fix(codex-bridge): ship zstd for the agent's Python and refuse to link without it

Codex compresses request bodies with zstd. The agent venv is Python 3.11, which has no zstd in its standard library, so every Codex request failed on a fresh install. The matching zstandard build now ships beside the bridge; the status endpoint reports ok=false when bodies cannot be read, and the IDE then leaves Codex's config alone. A running bridge from an older build is replaced, and "Sync now" makes the bridge re-read the IDE session at once. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01HRDGW1LiKcDUZYUdKBoWYW
2026-10-09 16:27:55 +03:00
13 changed files with 5317 additions and 10 deletions
@@ -26,9 +26,42 @@ from resim_context import resimleri_sadelestir
 import ide_catalog
 import ozet_kopru

-SURUM = "1"
+SURUM = "2"
 HERE = os.path.dirname(os.path.abspath(__file__))

+# Codex sends request bodies zstd-compressed. Python gained zstd in 3.14; for older
+# interpreters the matching `zstandard` build ships next to this file.
+_VENDOR = os.path.join(HERE, "vendor", "cp%d%d-win_amd64" % sys.version_info[:2])
+if os.path.isdir(_VENDOR):
+    sys.path.insert(0, _VENDOR)
+
+
+def _zstd_decompress(body):
+    try:
+        from compression import zstd as _z
+        return _z.decompress(body)
+    except ImportError:
+        import zstandard
+        return zstandard.ZstdDecompressor().decompressobj().decompress(body)
+
+
+def _zstd_ready():
+    try:
+        try:
+            from compression import zstd as _z
+            sample = _z.compress(b"aiturk")
+        except ImportError:
+            import zstandard
+            sample = zstandard.ZstdCompressor().compress(b"aiturk")
+        return _zstd_decompress(sample) == b"aiturk"
+    except Exception:
+        return False
+
+
+# Without zstd every Codex request would fail; the status endpoint then reports
+# ok=false so that the IDE never points Codex at this bridge.
+ZSTD_OK = _zstd_ready()
+

 def _arg(name, env):
    """--name <value> from the command line, else the environment variable."""
@@ -468,10 +501,7 @@ def decompress(body, enc):
    if enc == "deflate":
        import zlib; return zlib.decompress(body)
    if enc == "zstd":
-        try:
-            from compression import zstd as _z; return _z.decompress(body)
-        except ImportError:
-            import zstandard; return zstandard.ZstdDecompressor().decompressobj().decompress(body)
+        return _zstd_decompress(body)
    if enc == "br":
        import brotli; return brotli.decompress(body)
    return body
@@ -533,9 +563,18 @@ class H(http.server.BaseHTTPRequestHandler):

    def _control(self, path):
        """Local status/stop for AITURK IDE. Loopback only; carries no secrets."""
+        if self.command == "POST" and path.rstrip("/").endswith("/refresh"):
+            # The IDE asks for this after a sign-in or a settings change, so the model
+            # list does not wait for the next periodic refresh.
+            try:
+                load_ide_catalog()
+            except Exception as e:
+                log(f"katalog yenileme hatasi: {e}")
+            path = "/aiturk/codex-baglama/status"
+            self.command = "GET"
        if self.command == "GET" and path.rstrip("/").endswith("/status"):
            durum, _, sayi = (IDE_DURUM[0] or "kapali, 0 model").partition(", ")
-            return self._send_json(200, {"ok": True, "surum": SURUM, "ide_durum": durum,
+            return self._send_json(200, {"ok": ZSTD_OK, "surum": SURUM, "zstd": ZSTD_OK, "ide_durum": durum,
                                         "ide_model": int((sayi.split() or ["0"])[0]),
                                         "ozet": bool((CFG.get("ozet_kopru") or {}).get("enabled", False))})
        if self.command == "POST" and path.rstrip("/").endswith("/stop"):
@@ -0,0 +1,35 @@
+"""Codex compresses request bodies with zstd; the bridge must be able to read them."""
