Update app.py
Browse files
app.py
CHANGED
|
@@ -6,7 +6,9 @@ from bs4 import BeautifulSoup
|
|
| 6 |
|
| 7 |
app = FastAPI()
|
| 8 |
|
| 9 |
-
#
|
|
|
|
|
|
|
| 10 |
|
| 11 |
INDEX_HTML = """
|
| 12 |
<!DOCTYPE html>
|
|
@@ -62,7 +64,9 @@ async def index():
|
|
| 62 |
return INDEX_HTML
|
| 63 |
|
| 64 |
|
| 65 |
-
#
|
|
|
|
|
|
|
| 66 |
|
| 67 |
def normalize_url(raw: str) -> str:
|
| 68 |
"""如果没有 scheme，自动加上 http://"""
|
|
@@ -87,17 +91,22 @@ def is_allowed_url(url: str) -> bool:
|
|
| 87 |
# 禁止访问一些明显的本地 / 内网地址
|
| 88 |
if host in ("localhost", "127.0.0.1"):
|
| 89 |
return False
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 95 |
if any(host.startswith(p) for p in private_prefixes):
|
| 96 |
return False
|
| 97 |
|
| 98 |
return True
|
| 99 |
|
| 100 |
|
|
|
|
| 101 |
HOP_BY_HOP_HEADERS = {
|
| 102 |
"connection",
|
| 103 |
"keep-alive",
|
|
@@ -109,13 +118,20 @@ HOP_BY_HOP_HEADERS = {
|
|
| 109 |
"upgrade",
|
| 110 |
}
|
| 111 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 112 |
|
| 113 |
-
# ---- 重写 HTML 里的链接，使之继续走 /proxy ----
|
| 114 |
|
| 115 |
def rewrite_html(html: str, base_url: str) -> str:
|
|
|
|
|
|
|
|
|
|
|
|
|
| 116 |
soup = BeautifulSoup(html, "html.parser")
|
| 117 |
|
| 118 |
-
# 要处理的标签和对应属性
|
| 119 |
tag_attr_pairs = [
|
| 120 |
("a", "href"),
|
| 121 |
("link", "href"),
|
|
@@ -133,12 +149,12 @@ def rewrite_html(html: str, base_url: str) -> str:
|
|
| 133 |
if not value:
|
| 134 |
continue
|
| 135 |
|
| 136 |
-
# 锚点或 javascript: 等不处理
|
| 137 |
low = value.lower().strip()
|
|
|
|
| 138 |
if low.startswith("#") or low.startswith("javascript:"):
|
| 139 |
continue
|
| 140 |
|
| 141 |
-
#
|
| 142 |
absolute = urljoin(base_url, value)
|
| 143 |
proxied = f"/proxy?url={quote(absolute, safe='')}"
|
| 144 |
|
|
@@ -147,7 +163,9 @@ def rewrite_html(html: str, base_url: str) -> str:
|
|
| 147 |
return str(soup)
|
| 148 |
|
| 149 |
|
| 150 |
-
#
|
|
|
|
|
|
|
| 151 |
|
| 152 |
@app.api_route(
|
| 153 |
"/proxy",
|
|
@@ -166,7 +184,7 @@ async def proxy(request: Request, url: str):
|
|
| 166 |
detail="仅支持公网 http/https 地址，且不允许访问内网 / 本地地址。",
|
| 167 |
)
|
| 168 |
|
| 169 |
-
# 2. 准备要转发的请求头（过滤掉 hop-by-hop
|
| 170 |
outgoing_headers = {}
|
| 171 |
for k, v in request.headers.items():
|
| 172 |
lk = k.lower()
|
|
@@ -180,7 +198,10 @@ async def proxy(request: Request, url: str):
|
|
| 180 |
body = await request.body()
|
| 181 |
|
| 182 |
# 3. 用 httpx 转发请求
|
| 183 |
-
async with httpx.AsyncClient(
|
|
|
|
|
|
|
|
|
|
| 184 |
upstream_resp = await client.request(
|
| 185 |
request.method,
|
| 186 |
target_url,
|
|
@@ -190,30 +211,39 @@ async def proxy(request: Request, url: str):
|
|
| 190 |
|
| 191 |
content_type = upstream_resp.headers.get("content-type", "")
