main.py

import re
import urllib.parse
from fastapi import FastAPI, Request, HTTPException, Response
import httpx
from bs4 import BeautifulSoup

app = FastAPI()

# ---------------------------------------------------------------------------
# Utility: Rewrite CSS URLs inside CSS text
# ---------------------------------------------------------------------------
def rewrite_css(css_text: str, target_url: str, proxy_prefix="/?url=") -> str:
    """
    Finds all url(...) references in CSS and rewrites them so that
    the resources are loaded via the proxy.
    """
    # Matches: url('...'), url("..."), or url(...)
    pattern = re.compile(r'url\(\s*(?P<quote>["\']?)(?P<url>[^"\')]+)(?P=quote)\s*\)')
    
    def replace_url(match):
        original_url = match.group("url")
        # Skip if already proxied, is a data URL, or if it's already absolute and served correctly.
        if original_url.startswith(proxy_prefix) or original_url.startswith("data:"):
            return match.group(0)
        new_url = urllib.parse.urljoin(target_url, original_url)
        proxied = f'url({match.group("quote")}{proxy_prefix}{urllib.parse.quote(new_url)}{match.group("quote")})'
        return proxied

    return pattern.sub(replace_url, css_text)

# ---------------------------------------------------------------------------
# Injected JavaScript for Dynamic Interception & Real-Time Updates
# ---------------------------------------------------------------------------
# This script intercepts history changes, fetch, XHR, and anchor clicks,
# and uses MutationObserver to rewrite new elements. It also handles SVG
# attributes (including xlink:href) for icons and logos.
INJECTED_JS = """
<script>
(function() {
    // Intercept history.pushState to route dynamic navigations.
    const originalPushState = history.pushState;
    history.pushState = function(state, title, url) {
        if (url) {
            const proxiedUrl = '/?url=' + encodeURIComponent(url);
            return originalPushState.call(history, state, title, proxiedUrl);
        }
        return originalPushState.call(history, state, title, url);
    };

    // Intercept fetch() calls.
    const originalFetch = window.fetch;
    window.fetch = function(input, init) {
        let url;
        if (typeof input === 'string') {
            url = input;
        } else if (input && input.url) {
            url = input.url;
        } else {
            return originalFetch(input, init);
        }
        const proxiedUrl = '/?url=' + encodeURIComponent(url);
        if (typeof input === 'object') {
            input = new Request(proxiedUrl, input);
        } else {
            input = proxiedUrl;
        }
        return originalFetch(input, init);
    };

    // Intercept XMLHttpRequest.open().
    const originalOpen = XMLHttpRequest.prototype.open;
    XMLHttpRequest.prototype.open = function(method, url) {
        const proxiedUrl = '/?url=' + encodeURIComponent(url);
        return originalOpen.apply(this, [method, proxiedUrl, true]);
    };

    // Intercept anchor clicks to ensure navigation goes through the proxy.
    document.addEventListener('click', function(event) {
        const target = event.target.closest('a');
        if (target && target.href) {
            if (target.getAttribute('data-no-proxy') || target.href.indexOf('/?url=') === 0) {
                return;
            }
            event.preventDefault();
            window.location.href = '/?url=' + encodeURIComponent(target.href);
        }
    });

    // Use MutationObserver to catch and rewrite dynamically added elements.
    const observer = new MutationObserver(function(mutations) {
        mutations.forEach(function(mutation) {
            mutation.addedNodes.forEach(function(node) {
                if (node.nodeType === Node.ELEMENT_NODE) {
                    reProxyElement(node);
                }
            });
        });
    });
    observer.observe(document.body, { childList: true, subtree: true });

    // Rewrites URL attributes for an element and its children.
    function reProxyElement(element) {
        const urlAttrs = ['href', 'src', 'action', 'srcset', 'xlink:href'];
        urlAttrs.forEach(function(attr) {
            if (element.hasAttribute(attr)) {
                const value = element.getAttribute(attr);
                if (value && !value.startsWith('/?url=') &&
                    !value.startsWith('javascript:') && !value.startsWith('mailto:')) {
                    element.setAttribute(attr, '/?url=' + encodeURIComponent(value));
                }
            }
        });
        // Also rewrite inline style attribute.
        if (element.hasAttribute('style')) {
            let styleVal = element.getAttribute('style');
            // Simple client-side rewriting: prepend proxy to URLs.
            styleVal = styleVal.replace(/url\\((['"]?)(.*?)\\1\\)/g, function(match, quote, url) {
                if (url.startsWith('/?url=') || url.startsWith('data:')) return match;
                return "url(" + quote + "/?url=" + encodeURIComponent(url) + quote + ")";
            });
            element.setAttribute('style', styleVal);
        }
        Array.from(element.children).forEach(child => reProxyElement(child));
    }
})();
</script>
"""

