diff --git a/venv/Lib/site-packages/sqlalchemy/util/queue.py b/venv/Lib/site-packages/sqlalchemy/util/queue.py
new file mode 100644
index 0000000..bd55357
--- /dev/null
+++ b/venv/Lib/site-packages/sqlalchemy/util/queue.py
@@ -0,0 +1,322 @@
+# util/queue.py
+# Copyright (C) 2005-2026 the SQLAlchemy authors and contributors
+# <see AUTHORS file>
+#
+# This module is part of SQLAlchemy and is released under
+# the MIT License: https://www.opensource.org/licenses/mit-license.php
+# mypy: allow-untyped-defs, allow-untyped-calls
+
+"""An adaptation of Py2.3/2.4's Queue module which supports reentrant
+behavior, using RLock instead of Lock for its mutex object.  The
+Queue object is used exclusively by the sqlalchemy.pool.QueuePool
+class.
+
+This is to support the connection pool's usage of weakref callbacks to return
+connections to the underlying Queue, which can in extremely
+rare cases be invoked within the ``get()`` method of the Queue itself,
+producing a ``put()`` inside the ``get()`` and therefore a reentrant
+condition.
+
+"""
+
+from __future__ import annotations
+
+import asyncio
+from collections import deque
+import threading
+from time import time as _time
+import typing
+from typing import Any
+from typing import Awaitable
+from typing import Deque
+from typing import Generic
+from typing import Optional
+from typing import TypeVar
+
+from .concurrency import await_fallback
+from .concurrency import await_only
+from .langhelpers import memoized_property
+
+_T = TypeVar("_T", bound=Any)
+__all__ = ["Empty", "Full", "Queue"]
+
+
+class Empty(Exception):
+    "Exception raised by Queue.get(block=0)/get_nowait()."
+
+    pass
+
+
+class Full(Exception):
+    "Exception raised by Queue.put(block=0)/put_nowait()."
+
+    pass
+
+
+class QueueCommon(Generic[_T]):
+    maxsize: int
+    use_lifo: bool
+
+    def __init__(self, maxsize: int = 0, use_lifo: bool = False): ...
+
+    def empty(self) -> bool:
+        raise NotImplementedError()
+
+    def full(self) -> bool:
+        raise NotImplementedError()
+
+    def qsize(self) -> int:
+        raise NotImplementedError()
+
+    def put_nowait(self, item: _T) -> None:
+        raise NotImplementedError()
+
+    def put(
+        self, item: _T, block: bool = True, timeout: Optional[float] = None
+    ) -> None:
+        raise NotImplementedError()
+
+    def get_nowait(self) -> _T:
+        raise NotImplementedError()
+
+    def get(self, block: bool = True, timeout: Optional[float] = None) -> _T:
+        raise NotImplementedError()
+
+
+class Queue(QueueCommon[_T]):
+    queue: Deque[_T]
+
+    def __init__(self, maxsize: int = 0, use_lifo: bool = False):
+        """Initialize a queue object with a given maximum size.
+
+        If `maxsize` is <= 0, the queue size is infinite.
+
+        If `use_lifo` is True, this Queue acts like a Stack (LIFO).
+        """
+
+        self._init(maxsize)
+        # mutex must be held whenever the queue is mutating.  All methods
+        # that acquire mutex must release it before returning.  mutex
+        # is shared between the two conditions, so acquiring and
+        # releasing the conditions also acquires and releases mutex.
+        self.mutex = threading.RLock()
+        # Notify not_empty whenever an item is added to the queue; a
+        # thread waiting to get is notified then.
+        self.not_empty = threading.Condition(self.mutex)
+        # Notify not_full whenever an item is removed from the queue;
+        # a thread waiting to put is notified then.
+        self.not_full = threading.Condition(self.mutex)
+        # If this queue uses LIFO or FIFO
+        self.use_lifo = use_lifo
+
+    def qsize(self) -> int:
+        """Return the approximate size of the queue (not reliable!)."""
+
+        with self.mutex:
+            return self._qsize()
+
+    def empty(self) -> bool:
+        """Return True if the queue is empty, False otherwise (not
+        reliable!)."""
+
+        with self.mutex:
+            return self._empty()
+
+    def full(self) -> bool:
+        """Return True if the queue is full, False otherwise (not
+        reliable!)."""
+
+        with self.mutex:
+            return self._full()
+
+    def put(
+        self, item: _T, block: bool = True, timeout: Optional[float] = None
+    ) -> None:
+        """Put an item into the queue.
+
+        If optional args `block` is True and `timeout` is None (the
+        default), block if necessary until a free slot is
+        available. If `timeout` is a positive number, it blocks at
+        most `timeout` seconds and raises the ``Full`` exception if no
+        free slot was available within that time.  Otherwise (`block`
+        is false), put an item on the queue if a free slot is
+        immediately available, else raise the ``Full`` exception
+        (`timeout` is ignored in that case).
+        """
+
+        with self.not_full:
+            if not block:
+                if self._full():
+                    raise Full
+            elif timeout is None:
+                while self._full():
+                    self.not_full.wait()
+            else:
+                if timeout < 0:
+                    raise ValueError("'timeout' must be a positive number")
+                endtime = _time() + timeout
+                while self._full():
+                    remaining = endtime - _time()
+                    if remaining <= 0.0:
+                        raise Full
+                    self.not_full.wait(remaining)
+            self._put(item)
+            self.not_empty.notify()
+
+    def put_nowait(self, item: _T) -> None:
+        """Put an item into the queue without blocking.