|
| 192 |
|
| 193 |
-
# 4.
|
| 194 |
response_headers = {}
|
| 195 |
for k, v in upstream_resp.headers.items():
|
| 196 |
lk = k.lower()
|
| 197 |
if lk in HOP_BY_HOP_HEADERS:
|
| 198 |
continue
|
| 199 |
-
if lk
|
| 200 |
-
#
|
|
|
|
|
|
|
| 201 |
continue
|
| 202 |
response_headers[k] = v
|
| 203 |
|
| 204 |
# 5. 如果是 HTML，就重写里面的链接
|
| 205 |
if "text/html" in content_type:
|
|
|
|
| 206 |
rewritten_html = rewrite_html(
|
| 207 |
upstream_resp.text,
|
| 208 |
base_url=str(upstream_resp.url),
|
| 209 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 210 |
return HTMLResponse(
|
| 211 |
content=rewritten_html,
|
| 212 |
status_code=upstream_resp.status_code,
|
| 213 |
headers=response_headers,
|
| 214 |
)
|
| 215 |
|
| 216 |
-
# 6. 其他类型（CSS/JS
|
|
|
|
| 217 |
return Response(
|
| 218 |
content=upstream_resp.content,
|
| 219 |
status_code=upstream_resp.status_code,
|
|
|
|
| 6 |
|
| 7 |
app = FastAPI()
|
| 8 |
|
| 9 |
+
# ==========================
|
| 10 |
+
# 首页：输入网址的简单页面
|
| 11 |
+
# ==========================
|
| 12 |
|
| 13 |
INDEX_HTML = """
|
| 14 |
<!DOCTYPE html>
|
|
|
|
| 64 |
return INDEX_HTML
|
| 65 |
|
| 66 |
|
| 67 |
+
# ==========================
|
| 68 |
+
# 工具函数
|
| 69 |
+
# ==========================
|
| 70 |
|
| 71 |
def normalize_url(raw: str) -> str:
|
| 72 |
"""如果没有 scheme，自动加上 http://"""
|
|
|
|
| 91 |
# 禁止访问一些明显的本地 / 内网地址
|
| 92 |
if host in ("localhost", "127.0.0.1"):
|
| 93 |
return False
|
| 94 |
+
|
| 95 |
+
private_prefixes = (
|
| 96 |
+
"10.",
|
| 97 |
+
"192.168.",
|
| 98 |
+
"172.16.", "172.17.", "172.18.", "172.19.",
|
| 99 |
+
"172.20.", "172.21.", "172.22.", "172.23.",
|
| 100 |
+
"172.24.", "172.25.", "172.26.", "172.27.",
|
| 101 |
+
"172.28.", "172.29.", "172.30.", "172.31.",
|
| 102 |
+
)
|
| 103 |
if any(host.startswith(p) for p in private_prefixes):
|
| 104 |
return False
|
| 105 |
|
| 106 |
return True
|
| 107 |
|
| 108 |
|
| 109 |
+
# hop-by-hop 头：代理两边不应转发
|
| 110 |
HOP_BY_HOP_HEADERS = {
|
| 111 |
"connection",
|
| 112 |
"keep-alive",
|
|
|
|
| 118 |
"upgrade",
|
| 119 |
}
|
| 120 |
|
| 121 |
+
# 会导致被 iframe 拦截的头，在 Hugging Face 的 iframe 环境里可以适当去掉
|
| 122 |
+
BLOCKED_HEADERS = {
|
| 123 |
+
"x-frame-options",
|
| 124 |
+
"content-security-policy",
|
| 125 |
+
}
|
| 126 |
|
|
|
|
| 127 |
|
| 128 |
def rewrite_html(html: str, base_url: str) -> str:
|
| 129 |
+
"""
|
| 130 |
+
重写 HTML 里的链接，使之继续走 /proxy。
|
| 131 |
+
处理 a/link/img/script/iframe/video/source/form 等常见标签。
|
| 132 |
+
"""
|
| 133 |
soup = BeautifulSoup(html, "html.parser")
|
| 134 |
|
|
|
|
| 135 |
tag_attr_pairs = [
