add proxi

.dockerignore

@@ -14,3 +14,4 @@ coverage.xml
 *.egg-info/
 dist/
 build/
+artifacts/

Dockerfile

@@ -5,6 +5,10 @@ ENV PYTHONDONTWRITEBYTECODE=1 \
 
 WORKDIR /app
 
+# Xvfb for headed Chromium in container (IAAI blocks headless)
+RUN apt-get update -qq && apt-get install -y --no-install-recommends xvfb \
+    && rm -rf /var/lib/apt/lists/*
+
 COPY requirements.txt ./
 RUN pip install --no-cache-dir -r requirements.txt
 

entrypoint.sh

@@ -1,4 +1,31 @@
 #!/bin/bash
 set -e
 
+# Start HTTP→SOCKS5 proxy bridge if SOCKS5 upstream is configured
+if [ -n "$SOCKS5_PROXY_HOST" ]; then
+  echo "[entrypoint] Starting proxy bridge (HTTP :8899 → SOCKS5 $SOCKS5_PROXY_HOST:${SOCKS5_PROXY_PORT:-1002})..."
+  if [ -z "$IAAI_PROXY_SERVER" ]; then
+    export IAAI_PROXY_SERVER="http://127.0.0.1:8899"
+  elif [ "$IAAI_PROXY_SERVER" = "http://localhost:8899" ]; then
+    export IAAI_PROXY_SERVER="http://127.0.0.1:8899"
+  fi
+  echo "[entrypoint] Using browser proxy: $IAAI_PROXY_SERVER"
+  python -m iaai_scraper.proxy_bridge &
+  BRIDGE_PID=$!
+  sleep 1
+  if ! kill -0 $BRIDGE_PID 2>/dev/null; then
+    echo "[entrypoint] ERROR: iaai_scraper.proxy_bridge failed to start"
+    exit 1
+  fi
+  echo "[entrypoint] Proxy bridge started (PID $BRIDGE_PID)"
+fi
+
+# IAAI blocks headless Chromium on Linux; run headed via Xvfb virtual display
+export DISPLAY=:99
+export IAAI_HEADLESS=false
+Xvfb :99 -screen 0 1920x1080x24 -nolisten tcp &
+XVFB_PID=$!
+sleep 0.5
+echo "[entrypoint] Xvfb started (PID $XVFB_PID, DISPLAY=$DISPLAY)"
+
 exec "$@"