+import ast
+import unittest
+from pathlib import Path
+
+tree = ast.parse(Path(__file__).with_name("bridge.py").read_text(encoding="utf-8"))
+wanted = {"_zstd_decompress", "_zstd_ready", "decompress"}
+env = {}
+exec(compile(ast.Module(body=[n for n in tree.body if isinstance(n, ast.FunctionDef) and n.name in wanted],
+                        type_ignores=[]), "bridge.py", "exec"), env)
+
+
+class Zstd(unittest.TestCase):
+    def test_self_check_matches_reality(self):
+        try:
+            try:
+                from compression import zstd as z
+                packed = z.compress(b"govde" * 50)
+            except ImportError:
+                import zstandard
+                packed = zstandard.ZstdCompressor().compress(b"govde" * 50)
+        except ImportError:
+            self.assertFalse(env["_zstd_ready"]())      # no zstd here: the bridge must say so
+            return
+        self.assertTrue(env["_zstd_ready"]())
+        self.assertEqual(env["decompress"](packed, "zstd"), b"govde" * 50)
+
+    def test_other_encodings(self):
+        import gzip
+        self.assertEqual(env["decompress"](gzip.compress(b"x"), "gzip"), b"x")
+        self.assertEqual(env["decompress"](b"duz", None), b"duz")
+
+
+if __name__ == "__main__":
+    unittest.main()
@@ -0,0 +1,27 @@
+Copyright (c) 2016, Gregory Szorc
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+
+1. Redistributions of source code must retain the above copyright notice, this
+list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright notice,
+this list of conditions and the following disclaimer in the documentation
+and/or other materials provided with the distribution.
+
+3. Neither the name of the copyright holder nor the names of its contributors
+may be used to endorse or promote products derived from this software without
+specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
+ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
+WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR
+ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+(INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
+ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
@@ -0,0 +1,217 @@
+# Copyright (c) 2017-present, Gregory Szorc
+# All rights reserved.
+#
+# This software may be modified and distributed under the terms
+# of the BSD license. See the LICENSE file for details.
+
+# ruff: noqa: F403, F405
+
+"""Python interface to the Zstandard (zstd) compression library."""
+
+from __future__ import absolute_import, unicode_literals
+
+# This module serves 2 roles:
+#
+# 1) Export the C or CFFI "backend" through a central module.
+# 2) Implement additional functionality built on top of C or CFFI backend.
+import builtins
+import io
+import os
+import platform
+import sys
+
+if sys.version_info >= (3, 12):
+    from collections.abc import Buffer
+else:
+    from typing import ByteString as Buffer
+
+# Some Python implementations don't support C extensions. That's why we have
+# a CFFI implementation in the first place. The code here import one of our
+# "backends" then re-exports the symbols from this module. For convenience,
+# we support falling back to the CFFI backend if the C extension can't be
+# imported. But for performance reasons, we only do this on unknown Python
+# implementation. Notably, for CPython we require the C extension by default.
+# Because someone will inevitably want special behavior, the behavior is
+# configurable via an environment variable. A potentially better way to handle
+# this is to import a special ``__importpolicy__`` module or something
+# defining a variable and `setup.py` could write the file with whatever
+# policy was specified at build time. Until someone needs it, we go with
+# the hacky but simple environment variable approach.
+_module_policy = os.environ.get(
+    "PYTHON_ZSTANDARD_IMPORT_POLICY", "default"
+).strip()
+
+if _module_policy == "default":
+    if platform.python_implementation() in ("CPython",):
+        from .backend_c import *  # type: ignore
+
+        backend = "cext"
+    elif platform.python_implementation() in ("PyPy",):
+        from .backend_cffi import *  # type: ignore
+
+        backend = "cffi"
+    else:
+        try:
+            from .backend_c import *
+
+            backend = "cext"
+        except ImportError:
+            from .backend_cffi import *
+
+            backend = "cffi"
+elif _module_policy == "cffi_fallback":
+    try:
+        from .backend_c import *
+
+        backend = "cext"
+    except ImportError:
+        from .backend_cffi import *
+
+        backend = "cffi"
+elif _module_policy == "rust":
+    from .backend_rust import *  # type: ignore
+
+    backend = "rust"
+elif _module_policy == "cext":
+    from .backend_c import *
+
+    backend = "cext"
+elif _module_policy == "cffi":
+    from .backend_cffi import *
+
+    backend = "cffi"
+else:
+    raise ImportError(
+        "unknown module import policy: %s; use default, cffi_fallback, "
+        "cext, or cffi" % _module_policy
+    )
+
+# Keep this in sync with python-zstandard.h, rust-ext/src/lib.rs, and debian/changelog.