|
| 136 |
("a", "href"),
|
| 137 |
("link", "href"),
|
|
|
|
| 149 |
if not value:
|
| 150 |
continue
|
| 151 |
|
|
|
|
| 152 |
low = value.lower().strip()
|
| 153 |
+
# 锚点、javascript: 不改
|
| 154 |
if low.startswith("#") or low.startswith("javascript:"):
|
| 155 |
continue
|
| 156 |
|
| 157 |
+
# 相对链接 -> 绝对 URL
|
| 158 |
absolute = urljoin(base_url, value)
|
| 159 |
proxied = f"/proxy?url={quote(absolute, safe='')}"
|
| 160 |
|
|
|
|
| 163 |
return str(soup)
|
| 164 |
|
| 165 |
|
| 166 |
+
# ==========================
|
| 167 |
+
# 反向代理主逻辑
|
| 168 |
+
# ==========================
|
| 169 |
|
| 170 |
@app.api_route(
|
| 171 |
"/proxy",
|
|
|
|
| 184 |
detail="仅支持公网 http/https 地址，且不允许访问内网 / 本地地址。",
|
| 185 |
)
|
| 186 |
|
| 187 |
+
# 2. 准备要转发的请求头（过滤掉 hop-by-hop 头部和 Host）
|
| 188 |
outgoing_headers = {}
|
| 189 |
for k, v in request.headers.items():
|
| 190 |
lk = k.lower()
|
|
|
|
| 198 |
body = await request.body()
|
| 199 |
|
| 200 |
# 3. 用 httpx 转发请求
|
| 201 |
+
async with httpx.AsyncClient(
|
| 202 |
+
follow_redirects=True,
|
| 203 |
+
timeout=20.0,
|
| 204 |
+
) as client:
|
| 205 |
upstream_resp = await client.request(
|
| 206 |
request.method,
|
| 207 |
target_url,
|
|
|
|
| 211 |
|
| 212 |
content_type = upstream_resp.headers.get("content-type", "")
|
| 213 |
|
| 214 |
+
# 4. 回传响应头（删掉 hop-by-hop、content-length、以及会阻止 iframe 的头）
|
| 215 |
response_headers = {}
|
| 216 |
for k, v in upstream_resp.headers.items():
|
| 217 |
lk = k.lower()
|
| 218 |
if lk in HOP_BY_HOP_HEADERS:
|
| 219 |
continue
|
| 220 |
+
if lk == "content-length":
|
| 221 |
+
# 让 FastAPI 自己计算长度
|
| 222 |
+
continue
|
| 223 |
+
if lk in BLOCKED_HEADERS:
|
| 224 |
continue
|
| 225 |
response_headers[k] = v
|
| 226 |
|
| 227 |
# 5. 如果是 HTML，就重写里面的链接
|
| 228 |
if "text/html" in content_type:
|
| 229 |
+
# upstream_resp.text 会根据 charset 自动解码成字符串
|
| 230 |
rewritten_html = rewrite_html(
|
| 231 |
upstream_resp.text,
|
| 232 |
base_url=str(upstream_resp.url),
|
| 233 |
)
|
| 234 |
+
|
| 235 |
+
# 这里我们已经重新编码 HTML 了，所以不能再带 content-encoding
|
| 236 |
+
response_headers.pop("content-encoding", None)
|
| 237 |
+
response_headers.pop("Content-Encoding", None)
|
| 238 |
+
|
| 239 |
return HTMLResponse(
|
| 240 |
content=rewritten_html,
|
| 241 |
status_code=upstream_resp.status_code,
|
| 242 |
headers=response_headers,
|
| 243 |
)
|
| 244 |
|
| 245 |
+
# 6. 其他类型（CSS/JS/图片/二进制等）原样透传
|
| 246 |
+
# 保留 Content-Encoding，这样浏览器才能正确解压 / 展示，避免乱码
|
| 247 |
return Response(
|
| 248 |
content=upstream_resp.content,
|
| 249 |
status_code=upstream_resp.status_code,
|