iaai_scraper/proxy_bridge.py

@@ -0,0 +1,317 @@
+from __future__ import annotations
+
+import logging
+import os
+import select
+import socket
+import socketserver
+import struct
+import threading
+from urllib.parse import urlsplit
+
+BUFFER_SIZE = 65536
+CRLF = b"\r\n"
+DEFAULT_LISTEN_HOST = os.getenv("PROXY_BRIDGE_HOST", "0.0.0.0")
+DEFAULT_LISTEN_PORT = int(os.getenv("PROXY_BRIDGE_PORT", "8899"))
+SOCKS5_HOST = os.getenv("SOCKS5_PROXY_HOST", "")
+SOCKS5_PORT = int(os.getenv("SOCKS5_PROXY_PORT", "1002"))
+SOCKS5_USER = os.getenv("SOCKS5_PROXY_USER", "")
+SOCKS5_PASS = os.getenv("SOCKS5_PROXY_PASS", "")
+RELAY_IDLE_TIMEOUT_SECONDS = int(os.getenv("PROXY_BRIDGE_RELAY_IDLE_TIMEOUT_SECONDS", "60"))
+MAX_WORKERS = int(os.getenv("PROXY_BRIDGE_MAX_WORKERS", "64"))
+
+logger = logging.getLogger("proxy_bridge")
+
+
+class ThreadingTCPServer(socketserver.ThreadingMixIn, socketserver.TCPServer):
+    allow_reuse_address = True
+    daemon_threads = True
+
+    def __init__(self, server_address, request_handler_class):
+        super().__init__(server_address, request_handler_class)
+        self._worker_semaphore = threading.BoundedSemaphore(MAX_WORKERS)
+
+    def process_request_thread(self, request, client_address):
+        with self._worker_semaphore:
+            super().process_request_thread(request, client_address)
+
+
+def _recv_exact(sock: socket.socket, size: int) -> bytes:
+    data = b""
+    while len(data) < size:
+        chunk = sock.recv(size - len(data))
+        if not chunk:
+            raise ConnectionError("Unexpected EOF from SOCKS5 server")
+        data += chunk
+    return data
+
+
+def _socks5_connect(host: str, port: int) -> socket.socket:
+    if not SOCKS5_HOST:
+        raise RuntimeError("SOCKS5_PROXY_HOST is not configured")
+
+    upstream = socket.create_connection((SOCKS5_HOST, SOCKS5_PORT), timeout=30)
+    upstream.settimeout(30)
+
+    methods = [0x00]
+    if SOCKS5_USER or SOCKS5_PASS:
+        methods = [0x02]
+    upstream.sendall(bytes([0x05, len(methods), *methods]))
+    version, method = _recv_exact(upstream, 2)
+    if version != 0x05 or method == 0xFF:
+        upstream.close()
+        raise ConnectionError("SOCKS5 authentication negotiation failed")
+
+    if method == 0x02:
+        username = SOCKS5_USER.encode("utf-8")
+        password = SOCKS5_PASS.encode("utf-8")
+        if len(username) > 255 or len(password) > 255:
+            upstream.close()
+            raise ValueError("SOCKS5 username/password too long")
+        upstream.sendall(bytes([0x01, len(username)]) + username + bytes([len(password)]) + password)
+        auth_version, auth_status = _recv_exact(upstream, 2)
+        if auth_version != 0x01 or auth_status != 0x00:
+            upstream.close()
+            raise ConnectionError("SOCKS5 username/password authentication failed")
+
+    try:
+        socket.inet_aton(host)
+        addr_type = 0x01
+        addr_payload = socket.inet_aton(host)
+    except OSError:
+        host_bytes = host.encode("idna")
+        if len(host_bytes) > 255:
+            upstream.close()
+            raise ValueError("Target host is too long for SOCKS5 domain format")
+        addr_type = 0x03
+        addr_payload = bytes([len(host_bytes)]) + host_bytes
+
+    request = bytes([0x05, 0x01, 0x00, addr_type]) + addr_payload + struct.pack("!H", port)
+    upstream.sendall(request)
+
+    response_head = _recv_exact(upstream, 4)
+    version, reply, _reserved, reply_addr_type = response_head
+    if version != 0x05 or reply != 0x00:
+        upstream.close()
+        raise ConnectionError(f"SOCKS5 connect failed with code {reply}")
+
+    if reply_addr_type == 0x01:
+        _recv_exact(upstream, 4)
+    elif reply_addr_type == 0x03:
+        domain_len = _recv_exact(upstream, 1)[0]
+        _recv_exact(upstream, domain_len)
+    elif reply_addr_type == 0x04:
+        _recv_exact(upstream, 16)
+    _recv_exact(upstream, 2)
+
+    upstream.settimeout(RELAY_IDLE_TIMEOUT_SECONDS)
+    return upstream
+
+
+def _relay_bidirectional(left: socket.socket, right: socket.socket) -> None:
+    sockets = [left, right]
+    left.settimeout(RELAY_IDLE_TIMEOUT_SECONDS)
+    right.settimeout(RELAY_IDLE_TIMEOUT_SECONDS)
+    try:
+        while True:
+            readable, _, exceptional = select.select(sockets, [], sockets, RELAY_IDLE_TIMEOUT_SECONDS)
+            if exceptional:
+                break
+            if not readable:
+                logger.debug("Relay idle timeout reached; closing sockets")
+                return
+            for current in readable:
+                other = right if current is left else left
+                data = current.recv(BUFFER_SIZE)
+                if not data:
+                    return
+                other.sendall(data)
+    finally:
+        for sock in sockets:
+            try:
+                sock.shutdown(socket.SHUT_RDWR)
+            except OSError:
+                pass
+            try:
+                sock.close()
+            except OSError:
+                pass
+
+
+class ProxyHandler(socketserver.StreamRequestHandler):
+    def handle(self) -> None:
+        try:
+            request_line = self.rfile.readline(BUFFER_SIZE).decode("iso-8859-1").strip()
+            if not request_line:
+                return
+
+            method, target, version = request_line.split()
+            headers = self._read_headers()
+            logger.info("%s %s", method, target)
+
+            if method.upper() == "CONNECT":
+                host, port = self._parse_connect_target(target)
+                logger.info("CONNECT %s:%s", host, port)
+                upstream = _socks5_connect(host, port)
+                self.wfile.write(f"{version} 200 Connection Established".encode("ascii") + CRLF + CRLF)
+                self.wfile.flush()
+                _relay_bidirectional(self.connection, upstream)
+                return
+
+            host, port, path = self._parse_forward_target(target, headers)
+            upstream = _socks5_connect(host, port)
+            self._send_forward_request(upstream, method, path, version, headers)