+
+        Only enqueue the item if a free slot is immediately available.
+        Otherwise raise the ``Full`` exception.
+        """
+        return self.put(item, False)
+
+    def get(self, block: bool = True, timeout: Optional[float] = None) -> _T:
+        """Remove and return an item from the queue.
+
+        If optional args `block` is True and `timeout` is None (the
+        default), block if necessary until an item is available. If
+        `timeout` is a positive number, it blocks at most `timeout`
+        seconds and raises the ``Empty`` exception if no item was
+        available within that time.  Otherwise (`block` is false),
+        return an item if one is immediately available, else raise the
+        ``Empty`` exception (`timeout` is ignored in that case).
+
+        """
+        with self.not_empty:
+            if not block:
+                if self._empty():
+                    raise Empty
+            elif timeout is None:
+                while self._empty():
+                    self.not_empty.wait()
+            else:
+                if timeout < 0:
+                    raise ValueError("'timeout' must be a positive number")
+                endtime = _time() + timeout
+                while self._empty():
+                    remaining = endtime - _time()
+                    if remaining <= 0.0:
+                        raise Empty
+                    self.not_empty.wait(remaining)
+            item = self._get()
+            self.not_full.notify()
+            return item
+
+    def get_nowait(self) -> _T:
+        """Remove and return an item from the queue without blocking.
+
+        Only get an item if one is immediately available. Otherwise
+        raise the ``Empty`` exception.
+        """
+
+        return self.get(False)
+
+    def _init(self, maxsize: int) -> None:
+        self.maxsize = maxsize
+        self.queue = deque()
+
+    def _qsize(self) -> int:
+        return len(self.queue)
+
+    def _empty(self) -> bool:
+        return not self.queue
+
+    def _full(self) -> bool:
+        return self.maxsize > 0 and len(self.queue) == self.maxsize
+
+    def _put(self, item: _T) -> None:
+        self.queue.append(item)
+
+    def _get(self) -> _T:
+        if self.use_lifo:
+            # LIFO
+            return self.queue.pop()
+        else:
+            # FIFO
+            return self.queue.popleft()
+
+
+class AsyncAdaptedQueue(QueueCommon[_T]):
+    if typing.TYPE_CHECKING:
+
+        @staticmethod
+        def await_(coroutine: Awaitable[Any]) -> _T: ...
+
+    else:
+        await_ = staticmethod(await_only)
+
+    def __init__(self, maxsize: int = 0, use_lifo: bool = False):
+        self.use_lifo = use_lifo
+        self.maxsize = maxsize
+
+    def empty(self) -> bool:
+        return self._queue.empty()
+
+    def full(self):
+        return self._queue.full()
+
+    def qsize(self):
+        return self._queue.qsize()
+
+    @memoized_property
+    def _queue(self) -> asyncio.Queue[_T]:
+        # Delay creation of the queue until it is first used, to avoid
+        # binding it to a possibly wrong event loop.
+        # By delaying the creation of the pool we accommodate the common
+        # usage pattern of instantiating the engine at module level, where a
+        # different event loop is in present compared to when the application
+        # is actually run.
+
+        queue: asyncio.Queue[_T]
+
+        if self.use_lifo:
+            queue = asyncio.LifoQueue(maxsize=self.maxsize)
+        else:
+            queue = asyncio.Queue(maxsize=self.maxsize)
+        return queue
+
+    def put_nowait(self, item: _T) -> None:
+        try:
+            self._queue.put_nowait(item)
+        except asyncio.QueueFull as err:
+            raise Full() from err
+
+    def put(
+        self, item: _T, block: bool = True, timeout: Optional[float] = None
+    ) -> None:
+        if not block:
+            return self.put_nowait(item)
+
+        try:
+            if timeout is not None:
+                self.await_(asyncio.wait_for(self._queue.put(item), timeout))
+            else:
+                self.await_(self._queue.put(item))
+        except (asyncio.QueueFull, asyncio.TimeoutError) as err:
+            raise Full() from err
+
+    def get_nowait(self) -> _T:
+        try:
+            return self._queue.get_nowait()
+        except asyncio.QueueEmpty as err:
+            raise Empty() from err
+
+    def get(self, block: bool = True, timeout: Optional[float] = None) -> _T:
+        if not block:
+            return self.get_nowait()
+
+        try:
+            if timeout is not None:
+                return self.await_(
+                    asyncio.wait_for(self._queue.get(), timeout)
+                )
+            else:
+                return self.await_(self._queue.get())
+        except (asyncio.QueueEmpty, asyncio.TimeoutError) as err:
+            raise Empty() from err
+
+
+class FallbackAsyncAdaptedQueue(AsyncAdaptedQueue[_T]):
+    if not typing.TYPE_CHECKING:
+        await_ = staticmethod(await_fallback)
diff --git a/venv/Lib/site-packages/sqlalchemy/util/tool_support.py b/venv/Lib/site-packages/sqlalchemy/util/tool_support.py
new file mode 100644
index 0000000..7f02f20
--- /dev/null
+++ b/venv/Lib/site-packages/sqlalchemy/util/tool_support.py
@@ -0,0 +1,202 @@
+# util/tool_support.py
+# Copyright (C) 2005-2026 the SQLAlchemy authors and contributors
+# <see AUTHORS file>
+#
+# This module is part of SQLAlchemy and is released under
+# the MIT License: https://www.opensource.org/licenses/mit-license.php
+# mypy: allow-untyped-defs, allow-untyped-calls
+"""support routines for the helpers in tools/.