+__version__ = "0.25.0"
+
+_MODE_CLOSED = 0
+_MODE_READ = 1
+_MODE_WRITE = 2
+
+
+def open(
+    filename,
+    mode="rb",
+    cctx=None,
+    dctx=None,
+    encoding=None,
+    errors=None,
+    newline=None,
+    closefd=None,
+):
+    """Create a file object with zstd (de)compression.
+
+    The object returned from this function will be a
+    :py:class:`ZstdDecompressionReader` if opened for reading in binary mode,
+    a :py:class:`ZstdCompressionWriter` if opened for writing in binary mode,
+    or an ``io.TextIOWrapper`` if opened for reading or writing in text mode.
+
+    :param filename:
+       ``bytes``, ``str``, or ``os.PathLike`` defining a file to open or a
+       file object (with a ``read()`` or ``write()`` method).
+    :param mode:
+       ``str`` File open mode. Accepts any of the open modes recognized by
+       ``open()``.
+    :param cctx:
+       ``ZstdCompressor`` to use for compression. If not specified and file
+       is opened for writing, the default ``ZstdCompressor`` will be used.
+    :param dctx:
+       ``ZstdDecompressor`` to use for decompression. If not specified and file
+       is opened for reading, the default ``ZstdDecompressor`` will be used.
+    :param encoding:
+        ``str`` that defines text encoding to use when file is opened in text
+        mode.
+    :param errors:
+       ``str`` defining text encoding error handling mode.
+    :param newline:
+       ``str`` defining newline to use in text mode.
+    :param closefd:
+       ``bool`` whether to close the file when the returned object is closed.
+        Only used if a file object is passed. If a filename is specified, the
+        opened file is always closed when the returned object is closed.
+    """
+    normalized_mode = mode.replace("t", "")
+
+    if normalized_mode in ("r", "rb"):
+        dctx = dctx or ZstdDecompressor()
+        open_mode = "r"
+        raw_open_mode = "rb"
+    elif normalized_mode in ("w", "wb", "a", "ab", "x", "xb"):
+        cctx = cctx or ZstdCompressor()
+        open_mode = "w"
+        raw_open_mode = normalized_mode
+        if not raw_open_mode.endswith("b"):
+            raw_open_mode = raw_open_mode + "b"
+    else:
+        raise ValueError("Invalid mode: {!r}".format(mode))
+
+    if hasattr(os, "PathLike"):
+        types = (str, bytes, os.PathLike)
+    else:
+        types = (str, bytes)
+
+    if isinstance(filename, types):  # type: ignore
+        inner_fh = builtins.open(filename, raw_open_mode)
+        closefd = True
+    elif hasattr(filename, "read") or hasattr(filename, "write"):
+        inner_fh = filename
+        closefd = bool(closefd)
+    else:
+        raise TypeError(
+            "filename must be a str, bytes, file or PathLike object"
+        )
+
+    if open_mode == "r":
+        fh = dctx.stream_reader(inner_fh, closefd=closefd)
+    elif open_mode == "w":
+        fh = cctx.stream_writer(inner_fh, closefd=closefd)
+    else:
+        raise RuntimeError("logic error in zstandard.open() handling open mode")
+
+    if "b" not in normalized_mode:
+        return io.TextIOWrapper(
+            fh, encoding=encoding, errors=errors, newline=newline
+        )
+    else:
+        return fh
+
+
+def compress(data: Buffer, level: int = 3) -> bytes:
+    """Compress source data using the zstd compression format.
+
+    This performs one-shot compression using basic/default compression
+    settings.
+
+    This method is provided for convenience and is equivalent to calling
+    ``ZstdCompressor(level=level).compress(data)``.
+
+    If you find yourself calling this function in a tight loop,
+    performance will be greater if you construct a single ``ZstdCompressor``
+    and repeatedly call ``compress()`` on it.
+    """
+    cctx = ZstdCompressor(level=level)
+
+    return cctx.compress(data)
+
+
+def decompress(data: Buffer, max_output_size: int = 0) -> bytes:
+    """Decompress a zstd frame into its original data.
+
+    This performs one-shot decompression using basic/default compression
+    settings.
+
+    This method is provided for convenience and is equivalent to calling
+    ``ZstdDecompressor().decompress(data, max_output_size=max_output_size)``.
+
+    If you find yourself calling this function in a tight loop, performance
+    will be greater if you construct a single ``ZstdDecompressor`` and
+    repeatedly call ``decompress()`` on it.