+            body = self._read_request_body(headers)
+            if body:
+                upstream.sendall(body)
+            _relay_bidirectional(self.connection, upstream)
+        except Exception as exc:
+            logger.exception("Proxy bridge request failed: %s", exc)
+            try:
+                self.wfile.write(
+                    b"HTTP/1.1 502 Bad Gateway" + CRLF
+                    + b"Connection: close" + CRLF
+                    + b"Content-Type: text/plain; charset=utf-8" + CRLF + CRLF
+                    + b"Bad Gateway"
+                )
+                self.wfile.flush()
+            except OSError:
+                pass
+
+    def _read_headers(self) -> list[tuple[str, str]]:
+        headers: list[tuple[str, str]] = []
+        while True:
+            line = self.rfile.readline(BUFFER_SIZE)
+            if line in {CRLF, b"\n", b""}:
+                break
+            decoded = line.decode("iso-8859-1")
+            if ":" not in decoded:
+                continue
+            name, value = decoded.split(":", 1)
+            headers.append((name.strip(), value.strip()))
+        return headers
+
+    @staticmethod
+    def _parse_connect_target(target: str) -> tuple[str, int]:
+        if target.startswith("["):
+            end = target.find("]")
+            if end == -1 or len(target) <= end + 2 or target[end + 1] != ":":
+                raise ValueError("Invalid CONNECT target")
+            host = target[1:end]
+            port_text = target[end + 2 :]
+            return host, int(port_text)
+
+        host, port_text = target.rsplit(":", 1)
+        return host, int(port_text)
+
+    @staticmethod
+    def _parse_forward_target(target: str, headers: list[tuple[str, str]]) -> tuple[str, int, str]:
+        if target.startswith("http://"):
+            parts = urlsplit(target)
+            port = parts.port or 80
+            path = parts.path or "/"
+            if parts.query:
+                path += f"?{parts.query}"
+            return parts.hostname or "", port, path
+
+        if target.startswith("https://"):
+            raise ValueError("HTTPS absolute-form request must use CONNECT")
+
+        host_header = next((value for name, value in headers if name.lower() == "host"), "")
+        if not host_header:
+            raise ValueError("Missing Host header")
+        if ":" in host_header:
+            host, port_text = host_header.rsplit(":", 1)
+            return host, int(port_text), target
+        return host_header, 80, target
+
+    def _send_forward_request(
+        self,
+        upstream: socket.socket,
+        method: str,
+        path: str,
+        version: str,
+        headers: list[tuple[str, str]],
+    ) -> None:
+        filtered_headers: list[tuple[str, str]] = []
+        hop_by_hop = {
+            "proxy-connection",
+            "proxy-authorization",
+            "connection",
+            "keep-alive",
+            "te",
+            "trailer",
+            "transfer-encoding",
+            "upgrade",
+        }
+        for name, value in headers:
+            if name.lower() in hop_by_hop:
+                continue
+            filtered_headers.append((name, value))
+
+        request_head = [f"{method} {path} {version}\r\n"]
+        request_head.extend(f"{name}: {value}\r\n" for name, value in filtered_headers)
+        request_head.append("\r\n")
+        upstream.sendall("".join(request_head).encode("iso-8859-1"))
+
+    def _read_request_body(self, headers: list[tuple[str, str]]) -> bytes:
+        transfer_encoding = next((value for name, value in headers if name.lower() == "transfer-encoding"), "")
+        if "chunked" in transfer_encoding.lower():
+            return self._read_chunked_request_body()
+
+        content_length = next((value for name, value in headers if name.lower() == "content-length"), None)
+        if not content_length:
+            return b""
+        return self.rfile.read(int(content_length))
+
+    def _read_chunked_request_body(self) -> bytes:
+        chunks: list[bytes] = []
+        while True:
+            size_line = self.rfile.readline(BUFFER_SIZE)
+            if not size_line:
+                raise ConnectionError("Unexpected EOF in chunked request")
+            size_text = size_line.strip().split(b";", 1)[0]
+            chunk_size = int(size_text, 16)
+            chunks.append(size_line)
+            if chunk_size == 0:
+                while True:
+                    trailer_line = self.rfile.readline(BUFFER_SIZE)
+                    if not trailer_line:
+                        raise ConnectionError("Unexpected EOF in chunked trailers")
+                    chunks.append(trailer_line)
+                    if trailer_line in {CRLF, b"\n"}:
+                        return b"".join(chunks)
+
+            chunk_data = self.rfile.read(chunk_size)
+            if len(chunk_data) != chunk_size:
+                raise ConnectionError("Unexpected EOF in chunk body")
+            chunks.append(chunk_data)
+            chunk_end = self.rfile.read(2)
+            if chunk_end != CRLF:
+                raise ConnectionError("Invalid chunk terminator")
+            chunks.append(chunk_end)
+
+
+def main() -> None:
+    if not SOCKS5_HOST:
+        raise SystemExit("SOCKS5_PROXY_HOST is required")
+
+    logging.basicConfig(
+        level=os.getenv("PROXY_BRIDGE_LOG_LEVEL", "INFO").upper(),
+        format="[%(asctime)s] [proxy_bridge] %(levelname)s: %(message)s",
+    )
+
+    with ThreadingTCPServer((DEFAULT_LISTEN_HOST, DEFAULT_LISTEN_PORT), ProxyHandler) as server:
+        logger.info(
+            "Listening on %s:%s -> socks5://%s:%s (max_workers=%s)",
+            DEFAULT_LISTEN_HOST,
+            DEFAULT_LISTEN_PORT,
+            SOCKS5_HOST,
+            SOCKS5_PORT,
+            MAX_WORKERS,
+        )
+        server.serve_forever()
+
+
+if __name__ == "__main__":
+    main()

requirements.txt

@@ -2,5 +2,6 @@ playwright>=1.53.0
 python-dotenv>=1.0.1
 pydantic>=2.8.2
 SQLAlchemy>=2.0.32
+PySocks>=1.7.1
 pytest>=8.3.0
 pytest-cov>=5.0.0