+
+These aren't imported by the enclosing util package as the are not
+needed for normal library use.
+
+"""
+
+from __future__ import annotations
+
+from argparse import ArgumentParser
+from argparse import Namespace
+import contextlib
+import difflib
+import os
+from pathlib import Path
+import shlex
+import shutil
+import subprocess
+import sys
+from typing import Any
+from typing import Dict
+from typing import Iterator
+from typing import Optional
+from typing import Union
+
+from . import compat
+
+
+class code_writer_cmd:
+    parser: ArgumentParser
+    args: Namespace
+    suppress_output: bool
+    diffs_detected: bool
+    source_root: Path
+    pyproject_toml_path: Path
+
+    def __init__(self, tool_script: str):
+        self.source_root = Path(tool_script).parent.parent
+        self.pyproject_toml_path = self.source_root / Path("pyproject.toml")
+        assert self.pyproject_toml_path.exists()
+
+        self.parser = ArgumentParser()
+        self.parser.add_argument(
+            "--stdout",
+            action="store_true",
+            help="Write to stdout instead of saving to file",
+        )
+        self.parser.add_argument(
+            "-c",
+            "--check",
+            help="Don't write the files back, just return the "
+            "status. Return code 0 means nothing would change. "
+            "Return code 1 means some files would be reformatted",
+            action="store_true",
+        )
+
+    def run_zimports(self, tempfile: str) -> None:
+        self._run_console_script(
+            str(tempfile),
+            {
+                "entrypoint": "zimports",
+                "options": f"--toml-config {self.pyproject_toml_path}",
+            },
+        )
+
+    def run_black(self, tempfile: str) -> None:
+        self._run_console_script(
+            str(tempfile),
+            {
+                "entrypoint": "black",
+                "options": f"--config {self.pyproject_toml_path}",
+            },
+        )
+
+    def _run_console_script(self, path: str, options: Dict[str, Any]) -> None:
+        """Run a Python console application from within the process.
+
+        Used for black, zimports
+
+        """
+
+        is_posix = os.name == "posix"
+
+        entrypoint_name = options["entrypoint"]
+
+        for entry in compat.importlib_metadata_get("console_scripts"):
+            if entry.name == entrypoint_name:
+                impl = entry
+                break
+        else:
+            raise Exception(
+                f"Could not find entrypoint console_scripts.{entrypoint_name}"
+            )
+        cmdline_options_str = options.get("options", "")
+        cmdline_options_list = shlex.split(
+            cmdline_options_str, posix=is_posix
+        ) + [path]
+
+        kw: Dict[str, Any] = {}
+        if self.suppress_output:
+            kw["stdout"] = kw["stderr"] = subprocess.DEVNULL
+
+        subprocess.run(
+            [
+                sys.executable,
+                "-c",
+                "import %s; %s.%s()" % (impl.module, impl.module, impl.attr),
+            ]
+            + cmdline_options_list,
+            cwd=str(self.source_root),
+            **kw,
+        )
+
+    def write_status(self, *text: str) -> None:
+        if not self.suppress_output:
+            sys.stderr.write(" ".join(text))
+
+    def write_output_file_from_text(
+        self, text: str, destination_path: Union[str, Path]
+    ) -> None:
+        if self.args.check:
+            self._run_diff(destination_path, source=text)
+        elif self.args.stdout:
+            print(text)
+        else:
+            self.write_status(f"Writing {destination_path}...")
+            Path(destination_path).write_text(
+                text, encoding="utf-8", newline="\n"
+            )
+            self.write_status("done\n")
+
+    def write_output_file_from_tempfile(
+        self, tempfile: str, destination_path: str
+    ) -> None:
+        if self.args.check:
+            self._run_diff(destination_path, source_file=tempfile)
+            os.unlink(tempfile)
+        elif self.args.stdout:
+            with open(tempfile) as tf:
+                print(tf.read())
+            os.unlink(tempfile)
+        else:
+            self.write_status(f"Writing {destination_path}...")
