from flask import Flask, render_template, request, jsonify
import requests
import os
import time
import random
from collections import Counter
app = Flask(__name__)
# Generate dummy spaces in case of error
def generate_dummy_spaces(count):
"""
API 호출 실패 시 예시용 더미 스페이스 생성
"""
spaces = []
for i in range(count):
spaces.append({
'id': f'dummy/space-{i}',
'owner': 'dummy',
'title': f'Example Space {i+1}',
'description': 'Dummy space for fallback',
'likes': 100 - i,
'createdAt': '2023-01-01T00:00:00.000Z',
'hardware': 'cpu',
'user': {
'avatar_url': 'https://huggingface.co/front/thumbnails/huggingface/default-avatar.svg',
'name': 'dummyUser'
}
})
return spaces
# Function to fetch Zero-GPU (CPU-based) Spaces from Huggingface with pagination
def fetch_trending_spaces(offset=0, limit=72):
"""
Trending용 CPU 스페이스 목록 가져오기 (정렬은 Hugging Face 기본 정렬)
"""
try:
url = "https://huggingface.co/api/spaces"
params = {
"limit": 10000, # 더 많이 가져오기
"hardware": "cpu" # <-- Zero GPU(=CPU) 필터 적용
}
response = requests.get(url, params=params, timeout=30)
if response.status_code == 200:
spaces = response.json()
# owner나 id가 'None'인 경우 제외
filtered_spaces = [
space for space in spaces
if space.get('owner') != 'None'
and space.get('id', '').split('/', 1)[0] != 'None'
]
# 전체 목록에 대해 "글로벌 랭크"를 매긴다 (1부터 시작)
for i, sp in enumerate(filtered_spaces):
sp['global_rank'] = i + 1
# Slice according to requested offset and limit
start = min(offset, len(filtered_spaces))
end = min(offset + limit, len(filtered_spaces))
print(f"[fetch_trending_spaces] CPU기반 스페이스 총 {len(filtered_spaces)}개, "
f"요청 구간 {start}~{end-1} 반환")
return {
'spaces': filtered_spaces[start:end],
'total': len(filtered_spaces),
'offset': offset,
'limit': limit,
'all_spaces': filtered_spaces # 통계 산출용
}
else:
print(f"Error fetching spaces: {response.status_code}")
# 실패 시 더미 데이터 생성
return {
'spaces': generate_dummy_spaces(limit),
'total': 200,
'offset': offset,
'limit': limit,
'all_spaces': generate_dummy_spaces(500)
}
except Exception as e:
print(f"Exception when fetching spaces: {e}")
# 실패 시 더미 데이터 생성
return {
'spaces': generate_dummy_spaces(limit),
'total': 200,
'offset': offset,
'limit': limit,
'all_spaces': generate_dummy_spaces(500)
}
def fetch_latest_spaces(offset=0, limit=72):
"""
'createdAt' 기준 내림차순으로 최근 스페이스 500개를 추린 뒤,
offset ~ offset+limit 개만 반환
"""
try:
url = "https://huggingface.co/api/spaces"
params = {
"limit": 10000, # 충분히 많이 가져옴
"hardware": "cpu"
}
response = requests.get(url, params=params, timeout=30)
if response.status_code == 200:
spaces = response.json()
# owner나 id가 'None'인 경우 제외
filtered_spaces = [
space for space in spaces
if space.get('owner') != 'None'
and space.get('id', '').split('/', 1)[0] != 'None'
]
# createdAt 내림차순 정렬
# createdAt 예: "2023-01-01T00:00:00.000Z"
# 문자열 비교도 가능하지만, 안정성을 위해 time 파싱 후 비교할 수도 있음
def parse_time(sp):
return sp.get('createdAt') or ''
# 내림차순
filtered_spaces.sort(key=parse_time, reverse=True)
# 상위 500개만 추리기
truncated = filtered_spaces[:500]
# 필요한 구간 슬라이싱
start = min(offset, len(truncated))
end = min(offset + limit, len(truncated))
print(f"[fetch_latest_spaces] CPU기반 스페이스 총 {len(spaces)}개 중 필터 후 {len(filtered_spaces)}개, 상위 500개 중 {start}~{end-1} 반환")
return {
'spaces': truncated[start:end],
'total': len(truncated), # 500 이하
'offset': offset,
'limit': limit
}
else:
print(f"Error fetching spaces: {response.status_code}")
return {
'spaces': generate_dummy_spaces(limit),
'total': 500,
'offset': offset,
'limit': limit
}
except Exception as e:
print(f"Exception when fetching spaces: {e}")
return {
'spaces': generate_dummy_spaces(limit),
'total': 500,
'offset': offset,
'limit': limit
}
# Transform Huggingface URL to direct space URL
def transform_url(owner, name):
"""
Hugging Face Space -> 서브도메인 접근 URL
예) huggingface.co/spaces/owner/spaceName -> owner-spacename.hf.space
"""
# 1. Replace '.' with '-'
name = name.replace('.', '-')
# 2. Replace '_' with '-'
name = name.replace('_', '-')
# 3. Convert to lowercase
owner = owner.lower()
name = name.lower()
return f"https://{owner}-{name}.hf.space"
# Get space details
def get_space_details(space_data, index, offset):
"""
스페이스 상세 필드 추출
- rank는 offset 기반 (현재 페이지)
"""
try:
if '/' in space_data.get('id', ''):
owner, name = space_data.get('id', '').split('/', 1)
else:
owner = space_data.get('owner', '')
name = space_data.get('id', '')
# Ignore if contains None
if owner == 'None' or name == 'None':
return None