+    """
+    dctx = ZstdDecompressor()
+
+    return dctx.decompress(data, max_output_size=max_output_size)
@@ -0,0 +1,481 @@
+# Copyright (c) 2016-present, Gregory Szorc
+# All rights reserved.
+#
+# This software may be modified and distributed under the terms
+# of the BSD license. See the LICENSE file for details.
+
+import os
+from typing import (
+    IO,
+    BinaryIO,
+    ByteString,
+    Generator,
+    Iterable,
+    List,
+    Optional,
+    Set,
+    Tuple,
+    Union,
+)
+
+FLUSH_BLOCK: int
+FLUSH_FRAME: int
+
+COMPRESSOBJ_FLUSH_FINISH: int
+COMPRESSOBJ_FLUSH_BLOCK: int
+
+CONTENTSIZE_UNKNOWN: int
+CONTENTSIZE_ERROR: int
+
+MAX_COMPRESSION_LEVEL: int
+
+COMPRESSION_RECOMMENDED_INPUT_SIZE: int
+COMPRESSION_RECOMMENDED_OUTPUT_SIZE: int
+
+DECOMPRESSION_RECOMMENDED_INPUT_SIZE: int
+DECOMPRESSION_RECOMMENDED_OUTPUT_SIZE: int
+
+BLOCKSIZELOG_MAX: int
+BLOCKSIZE_MAX: int
+
+WINDOWLOG_MIN: int
+WINDOWLOG_MAX: int
+
+CHAINLOG_MIN: int
+CHAINLOG_MAX: int
+HASHLOG_MIN: int
+HASHLOG_MAX: int
+MINMATCH_MIN: int
+MINMATCH_MAX: int
+SEARCHLOG_MIN: int
+SEARCHLOG_MAX: int
+SEARCHLENGTH_MIN: int
+SEARCHLENGTH_MAX: int
+TARGETLENGTH_MIN: int
+TARGETLENGTH_MAX: int
+LDM_MINMATCH_MIN: int
+LDM_MINMATCH_MAX: int
+LDM_BUCKETSIZELOG_MAX: int
+
+STRATEGY_FAST: int
+STRATEGY_DFAST: int
+STRATEGY_GREEDY: int
+STRATEGY_LAZY: int
+STRATEGY_LAZY2: int
+STRATEGY_BTLAZY2: int
+STRATEGY_BTOPT: int
+STRATEGY_BTULTRA: int
+STRATEGY_BTULTRA2: int
+
+DICT_TYPE_AUTO: int
+DICT_TYPE_RAWCONTENT: int
+DICT_TYPE_FULLDICT: int
+
+FORMAT_ZSTD1: int
+FORMAT_ZSTD1_MAGICLESS: int
+
+ZSTD_VERSION: Tuple[int, int, int]
+FRAME_HEADER: bytes
+MAGIC_NUMBER: int
+
+backend: str
+backend_features: Set[str]
+__version__: str
+
+class ZstdError(Exception): ...
+
+class BufferSegment(object):
+    offset: int
+    def __len__(self) -> int: ...
+    def tobytes(self) -> bytes: ...
+
+class BufferSegments(object):
+    def __len__(self) -> int: ...
+    def __getitem__(self, i: int) -> BufferSegment: ...
+
+class BufferWithSegments(object):
+    size: int
+    def __init__(self, data: ByteString, segments: ByteString): ...
+    def __len__(self) -> int: ...
+    def __getitem__(self, i: int) -> BufferSegment: ...
+    def segments(self): ...
+    def tobytes(self) -> bytes: ...
+
+class BufferWithSegmentsCollection(object):
+    def __init__(self, *args): ...
+    def __len__(self) -> int: ...
+    def __getitem__(self, i: int) -> BufferSegment: ...
+    def size(self) -> int: ...
+
+class ZstdCompressionParameters(object):
+    @staticmethod
+    def from_level(
+        level: int, source_size: int = ..., dict_size: int = ..., **kwargs
+    ) -> "ZstdCompressionParameters": ...
+    def __init__(
+        self,
+        format: int = ...,
+        compression_level: int = ...,
+        window_log: int = ...,
+        hash_log: int = ...,
+        chain_log: int = ...,
+        search_log: int = ...,
+        min_match: int = ...,
+        target_length: int = ...,
+        strategy: int = ...,
+        write_content_size: int = ...,
+        write_checksum: int = ...,
+        write_dict_id: int = ...,
+        job_size: int = ...,
+        overlap_log: int = ...,
+        force_max_window: int = ...,
+        enable_ldm: int = ...,
+        ldm_hash_log: int = ...,
+        ldm_min_match: int = ...,
+        ldm_bucket_size_log: int = ...,
+        ldm_hash_rate_log: int = ...,
+        threads: int = ...,
+    ): ...