+            shutil.move(tempfile, destination_path)
+            self.write_status("done\n")
+
+    def _run_diff(
+        self,
+        destination_path: Union[str, Path],
+        *,
+        source: Optional[str] = None,
+        source_file: Optional[str] = None,
+    ) -> None:
+        if source_file:
+            with open(source_file, encoding="utf-8") as tf:
+                source_lines = list(tf)
+        elif source is not None:
+            source_lines = source.splitlines(keepends=True)
+        else:
+            assert False, "source or source_file is required"
+
+        with open(destination_path, encoding="utf-8") as dp:
+            d = difflib.unified_diff(
+                list(dp),
+                source_lines,
+                fromfile=Path(destination_path).as_posix(),
+                tofile="<proposed changes>",
+                n=3,
+                lineterm="\n",
+            )
+            d_as_list = list(d)
+            if d_as_list:
+                self.diffs_detected = True
+                print("".join(d_as_list))
+
+    @contextlib.contextmanager
+    def add_arguments(self) -> Iterator[ArgumentParser]:
+        yield self.parser
+
+    @contextlib.contextmanager
+    def run_program(self) -> Iterator[None]:
+        self.args = self.parser.parse_args()
+        if self.args.check:
+            self.diffs_detected = False
+            self.suppress_output = True
+        elif self.args.stdout:
+            self.suppress_output = True
+        else:
+            self.suppress_output = False
+        yield
+
+        if self.args.check and self.diffs_detected:
+            sys.exit(1)
+        else:
+            sys.exit(0)
diff --git a/venv/Lib/site-packages/sqlalchemy/util/topological.py b/venv/Lib/site-packages/sqlalchemy/util/topological.py
new file mode 100644
index 0000000..b051ca7
--- /dev/null
+++ b/venv/Lib/site-packages/sqlalchemy/util/topological.py
@@ -0,0 +1,120 @@
+# util/topological.py
+# Copyright (C) 2005-2026 the SQLAlchemy authors and contributors
+# <see AUTHORS file>
+#
+# This module is part of SQLAlchemy and is released under
+# the MIT License: https://www.opensource.org/licenses/mit-license.php
+
+"""Topological sorting algorithms."""
+
+from __future__ import annotations
+
+from typing import Any
+from typing import Collection
+from typing import DefaultDict
+from typing import Iterable
+from typing import Iterator
+from typing import Sequence
+from typing import Set
+from typing import Tuple
+from typing import TypeVar
+
+from .. import util
+from ..exc import CircularDependencyError
+
+_T = TypeVar("_T", bound=Any)
+
+__all__ = ["sort", "sort_as_subsets", "find_cycles"]
+
+
+def sort_as_subsets(
+    tuples: Collection[Tuple[_T, _T]], allitems: Collection[_T]
+) -> Iterator[Sequence[_T]]:
+    edges: DefaultDict[_T, Set[_T]] = util.defaultdict(set)
+    for parent, child in tuples:
+        edges[child].add(parent)
+
+    todo = list(allitems)
+    todo_set = set(allitems)
+
+    while todo_set:
+        output = []
+        for node in todo:
+            if todo_set.isdisjoint(edges[node]):
+                output.append(node)
+
+        if not output:
+            raise CircularDependencyError(
+                "Circular dependency detected.",
+                find_cycles(tuples, allitems),
+                _gen_edges(edges),
+            )
+
+        todo_set.difference_update(output)
+        todo = [t for t in todo if t in todo_set]
+        yield output
+
+
+def sort(
+    tuples: Collection[Tuple[_T, _T]],
+    allitems: Collection[_T],
+    deterministic_order: bool = True,
+) -> Iterator[_T]:
+    """sort the given list of items by dependency.
+
+    'tuples' is a list of tuples representing a partial ordering.
+
+    deterministic_order is no longer used, the order is now always
+    deterministic given the order of "allitems".    the flag is there
+    for backwards compatibility with Alembic.
+
+    """
+
+    for set_ in sort_as_subsets(tuples, allitems):
+        yield from set_
+
+
+def find_cycles(
+    tuples: Iterable[Tuple[_T, _T]], allitems: Iterable[_T]
+) -> Set[_T]:
+    # adapted from:
+    # https://neopythonic.blogspot.com/2009/01/detecting-cycles-in-directed-graph.html
+
+    edges: DefaultDict[_T, Set[_T]] = util.defaultdict(set)
+    for parent, child in tuples:
+        edges[parent].add(child)
+    nodes_to_test = set(edges)
+
+    output = set()
+
+    # we'd like to find all nodes that are
+    # involved in cycles, so we do the full
+    # pass through the whole thing for each
+    # node in the original list.
+
+    # we can go just through parent edge nodes.
+    # if a node is only a child and never a parent,
+    # by definition it can't be part of a cycle.  same
+    # if it's not in the edges at all.