# ---------------------------------------------------------------------------
# Helper Function: Rewrite URLs for All Relevant Attributes in HTML
# ---------------------------------------------------------------------------
def rewrite_urls(soup, target_url, proxy_prefix="/?url="):
    """
    Iterates over all elements in the parsed HTML and rewrites URL-like attributes.
    Supports attributes such as: href, src, action, srcset, xlink:href.
    Also rewrites inline style attributes that contain CSS url() references.
    """
    url_attrs = ['href', 'src', 'action', 'srcset', 'xlink:href']
    for element in soup.find_all(True):
        # Rewrite attributes that hold URLs.
        for attr in list(element.attrs):
            if attr in url_attrs:
                orig_value = element.get(attr)
                if not orig_value or orig_value.startswith(proxy_prefix) or orig_value.startswith(("mailto:", "javascript:")):
                    continue
                new_url = urllib.parse.urljoin(target_url, orig_value)
                element[attr] = proxy_prefix + urllib.parse.quote(new_url)
        # Rewrite inline style attributes containing CSS url() patterns.
        if element.has_attr("style"):
            original_style = element.get("style")
            new_style = rewrite_css(original_style, target_url, proxy_prefix)
            element["style"] = new_style
    return soup

# ---------------------------------------------------------------------------
# Core Function: Fetch and Rewrite the Target Resource
# ---------------------------------------------------------------------------
async def fetch_and_rewrite(target_url: str) -> Response:
    """
    Fetches the target URL and processes the response.
    - If the content is HTML, removes conflicting CSP tags, inserts a <base> tag,
      injects dynamic JavaScript, and rewrites URL attributes.
    - If the content is CSS, rewrites any url() references so that the resources are proxied.
    - Other content types are returned as-is.
    """
    async with httpx.AsyncClient() as client:
        resp = await client.get(target_url)
    
    content_type = resp.headers.get("Content-Type", "")
    
    # Process CSS files to rewrite url() references.
    if "text/css" in content_type:
        new_css = rewrite_css(resp.text, target_url)
        return Response(content=new_css, media_type="text/css", status_code=resp.status_code)
    
    # Process HTML content.
    if "text/html" in content_type:
        soup = BeautifulSoup(resp.text, "html.parser")
    
        # Remove any Content Security Policy meta tags that might block our scripts.
        for meta in soup.find_all("meta", attrs={"http-equiv": "Content-Security-Policy"}):
            meta.decompose()
    
        # Insert (or replace) a <base> tag to ensure relative URLs resolve properly.
        parsed_target = urllib.parse.urlparse(target_url)
        base_href = f"{parsed_target.scheme}://{parsed_target.netloc}"
        if soup.head:
            for base in soup.head.find_all("base"):
                base.decompose()
            base_tag = soup.new_tag("base", href=base_href)
            soup.head.insert(0, base_tag)
        else:
            head_tag = soup.new_tag("head")
            base_tag = soup.new_tag("base", href=base_href)
            head_tag.insert(0, base_tag)
            soup.insert(0, head_tag)
    
        # Inject dynamic JavaScript for real-time interception.
        if soup.body:
            soup.body.insert(0, BeautifulSoup(INJECTED_JS, "html.parser"))
        else:
            soup.insert(0, BeautifulSoup(INJECTED_JS, "html.parser"))
    
        # Rewrite URLs (including SVG icons, CSS backgrounds in style attributes, etc.).
        soup = rewrite_urls(soup, target_url)
    
        return Response(content=str(soup), media_type="text/html", status_code=resp.status_code)
    
    # For other content types (JS, images, etc.) return the response directly.
    return Response(content=resp.content, media_type=content_type, status_code=resp.status_code)

# ---------------------------------------------------------------------------
# Simple Proxy Endpoint: /proxy_full
# ---------------------------------------------------------------------------
@app.get("/proxy_full")
async def proxy_full(url: str):
    """
    A simple proxy endpoint that fetches the given URL and rewrites its HTML/CSS so that all resource URLs route via the proxy.
    """
    if not url:
        raise HTTPException(status_code=400, detail="Missing 'url' query parameter")
    
    async with httpx.AsyncClient() as client:
        resp = await client.get(url)
    
    content_type = resp.headers.get("Content-Type", "")
    if "text/html" in content_type:
        soup = BeautifulSoup(resp.text, "html.parser")
        soup = rewrite_urls(soup, url)
        return Response(content=str(soup), media_type="text/html")
    elif "text/css" in content_type:
        new_css = rewrite_css(resp.text, url)
        return Response(content=new_css, media_type="text/css")
    
    return Response(content=resp.content, media_type=content_type, status_code=resp.status_code)

# ---------------------------------------------------------------------------
# Catch-All Dynamic Proxy Endpoint
# ---------------------------------------------------------------------------
@app.get("/{full_path:path}")
async def catch_all(full_path: str, request: Request):
    """
    Catch-all endpoint for dynamic proxying.
    Determines the target URL via a query parameter or a stored cookie,
    then processes the resource (HTML or CSS) to inject dynamic JS and rewrite URLs.
    """
    query_params = dict(request.query_params)
    if "url" in query_params:
        target_url = query_params["url"]
    else:
        target_base = request.cookies.get("target_base")
        if not target_base:
            return Response("No target URL provided.", status_code=400)
        qs = request.url.query
        target_url = urllib.parse.urljoin(target_base, full_path)
        if qs:
            target_url += "?" + qs

    response = await fetch_and_rewrite(target_url)
    
    # Store the target's base URL in a cookie for subsequent requests.
    parsed_target = urllib.parse.urlparse(target_url)
    base_url = f"{parsed_target.scheme}://{parsed_target.netloc}"
    response.set_cookie("target_base", base_url)
    
    return response

# ---------------------------------------------------------------------------
# Run the Application on Port 7860
# ---------------------------------------------------------------------------
if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="0.0.0.0", port=7860)