+    @property
+    def format(self) -> int: ...
+    @property
+    def compression_level(self) -> int: ...
+    @property
+    def window_log(self) -> int: ...
+    @property
+    def hash_log(self) -> int: ...
+    @property
+    def chain_log(self) -> int: ...
+    @property
+    def search_log(self) -> int: ...
+    @property
+    def min_match(self) -> int: ...
+    @property
+    def target_length(self) -> int: ...
+    @property
+    def strategy(self) -> int: ...
+    @property
+    def write_content_size(self) -> int: ...
+    @property
+    def write_checksum(self) -> int: ...
+    @property
+    def write_dict_id(self) -> int: ...
+    @property
+    def job_size(self) -> int: ...
+    @property
+    def overlap_log(self) -> int: ...
+    @property
+    def force_max_window(self) -> int: ...
+    @property
+    def enable_ldm(self) -> int: ...
+    @property
+    def ldm_hash_log(self) -> int: ...
+    @property
+    def ldm_min_match(self) -> int: ...
+    @property
+    def ldm_bucket_size_log(self) -> int: ...
+    @property
+    def ldm_hash_rate_log(self) -> int: ...
+    @property
+    def threads(self) -> int: ...
+    def estimated_compression_context_size(self) -> int: ...
+
+class CompressionParameters(ZstdCompressionParameters): ...
+
+class ZstdCompressionDict(object):
+    k: int
+    d: int
+    def __init__(
+        self,
+        data: ByteString,
+        dict_type: int = ...,
+        k: int = ...,
+        d: int = ...,
+    ): ...
+    def __len__(self) -> int: ...
+    def dict_id(self) -> int: ...
+    def as_bytes(self) -> bytes: ...
+    def precompute_compress(
+        self,
+        level: int = ...,
+        compression_params: ZstdCompressionParameters = ...,
+    ): ...
+
+class ZstdCompressionObj(object):
+    def compress(self, data: ByteString) -> bytes: ...
+    def flush(self, flush_mode: int = ...) -> bytes: ...
+
+class ZstdCompressionChunker(object):
+    def compress(self, data: ByteString): ...
+    def flush(self): ...
+    def finish(self): ...
+
+class ZstdCompressionReader(BinaryIO):
+    def __enter__(self) -> "ZstdCompressionReader": ...
+    def __exit__(self, exc_type, exc_value, exc_tb): ...
+    def readable(self) -> bool: ...
+    def writable(self) -> bool: ...
+    def seekable(self) -> bool: ...
+    def readline(self, limit: int = ...) -> bytes: ...
+    def readlines(self, hint: int = ...) -> List[bytes]: ...
+    def write(self, data: ByteString): ...
+    def writelines(self, data: Iterable[bytes]): ...
+    def isatty(self) -> bool: ...
+    def flush(self): ...
+    def close(self): ...
+    @property
+    def closed(self) -> bool: ...
+    def tell(self) -> int: ...
+    def readall(self) -> bytes: ...
+    def __iter__(self): ...
+    def __next__(self): ...
+    def next(self): ...
+    def read(self, size: int = ...) -> bytes: ...
+    def read1(self, size: int = ...) -> bytes: ...
+    def readinto(self, b) -> int: ...
+    def readinto1(self, b) -> int: ...
+
+class ZstdCompressionWriter(BinaryIO):
+    def __enter__(self) -> "ZstdCompressionWriter": ...
+    def __exit__(self, exc_type, exc_value, exc_tb): ...
+    def memory_size(self) -> int: ...
+    def fileno(self) -> int: ...
+    def close(self): ...
+    @property
+    def closed(self) -> bool: ...
+    def isatty(self) -> bool: ...
+    def readable(self) -> bool: ...
+    def readline(self, size: int = ...) -> bytes: ...
+    def readlines(self, hint: int = ...) -> List[bytes]: ...
+    def seek(self, offset: int, whence: int = ...): ...
+    def seekable(self) -> bool: ...
+    def truncate(self, size: int = ...): ...
+    def writable(self) -> bool: ...
+    def writelines(self, lines: Iterable[bytes]): ...