+    for node in nodes_to_test:
+        stack = [node]
+        todo = nodes_to_test.difference(stack)
+        while stack:
+            top = stack[-1]
+            for node in edges[top]:
+                if node in stack:
+                    cyc = stack[stack.index(node) :]
+                    todo.difference_update(cyc)
+                    output.update(cyc)
+
+                if node in todo:
+                    stack.append(node)
+                    todo.remove(node)
+                    break
+            else:
+                stack.pop()
+    return output
+
+
+def _gen_edges(edges: DefaultDict[_T, Set[_T]]) -> Set[Tuple[_T, _T]]:
+    return {(right, left) for left in edges for right in edges[left]}
diff --git a/venv/Lib/site-packages/sqlalchemy/util/typing.py b/venv/Lib/site-packages/sqlalchemy/util/typing.py
new file mode 100644
index 0000000..2ee78d6
--- /dev/null
+++ b/venv/Lib/site-packages/sqlalchemy/util/typing.py
@@ -0,0 +1,734 @@
+# util/typing.py
+# Copyright (C) 2022-2026 the SQLAlchemy authors and contributors
+# <see AUTHORS file>
+#
+# This module is part of SQLAlchemy and is released under
+# the MIT License: https://www.opensource.org/licenses/mit-license.php
+# mypy: allow-untyped-defs, allow-untyped-calls
+
+from __future__ import annotations
+
+import builtins
+from collections import deque
+import collections.abc as collections_abc
+import re
+import sys
+import typing
+from typing import Any
+from typing import Callable
+from typing import Dict
+from typing import ForwardRef
+from typing import Generic
+from typing import Iterable
+from typing import Mapping
+from typing import NewType
+from typing import NoReturn
+from typing import Optional
+from typing import overload
+from typing import Set
+from typing import Tuple
+from typing import Type
+from typing import TYPE_CHECKING
+from typing import TypeVar
+from typing import Union
+
+import typing_extensions
+
+from . import compat
+
+if True:  # zimports removes the tailing comments
+    from typing_extensions import Annotated as Annotated  # 3.8
+    from typing_extensions import Concatenate as Concatenate  # 3.10
+    from typing_extensions import (
+        dataclass_transform as dataclass_transform,  # 3.11,
+    )
+    from typing_extensions import Final as Final  # 3.8
+    from typing_extensions import final as final  # 3.8
+    from typing_extensions import get_args as get_args  # 3.10
+    from typing_extensions import get_origin as get_origin  # 3.10
+    from typing_extensions import Literal as Literal  # 3.8
+    from typing_extensions import NotRequired as NotRequired  # 3.11
+    from typing_extensions import ParamSpec as ParamSpec  # 3.10
+    from typing_extensions import Protocol as Protocol  # 3.8
+    from typing_extensions import SupportsIndex as SupportsIndex  # 3.8
+    from typing_extensions import TypeAlias as TypeAlias  # 3.10
+    from typing_extensions import TypedDict as TypedDict  # 3.8
+    from typing_extensions import TypeGuard as TypeGuard  # 3.10
+    from typing_extensions import Self as Self  # 3.11
+    from typing_extensions import TypeAliasType as TypeAliasType  # 3.12
+    from typing_extensions import Never as Never  # 3.11
+    from typing_extensions import LiteralString as LiteralString  # 3.11
+
+_T = TypeVar("_T", bound=Any)
+_KT = TypeVar("_KT")
+_KT_co = TypeVar("_KT_co", covariant=True)
+_KT_contra = TypeVar("_KT_contra", contravariant=True)
+_VT = TypeVar("_VT")
+_VT_co = TypeVar("_VT_co", covariant=True)
+
+if compat.py310:
+    # why they took until py310 to put this in stdlib is beyond me,
+    # I've been wanting it since py27
+    from types import NoneType as NoneType
+else:
+    NoneType = type(None)  # type: ignore
+
+
+def is_fwd_none(typ: Any) -> bool:
+    return isinstance(typ, ForwardRef) and typ.__forward_arg__ == "None"
+
+
+_AnnotationScanType = Union[
+    Type[Any], str, ForwardRef, NewType, TypeAliasType, "GenericProtocol[Any]"
+]
+
+
+class ArgsTypeProtocol(Protocol):
+    """protocol for types that have ``__args__``
+
+    there's no public interface for this AFAIK
+
+    """
+
+    __args__: Tuple[_AnnotationScanType, ...]
+
+
+class GenericProtocol(Protocol[_T]):
+    """protocol for generic types.
+
+    this since Python.typing _GenericAlias is private
+
+    """
+
+    __args__: Tuple[_AnnotationScanType, ...]
+    __origin__: Type[_T]
+
+    # Python's builtin _GenericAlias has this method, however builtins like
+    # list, dict, etc. do not, even though they have ``__origin__`` and
+    # ``__args__``
+    #
+    # def copy_with(self, params: Tuple[_AnnotationScanType, ...]) -> Type[_T]:
+    #     ...
+
+
+# copied from TypeShed, required in order to implement
+# MutableMapping.update()
+class SupportsKeysAndGetItem(Protocol[_KT, _VT_co]):
+    def keys(self) -> Iterable[_KT]: ...
+
+    def __getitem__(self, __k: _KT) -> _VT_co: ...
+
+
+# work around https://github.com/microsoft/pyright/issues/3025
+_LiteralStar = Literal["*"]
+
+
+def de_stringify_annotation(
+    cls: Type[Any],
+    annotation: _AnnotationScanType,
+    originating_module: str,
+    locals_: Mapping[str, Any],
+    *,
+    str_cleanup_fn: Optional[Callable[[str, str], str]] = None,
+    include_generic: bool = False,
+    _already_seen: Optional[Set[Any]] = None,
+) -> Type[Any]:
+    """Resolve annotations that may be string based into real objects.
+
+    This is particularly important if a module defines "from __future__ import
+    annotations", as everything inside of __annotations__ is a string. We want
+    to at least have generic containers like ``Mapped``, ``Union``, ``List``,
+    etc.
+
+    """
+    # looked at typing.get_type_hints(), looked at pydantic.  We need much
+    # less here, and we here try to not use any private typing internals
+    # or construct ForwardRef objects which is documented as something
+    # that should be avoided.
