or 0}<span style='margin-left: .5rem'></span>")
info.append(f"<strong>评论数</strong>:{article.numComments or 0}</p>")
if article.paper.discussionId:
info.append(f"<a href='https://huggingface.co/papers/{article.paper.id}/discussion/{article.paper.discussionId}'>进入讨论</a></p>")
# 提交者信息
if article.submittedBy:
submitter = article.submittedBy
info.append(f"<hr><p><strong>提交者</strong>: ")
info.append(
f"<span><img src='{submitter.avatarUrl}' class='author' /></span>{submitter.fullname}(<a href='https://huggingface.co/{submitter.name}'>@{submitter.name}</a>) ")
info.append(f"粉丝数:{submitter.followerCount or 0}</p>")
return "".join(info)
def generate_table_html(papers):
"""生成带可点击标题的表格 HTML"""
html = ['<table class="paper-table"><tr><th>标题</th><th>👍点赞</th><th>💬评论</th><th>📅日期</th></tr>']
for article in papers:
title = article.title or "无标题"
upvotes = article.paper.upvotes or 0
comments = article.numComments or 0
date = article.paper.publishedAt.strftime("%Y-%m-%d") if article.paper.publishedAt else "未知"
paper_id = article.paper.id
row = f"""
<tr>
<td><a class="paper-title" href="javascript:void(0)" onclick="showDetail('{paper_id}')">{title}</a></td>
<td>{upvotes}</td>
<td>{comments}</td>
<td>{date}</td>
</tr>
"""
html.append(row)
html.append("</table>")
return "".join(html)
def build_html(papers):
# 将所有的papers转换为一个html字符串,每个paper用一个div包裹,div内部包含paper的信息,div的id为paper的id
html = ""
for article in papers:
article_html = format_paper_info(article)
html += f"<div id='smartflow-paper-{article.paper.id.replace('.', '-')}' style='display: none'>{article_html}</div>"
return html
def query_papers(start_date_str, end_date_str):
"""处理日期查询"""
try:
start_date = datetime.strptime(start_date_str, "%Y-%m-%d")
end_date = datetime.strptime(end_date_str, "%Y-%m-%d")
papers = fetch_papers_with_daterange(start_date, end_date)
papers = sort_by_upvotes(papers)
return generate_table_html(papers), build_html(papers)
except Exception as e:
print(f"查询出错: {e}")
return "<p>⚠️ 查询失败,请检查日期格式(YYYY-MM-DD)</p>", "<p>⚠️ 查询失败,请检查日期格式(YYYY-MM-DD)</p>"
def show_detail(paper_id, papers):
"""显示论文详情"""
if not papers:
return "请先进行查询"
return build_html(papers)
# CSS 样式(可放入单独文件)
custom_css = """
.paper-table { width: 100%; border-collapse: collapse; }
.paper-table td { padding: 12px; border-bottom: 1px solid #ddd; }
.paper-table th { font-weight: bold; background: #f9f9f920; }
.paper-table tr:hover { background: #f9f9f920; }
.paper-title { color: #1a73e8; cursor: pointer; text-decoration: none !important; }
.paper-title:hover { text-decoration: underline !important; }
.paper-table td:nth-child(2), .paper-table td:nth-child(3), .paper-table td:nth-child(4) { text-align: center; }
.paper-table th:nth-child(2), .paper-table th:nth-child(3), .paper-table th:nth-child(4) { text-align: center; }
.detail-area { margin-top: 20px; padding: 20px; border: 1px solid #ddd; border-radius: 5px; }
"""
custom_js = """
function showDetail(paperId) {
// 隐藏 smartflow-paper-paperId 的所有兄弟节点
var siblings = document.querySelectorAll(`div[id^='smartflow-paper-']:not(#smartflow-paper-${paperId.replace('.', '-')})`);
siblings.forEach(sibling => sibling.style.display = 'none');
// 显示当前节点
var paper = document.getElementById(`smartflow-paper-${paperId.replace('.', '-')}`);
if (paper) {
paper.style.display = 'block';
}
}
"""
def create_interface():
"""创建新的界面布局"""
with gr.Blocks(title="Hugging Face Daily Paper", css=custom_css, head=f"<script>{custom_js}</script>") as app:
# 主界面
gr.Markdown("# 📚 Hugging Face Daily Paper")
# 查询控制区
with gr.Row():
start_date = gr.Textbox(label="起始日期", placeholder="YYYY-MM-DD", value=datetime.now().strftime("%Y-%m-%d"))
end_date = gr.Textbox(label="结束日期", placeholder="YYYY-MM-DD", value=datetime.now().strftime("%Y-%m-%d"))
query_btn = gr.Button("🔍 查询", variant="primary")
# 结果显示区
with gr.Column(visible=True):
results_html = gr.HTML(label="查询结果")
# 论文详情区
with gr.Column(visible=True, elem_classes="detail-area"):
gr.Markdown("## 论文详情")
detail_html = gr.HTML(elem_id="detail-html")
# 事件处理
query_btn.click(
fn=query_papers,
inputs=[start_date, end_date],
outputs=[results_html, detail_html]
)
return app
if __name__ == "__main__":
gr.close_all()
app = create_interface()
app.launch(
# server_name="localhost",
# server_port=7860,
# share=True
)