+    def read(self, size: int = ...) -> bytes: ...
+    def readall(self) -> bytes: ...
+    def readinto(self, b): ...
+    def write(self, data: ByteString) -> int: ...
+    def flush(self, flush_mode: int = ...) -> int: ...
+    def tell(self) -> int: ...
+
+class ZstdCompressor(object):
+    def __init__(
+        self,
+        level: int = ...,
+        dict_data: Optional[ZstdCompressionDict] = ...,
+        compression_params: Optional[ZstdCompressionParameters] = ...,
+        write_checksum: Optional[bool] = ...,
+        write_content_size: Optional[bool] = ...,
+        write_dict_id: Optional[bool] = ...,
+        threads: int = ...,
+    ): ...
+    def memory_size(self) -> int: ...
+    def compress(self, data: ByteString) -> bytes: ...
+    def compressobj(self, size: int = ...) -> ZstdCompressionObj: ...
+    def chunker(
+        self, size: int = ..., chunk_size: int = ...
+    ) -> ZstdCompressionChunker: ...
+    def copy_stream(
+        self,
+        ifh: IO[bytes],
+        ofh: IO[bytes],
+        size: int = ...,
+        read_size: int = ...,
+        write_size: int = ...,
+    ) -> Tuple[int, int]: ...
+    def stream_reader(
+        self,
+        source: Union[IO[bytes], ByteString],
+        size: int = ...,
+        read_size: int = ...,
+        *,
+        closefd: bool = ...,
+    ) -> ZstdCompressionReader: ...
+    def stream_writer(
+        self,
+        writer: IO[bytes],
+        size: int = ...,
+        write_size: int = ...,
+        write_return_read: bool = ...,
+        *,
+        closefd: bool = ...,
+    ) -> ZstdCompressionWriter: ...
+    def read_to_iter(
+        self,
+        reader: Union[IO[bytes], ByteString],
+        size: int = ...,
+        read_size: int = ...,
+        write_size: int = ...,
+    ) -> Generator[bytes, None, None]: ...
+    def frame_progression(self) -> Tuple[int, int, int]: ...
+    def multi_compress_to_buffer(
+        self,
+        data: Union[
+            BufferWithSegments,
+            BufferWithSegmentsCollection,
+            List[ByteString],
+        ],
+        threads: int = ...,
+    ) -> BufferWithSegmentsCollection: ...
+
+class ZstdDecompressionObj(object):
+    def decompress(self, data: ByteString) -> bytes: ...
+    def flush(self, length: int = ...) -> bytes: ...
+    @property
+    def unused_data(self) -> bytes: ...
+    @property
+    def unconsumed_tail(self) -> bytes: ...
+    @property
+    def eof(self) -> bool: ...
+
+class ZstdDecompressionReader(BinaryIO):
+    def __enter__(self) -> "ZstdDecompressionReader": ...
+    def __exit__(self, exc_type, exc_value, exc_tb): ...
+    def readable(self) -> bool: ...
+    def writable(self) -> bool: ...
+    def seekable(self) -> bool: ...
+    def readline(self, size: int = ...): ...
+    def readlines(self, hint: int = ...): ...
+    def write(self, data: ByteString): ...
+    def writelines(self, lines: Iterable[bytes]): ...
+    def isatty(self) -> bool: ...
+    def flush(self): ...
+    def close(self): ...
+    @property
+    def closed(self) -> bool: ...
+    def tell(self) -> int: ...
+    def readall(self) -> bytes: ...
+    def __iter__(self): ...
+    def __next__(self): ...
+    def next(self): ...
+    def read(self, size: int = ...) -> bytes: ...
+    def readinto(self, b) -> int: ...
+    def read1(self, size: int = ...) -> bytes: ...
+    def readinto1(self, b) -> int: ...
+    def seek(self, pos: int, whence: int = ...) -> int: ...
+
+class ZstdDecompressionWriter(BinaryIO):
+    def __enter__(self) -> "ZstdDecompressionWriter": ...
+    def __exit__(self, exc_type, exc_value, exc_tb): ...
+    def memory_size(self) -> int: ...
+    def close(self): ...
+    @property
+    def closed(self) -> bool: ...
+    def fileno(self) -> int: ...
+    def flush(self): ...
+    def isatty(self) -> bool: ...
+    def readable(self) -> bool: ...
+    def readline(self, size: int = ...): ...