+
+    original_annotation = annotation
+
+    if is_fwd_ref(annotation):
+        annotation = annotation.__forward_arg__
+
+    if isinstance(annotation, str):
+        if str_cleanup_fn:
+            annotation = str_cleanup_fn(annotation, originating_module)
+
+        annotation = eval_expression(
+            annotation, originating_module, locals_=locals_, in_class=cls
+        )
+
+    if (
+        include_generic
+        and is_generic(annotation)
+        and not is_literal(annotation)
+    ):
+        if _already_seen is None:
+            _already_seen = set()
+
+        if annotation in _already_seen:
+            # only occurs recursively.  outermost return type
+            # will always be Type.
+            # the element here will be either ForwardRef or
+            # Optional[ForwardRef]
+            return original_annotation  # type: ignore
+        else:
+            _already_seen.add(annotation)
+
+        elements = tuple(
+            de_stringify_annotation(
+                cls,
+                elem,
+                originating_module,
+                locals_,
+                str_cleanup_fn=str_cleanup_fn,
+                include_generic=include_generic,
+                _already_seen=_already_seen,
+            )
+            for elem in annotation.__args__
+        )
+
+        return _copy_generic_annotation_with(annotation, elements)
+
+    return annotation  # type: ignore
+
+
+def fixup_container_fwd_refs(
+    type_: _AnnotationScanType,
+) -> _AnnotationScanType:
+    """Correct dict['x', 'y'] into dict[ForwardRef('x'), ForwardRef('y')]
+    and similar for list, set
+
+    """
+
+    if (
+        is_generic(type_)
+        and get_origin(type_)
+        in (
+            dict,
+            set,
+            list,
+            collections_abc.MutableSet,
+            collections_abc.MutableMapping,
+            collections_abc.MutableSequence,
+            collections_abc.Mapping,
+            collections_abc.Sequence,
+        )
+        # fight, kick and scream to struggle to tell the difference between
+        # dict[] and typing.Dict[] which DO NOT compare the same and DO NOT
+        # behave the same yet there is NO WAY to distinguish between which type
+        # it is using public attributes
+        and not re.match(
+            "typing.(?:Dict|List|Set|.*Mapping|.*Sequence|.*Set)", repr(type_)
+        )
+    ):
+        # compat with py3.10 and earlier
+        return get_origin(type_).__class_getitem__(  # type: ignore
+            tuple(
+                [
+                    ForwardRef(elem) if isinstance(elem, str) else elem
+                    for elem in get_args(type_)
+                ]
+            )
+        )
+    return type_
+
+
+def _copy_generic_annotation_with(
+    annotation: GenericProtocol[_T], elements: Tuple[_AnnotationScanType, ...]
+) -> Type[_T]:
+    if hasattr(annotation, "copy_with"):
+        # List, Dict, etc. real generics
+        return annotation.copy_with(elements)  # type: ignore
+    else:
+        # Python builtins list, dict, etc.
+        return annotation.__origin__[elements]  # type: ignore
+
+
+def eval_expression(
+    expression: str,
+    module_name: str,
+    *,
+    locals_: Optional[Mapping[str, Any]] = None,
+    in_class: Optional[Type[Any]] = None,
+) -> Any:
+    try:
+        base_globals: Dict[str, Any] = sys.modules[module_name].__dict__
+    except KeyError as ke:
+        raise NameError(
+            f"Module {module_name} isn't present in sys.modules; can't "
+            f"evaluate expression {expression}"
+        ) from ke
+
+    try:
+        if in_class is not None:
+            cls_namespace = dict(in_class.__dict__)
+            cls_namespace.setdefault(in_class.__name__, in_class)
+
+            # see #10899.  We want the locals/globals to take precedence
+            # over the class namespace in this context, even though this
+            # is not the usual way variables would resolve.
+            cls_namespace.update(base_globals)
+
+            annotation = eval(expression, cls_namespace, locals_)
+        else:
+            annotation = eval(expression, base_globals, locals_)
+    except Exception as err:
+        raise NameError(
+            f"Could not de-stringify annotation {expression!r}"
+        ) from err
+    else:
+        return annotation
+
+
+def eval_name_only(
+    name: str,
+    module_name: str,
+    *,
+    locals_: Optional[Mapping[str, Any]] = None,
+) -> Any:
+    if "." in name:
+        return eval_expression(name, module_name, locals_=locals_)
+
+    try:
+        base_globals: Dict[str, Any] = sys.modules[module_name].__dict__
+    except KeyError as ke:
+        raise NameError(
+            f"Module {module_name} isn't present in sys.modules; can't "
+            f"resolve name {name}"
+        ) from ke
+
+    # name only, just look in globals.  eval() works perfectly fine here,
+    # however we are seeking to have this be faster, as this occurs for
+    # every Mapper[] keyword, etc. depending on configuration
+    try:
+        return base_globals[name]
+    except KeyError as ke:
+        # check in builtins as well to handle `list`, `set` or `dict`, etc.