# Construct URLs
original_url = f"https://huggingface.co/spaces/{owner}/{name}"
embed_url = transform_url(owner, name)
# Likes count
likes_count = space_data.get('likes', 0)
# Title
title = space_data.get('title') or name
# Description
short_desc = space_data.get('description', '')
# User info
user_info = space_data.get('user', {})
avatar_url = user_info.get('avatar_url', '')
author_name = user_info.get('name') or owner
return {
'url': original_url,
'embedUrl': embed_url,
'title': title,
'owner': owner,
'name': name,
'likes_count': likes_count,
'description': short_desc,
'avatar_url': avatar_url,
'author_name': author_name,
'rank': offset + index + 1 # 현재 페이지 표시용 랭크
}
except Exception as e:
print(f"Error processing space data: {e}")
# 에러 시 기본 데이터로 대체
return {
'url': 'https://huggingface.co/spaces',
'embedUrl': 'https://huggingface.co/spaces',
'title': 'Error Loading Space',
'owner': 'huggingface',
'name': 'error',
'likes_count': 0,
'description': '',
'avatar_url': '',
'author_name': 'huggingface',
'rank': offset + index + 1
}
# Get owner statistics from all spaces (for the "Trending" tab's top owners)
def get_owner_stats(all_spaces):
"""
상위 500위(global_rank <= 500) 이내에 배치된 스페이스들의 owner를 추출해,
각 owner가 몇 번 등장했는지 센 뒤 상위 30명만 반환
"""
# Top 500
top_500 = [s for s in all_spaces if s.get('global_rank', 999999) <= 500]
owners = []
for space in top_500:
if '/' in space.get('id', ''):
owner, _ = space.get('id', '').split('/', 1)
else:
owner = space.get('owner', '')
if owner and owner != 'None':
owners.append(owner)
# Count occurrences of each owner in top 500
owner_counts = Counter(owners)
# Get top 30 owners by count
top_owners = owner_counts.most_common(30)
return top_owners
# Homepage route
@app.route('/')
def home():
"""
index.html 템플릿 렌더링 (메인 페이지)
"""
return render_template('index.html')
# Zero-GPU spaces API (Trending)
@app.route('/api/trending-spaces', methods=['GET'])
def trending_spaces():
"""
hardware=cpu 스페이스 목록을 불러와 검색, 페이징, 통계 등을 적용 (기존 'Trending')
"""
search_query = request.args.get('search', '').lower()
offset = int(request.args.get('offset', 0))
limit = int(request.args.get('limit', 72))
# Fetch zero-gpu (cpu) spaces
spaces_data = fetch_trending_spaces(offset, limit)
# Process and filter spaces
results = []
for index, space_data in enumerate(spaces_data['spaces']):
space_info = get_space_details(space_data, index, offset)
if not space_info:
continue
# 검색어 필터
if search_query:
if (search_query not in space_info['title'].lower()
and search_query not in space_info['owner'].lower()
and search_query not in space_info['url'].lower()
and search_query not in space_info['description'].lower()):
continue
results.append(space_info)
# 오너 통계 (Top 500 → Top 30)
top_owners = get_owner_stats(spaces_data.get('all_spaces', []))
return jsonify({
'spaces': results,
'total': spaces_data['total'],
'offset': offset,
'limit': limit,
'top_owners': top_owners
})
# Zero-GPU spaces API (Latest Releases)
@app.route('/api/latest-spaces', methods=['GET'])
def latest_spaces():
"""
hardware=cpu 스페이스 중에서 createdAt 기준으로 최신순 500개를 페이징, 검색
"""
search_query = request.args.get('search', '').lower()
offset = int(request.args.get('offset', 0))
limit = int(request.args.get('limit', 72))
spaces_data = fetch_latest_spaces(offset, limit)
results = []
for index, space_data in enumerate(spaces_data['spaces']):
space_info = get_space_details(space_data, index, offset)
if not space_info:
continue
# 검색어 필터
if search_query:
if (search_query not in space_info['title'].lower()
and search_query not in space_info['owner'].lower()
and search_query not in space_info['url'].lower()
and search_query not in space_info['description'].lower()):
continue
results.append(space_info)
return jsonify({
'spaces': results,
'total': spaces_data['total'],
'offset': offset,
'limit': limit
})
if __name__ == '__main__':
"""
서버 구동 시, templates/index.html 파일을 생성 후 Flask 실행
"""
# Create templates folder if not exists
os.makedirs('templates', exist_ok=True)
# index.html 전체를 새로 작성
with open('templates/index.html', 'w', encoding='utf-8') as f:
f.write('''
Huggingface Zero-GPU Spaces
Huggingface Explorer
ZeroGPU Spaces Leaderboard
Discover Zero GPU(Shared A100) spaces from Hugging Face
Creator Statistics
Top 30 Creators by Number of Spaces Ranked within Top 500
Loading Zero-GPU spaces...
If this takes too long, try refreshing the page.
''')
if __name__ == '__main__':
port = int(os.environ.get("PORT", 7860))
app.run(host='0.0.0.0', port=port)