+    def readlines(self, hint: int = ...): ...
+    def seek(self, offset: int, whence: int = ...): ...
+    def seekable(self) -> bool: ...
+    def tell(self): ...
+    def truncate(self, size: int = ...): ...
+    def writable(self) -> bool: ...
+    def writelines(self, lines: Iterable[bytes]): ...
+    def read(self, size: int = ...): ...
+    def readall(self): ...
+    def readinto(self, b): ...
+    def write(self, data: ByteString) -> int: ...
+
+class ZstdDecompressor(object):
+    def __init__(
+        self,
+        dict_data: Optional[ZstdCompressionDict] = ...,
+        max_window_size: int = ...,
+        format: int = ...,
+    ): ...
+    def memory_size(self) -> int: ...
+    def decompress(
+        self,
+        data: ByteString,
+        max_output_size: int = ...,
+        read_across_frames: bool = ...,
+        allow_extra_data: bool = ...,
+    ) -> bytes: ...
+    def stream_reader(
+        self,
+        source: Union[IO[bytes], ByteString],
+        read_size: int = ...,
+        read_across_frames: bool = ...,
+        *,
+        closefd=False,
+    ) -> ZstdDecompressionReader: ...
+    def decompressobj(
+        self, write_size: int = ..., read_across_frames: bool = False
+    ) -> ZstdDecompressionObj: ...
+    def read_to_iter(
+        self,
+        reader: Union[IO[bytes], ByteString],
+        read_size: int = ...,
+        write_size: int = ...,
+        skip_bytes: int = ...,
+    ) -> Generator[bytes, None, None]: ...
+    def stream_writer(
+        self,
+        writer: IO[bytes],
+        write_size: int = ...,
+        write_return_read: bool = ...,
+        *,
+        closefd: bool = ...,
+    ) -> ZstdDecompressionWriter: ...
+    def copy_stream(
+        self,
+        ifh: IO[bytes],
+        ofh: IO[bytes],
+        read_size: int = ...,
+        write_size: int = ...,
+    ) -> Tuple[int, int]: ...
+    def decompress_content_dict_chain(
+        self, frames: list[ByteString]
+    ) -> bytes: ...
+    def multi_decompress_to_buffer(
+        self,
+        frames: Union[
+            BufferWithSegments,
+            BufferWithSegmentsCollection,
+            List[ByteString],
+        ],
+        decompressed_sizes: ByteString = ...,
+        threads: int = ...,
+    ) -> BufferWithSegmentsCollection: ...
+
+class FrameParameters(object):
+    content_size: int
+    window_size: int
+    dict_id: int
+    has_checksum: bool
+
+def estimate_decompression_context_size() -> int: ...
+def frame_content_size(data: ByteString) -> int: ...
+def frame_header_size(data: ByteString) -> int: ...
+def get_frame_parameters(
+    data: ByteString, format: Optional[int] = None
+) -> FrameParameters: ...
+def train_dictionary(
+    dict_size: int,
+    samples: list[ByteString],
+    k: int = ...,
+    d: int = ...,
+    f: int = ...,
+    split_point: float = ...,
+    accel: int = ...,
+    notifications: int = ...,
+    dict_id: int = ...,
+    level: int = ...,
+    steps: int = ...,
+    threads: int = ...,
+) -> ZstdCompressionDict: ...
+def open(
+    filename: Union[bytes, str, os.PathLike, BinaryIO],
+    mode: str = ...,
+    cctx: Optional[ZstdCompressor] = ...,
+    dctx: Optional[ZstdDecompressor] = ...,
+    encoding: Optional[str] = ...,
+    errors: Optional[str] = ...,
+    newline: Optional[str] = ...,
+    closefd: bool = ...,
+): ...
+def compress(data: ByteString, level: int = ...) -> bytes: ...
+def decompress(data: ByteString, max_output_size: int = ...) -> bytes: ...
@@ -21,12 +21,13 @@ export const CODEX_LINK_PORT = 8795
 export const CODEX_LINK_ORIGIN = `http://127.0.0.1:${CODEX_LINK_PORT}`
 export const CODEX_LINK_STATUS_URL = `${CODEX_LINK_ORIGIN}/aiturk/codex-baglama/status`
 export const CODEX_LINK_STOP_URL = `${CODEX_LINK_ORIGIN}/aiturk/codex-baglama/stop`
+export const CODEX_LINK_REFRESH_URL = `${CODEX_LINK_ORIGIN}/aiturk/codex-baglama/refresh`

 /** The trailing comment that marks the line as ours. */
 export const CODEX_LINK_MARKER = '# AITURK IDE Codex Baglama'
 // Must equal SURUM in assets/codex-bridge/bridge.py; a running bridge that reports
 // another version is restarted so an updated IDE never keeps an old bridge alive.