+        try:
+            return builtins.__dict__[name]
+        except KeyError:
+            pass
+
+        raise NameError(
+            f"Could not locate name {name} in module {module_name}"
+        ) from ke
+
+
+def resolve_name_to_real_class_name(name: str, module_name: str) -> str:
+    try:
+        obj = eval_name_only(name, module_name)
+    except NameError:
+        return name
+    else:
+        return getattr(obj, "__name__", name)
+
+
+def is_pep593(type_: Optional[Any]) -> bool:
+    return type_ is not None and get_origin(type_) in _type_tuples.Annotated
+
+
+def is_non_string_iterable(obj: Any) -> TypeGuard[Iterable[Any]]:
+    return isinstance(obj, collections_abc.Iterable) and not isinstance(
+        obj, (str, bytes)
+    )
+
+
+def is_literal(type_: Any) -> bool:
+    return get_origin(type_) in _type_tuples.Literal
+
+
+def is_newtype(type_: Optional[_AnnotationScanType]) -> TypeGuard[NewType]:
+    return hasattr(type_, "__supertype__")
+
+    # doesn't work in 3.8, 3.7 as it passes a closure, not an
+    # object instance
+    # isinstance(type, type_instances.NewType)
+
+
+def is_generic(type_: _AnnotationScanType) -> TypeGuard[GenericProtocol[Any]]:
+    return hasattr(type_, "__args__") and hasattr(type_, "__origin__")
+
+
+def is_pep695(type_: _AnnotationScanType) -> TypeGuard[TypeAliasType]:
+    # NOTE: a generic TAT does not instance check as TypeAliasType outside of
+    # python 3.10. For sqlalchemy use cases it's fine to consider it a TAT
+    # though.
+    # NOTE: things seems to work also without this additional check
+    if is_generic(type_):
+        return is_pep695(type_.__origin__)
+    return isinstance(type_, _type_instances.TypeAliasType)
+
+
+def flatten_newtype(type_: NewType) -> Type[Any]:
+    super_type = type_.__supertype__
+    while is_newtype(super_type):
+        super_type = super_type.__supertype__
+    return super_type  # type: ignore[return-value]
+
+
+def pep695_values(type_: _AnnotationScanType) -> Set[Any]:
+    """Extracts the value from a TypeAliasType, recursively exploring unions
+    and inner TypeAliasType to flatten them into a single set.
+
+    Forward references are not evaluated, so no recursive exploration happens
+    into them.
+    """
+    _seen = set()
+
+    def recursive_value(inner_type):
+        if inner_type in _seen:
+            # recursion are not supported (at least it's flagged as
+            # an error by pyright). Just avoid infinite loop
+            return inner_type
+        _seen.add(inner_type)
+        if not is_pep695(inner_type):
+            return inner_type
+        value = inner_type.__value__
+        if not is_union(value):
+            return value
+        return [recursive_value(t) for t in value.__args__]
+
+    res = recursive_value(type_)
+    if isinstance(res, list):
+        types = set()
+        stack = deque(res)
+        while stack:
+            t = stack.popleft()
+            if isinstance(t, list):
+                stack.extend(t)
+            else:
+                types.add(None if t is NoneType or is_fwd_none(t) else t)
+        return types
+    else:
+        return {res}
+
+
+def is_fwd_ref(
+    type_: _AnnotationScanType,
+    check_generic: bool = False,
+    check_for_plain_string: bool = False,
+) -> TypeGuard[ForwardRef]:
+    if check_for_plain_string and isinstance(type_, str):
+        return True
+    elif isinstance(type_, _type_instances.ForwardRef):
+        return True
+    elif check_generic and is_generic(type_):
+        return any(
+            is_fwd_ref(
+                arg, True, check_for_plain_string=check_for_plain_string
+            )
+            for arg in type_.__args__
+        )
+    else:
+        return False
+
+
+@overload
+def de_optionalize_union_types(type_: str) -> str: ...
+
+
+@overload
+def de_optionalize_union_types(type_: Type[Any]) -> Type[Any]: ...
+
+
+@overload
+def de_optionalize_union_types(
+    type_: _AnnotationScanType,
+) -> _AnnotationScanType: ...
+
+
+def de_optionalize_union_types(
+    type_: _AnnotationScanType,
+) -> _AnnotationScanType:
+    """Given a type, filter out ``Union`` types that include ``NoneType``
+    to not include the ``NoneType``.
+
+    Contains extra logic to work on non-flattened unions, unions that contain
+    ``None`` (seen in py38, 37)
+
+    """
+
+    if is_fwd_ref(type_):
+        return _de_optionalize_fwd_ref_union_types(type_, False)
+
+    elif is_union(type_) and includes_none(type_):
+        if compat.py39:
+            typ = set(type_.__args__)
+        else:
+            # py38, 37 - unions are not automatically flattened, can contain
+            # None rather than NoneType
+            stack_of_unions = deque([type_])
+            typ = set()
+            while stack_of_unions:
+                u_typ = stack_of_unions.popleft()
+                for elem in u_typ.__args__:
+                    if is_union(elem):
+                        stack_of_unions.append(elem)
+                    else:
+                        typ.add(elem)
+
+            typ.discard(None)  # type: ignore[arg-type,unused-ignore]
+
+        typ = {t for t in typ if t is not NoneType and not is_fwd_none(t)}
+
+        return make_union_type(*typ)
+
+    else:
+        return type_
+
+
+@overload
+def _de_optionalize_fwd_ref_union_types(
+    type_: ForwardRef, return_has_none: Literal[True]
+) -> bool: ...
+
+
+@overload
+def _de_optionalize_fwd_ref_union_types(
+    type_: ForwardRef, return_has_none: Literal[False]
+) -> _AnnotationScanType: ...