-export const CODEX_BRIDGE_VERSION = '1'
+export const CODEX_BRIDGE_VERSION = '2'
 export const CODEX_LINK_LINE = `openai_base_url = "${CODEX_LINK_ORIGIN}/backend-api/codex"  ${CODEX_LINK_MARKER}`

 export const CODEX_LINK_BACKUP_SUFFIX = '.aiturk-yedek'
@@ -3,6 +3,7 @@ import { describe, expect, it } from 'vitest'
 import { CODEX_BRIDGE_START_BUDGET_MS, type CodexLinkDeps, type CodexLinkPaths, createCodexLink } from './codex-link'
 import {
  CODEX_LINK_LINE,
+  CODEX_LINK_REFRESH_URL,
  CODEX_LINK_STATUS_URL,
  CODEX_LINK_STOP_URL,
  codexBridgeArgs,
@@ -21,7 +22,7 @@ const PATHS: CodexLinkPaths = {

 const BRIDGE_JSON = `${PATHS.bridgeHome}\\bridge.json`
 const BACKUP = `${PATHS.codexConfig}.aiturk-yedek`
-const STATUS_OK = { ok: true, surum: '1', ide_durum: 'tamam', ide_model: 7, ozet: true }
+const STATUS_OK = { ok: true, surum: '2', ide_durum: 'tamam', ide_model: 7, ozet: true }
 const USER_CONFIG = 'model = "gpt-6"\r\n\r\n[features]\r\nweb = true\r\n'

 interface World {
@@ -103,6 +104,12 @@ function createWorld(patch: Partial<World> & { files?: Map<string, string> } = {
        return null
      }

+      if (method === 'POST' && url === CODEX_LINK_REFRESH_URL) {
+        world.events.push('refresh')
+
+        return world.listening ? STATUS_OK : null
+      }
+
      if (method === 'POST' && url === CODEX_LINK_STOP_URL) {
        world.events.push('stop')

@@ -273,6 +280,15 @@ describe('applyCodexLink', () => {
    expect(world.files.get(PATHS.codexConfig)).toBe(`${CODEX_LINK_LINE}\r\n${USER_CONFIG}`)
  })

+  it('asks a running bridge to re-read the IDE session on every apply', async () => {
+    const { link, world } = createWorld({ listening: true })
+
+    await link.apply({ sync: true })
+
+    expect(world.events).toContain('refresh')
+    expect(spawns(world)).toEqual([])
+  })
+
  it('never overwrites the first backup', async () => {
    const { link, world } = createWorld({ listening: true })
    world.files.set(BACKUP, 'the original original')
@@ -25,6 +25,7 @@ import {
  CODEX_BRIDGE_VERSION,
  CODEX_LINK_BACKUP_SUFFIX,
  CODEX_LINK_LAUNCHER_NAME,
+  CODEX_LINK_REFRESH_URL,
  CODEX_LINK_RUN_KEY,
  CODEX_LINK_RUN_VALUE,
  CODEX_LINK_STATUS_URL,
@@ -306,7 +307,17 @@ export function createCodexLink(deps: CodexLinkDeps): CodexLink {
      bridge = NOT_RUNNING
    }

-    if (!bridge.running) {
+    if (bridge.running) {
+      // Already up: have it re-read the IDE session and model list now (a sign-in
+      // since its last periodic refresh would otherwise show up minutes later).
+      try {
+        const refreshed = readProbe(await deps.request('POST', CODEX_LINK_REFRESH_URL, CODEX_BRIDGE_START_BUDGET_MS))
+
+        bridge = refreshed.running ? refreshed : bridge
+      } catch {
+        // The status it already gave stays valid.
+      }
+    } else {
      bridge = await startBridge(python)
    }

@@ -304,7 +304,9 @@
          "from": "assets/codex-bridge",
          "to": "codex-bridge",
          "filter": [
-            "*.py",
+            "**/*.py",
+            "**/*.pyd",
+            "**/LICENSE",
            "!test_*.py"
          ]
        }