+
+
+def _de_optionalize_fwd_ref_union_types(
+    type_: ForwardRef, return_has_none: bool
+) -> Union[_AnnotationScanType, bool]:
+    """return the non-optional type for Optional[], Union[None, ...], x|None,
+    etc. without de-stringifying forward refs.
+
+    unfortunately this seems to require lots of hardcoded heuristics
+
+    """
+
+    annotation = type_.__forward_arg__
+
+    mm = re.match(r"^(.+?)\[(.+)\]$", annotation)
+    if mm:
+        g1 = mm.group(1).split(".")[-1]
+        if g1 == "Optional":
+            return True if return_has_none else ForwardRef(mm.group(2))
+        elif g1 == "Union":
+            if "[" in mm.group(2):
+                # cases like "Union[Dict[str, int], int, None]"
+                elements: list[str] = []
+                current: list[str] = []
+                ignore_comma = 0
+                for char in mm.group(2):
+                    if char == "[":
+                        ignore_comma += 1
+                    elif char == "]":
+                        ignore_comma -= 1
+                    elif ignore_comma == 0 and char == ",":
+                        elements.append("".join(current).strip())
+                        current.clear()
+                        continue
+                    current.append(char)
+            else:
+                elements = re.split(r",\s*", mm.group(2))
+            parts = [ForwardRef(elem) for elem in elements if elem != "None"]
+            if return_has_none:
+                return len(elements) != len(parts)
+            else:
+                return make_union_type(*parts) if parts else Never  # type: ignore[return-value] # noqa: E501
+        else:
+            return False if return_has_none else type_
+
+    pipe_tokens = re.split(r"\s*\|\s*", annotation)
+    has_none = "None" in pipe_tokens
+    if return_has_none:
+        return has_none
+    if has_none:
+        anno_str = "|".join(p for p in pipe_tokens if p != "None")
+        return ForwardRef(anno_str) if anno_str else Never  # type: ignore[return-value] # noqa: E501
+
+    return type_
+
+
+def make_union_type(*types: _AnnotationScanType) -> Type[Any]:
+    """Make a Union type."""
+
+    return Union[types]  # type: ignore
+
+
+def includes_none(type_: Any) -> bool:
+    """Returns if the type annotation ``type_`` allows ``None``.
+
+    This function supports:
+    * forward refs
+    * unions
+    * pep593 - Annotated
+    * pep695 - TypeAliasType (does not support looking into
+    fw reference of other pep695)
+    * NewType
+    * plain types like ``int``, ``None``, etc
+    """
+    if is_fwd_ref(type_):
+        return _de_optionalize_fwd_ref_union_types(type_, True)
+    if is_union(type_):
+        return any(includes_none(t) for t in get_args(type_))
+    if is_pep593(type_):
+        return includes_none(get_args(type_)[0])
+    if is_pep695(type_):
+        return any(includes_none(t) for t in pep695_values(type_))
+    if is_newtype(type_):
+        return includes_none(type_.__supertype__)
+    try:
+        return type_ in (NoneType, None) or is_fwd_none(type_)
+    except TypeError:
+        # if type_ is Column, mapped_column(), etc. the use of "in"
+        # resolves to ``__eq__()`` which then gives us an expression object
+        # that can't resolve to boolean.  just catch it all via exception
+        return False
+
+
+def is_a_type(type_: Any) -> bool:
+    return (
+        isinstance(type_, type)
+        or get_origin(type_) is not None
+        or getattr(type_, "__module__", None)
+        in ("typing", "typing_extensions")
+        or type(type_).__mro__[0].__module__ in ("typing", "typing_extensions")
+    )
+
+
+def is_union(type_: Any) -> TypeGuard[ArgsTypeProtocol]:
+    return is_origin_of(type_, "Union", "UnionType")
+
+
+def is_origin_of_cls(
+    type_: Any, class_obj: Union[Tuple[Type[Any], ...], Type[Any]]
+) -> bool:
+    """return True if the given type has an __origin__ that shares a base
+    with the given class"""
+
+    origin = get_origin(type_)
+    if origin is None:
+        return False
+
+    return isinstance(origin, type) and issubclass(origin, class_obj)
+
+
+def is_origin_of(
+    type_: Any, *names: str, module: Optional[str] = None
+) -> bool:
+    """return True if the given type has an __origin__ with the given name
+    and optional module."""
+
+    origin = get_origin(type_)
+    if origin is None:
+        return False
+
+    return _get_type_name(origin) in names and (
+        module is None or origin.__module__.startswith(module)
+    )
+
+
+def _get_type_name(type_: Type[Any]) -> str:
+    if compat.py310:
+        return type_.__name__
+    else:
+        typ_name = getattr(type_, "__name__", None)
+        if typ_name is None:
+            typ_name = getattr(type_, "_name", None)
+
+        return typ_name  # type: ignore
+
+
+class DescriptorProto(Protocol):
+    def __get__(self, instance: object, owner: Any) -> Any: ...
+
+    def __set__(self, instance: Any, value: Any) -> None: ...
+
+    def __delete__(self, instance: Any) -> None: ...
+
+
+_DESC = TypeVar("_DESC", bound=DescriptorProto)
+
+
+class DescriptorReference(Generic[_DESC]):
+    """a descriptor that refers to a descriptor.
+
+    used for cases where we need to have an instance variable referring to an
+    object that is itself a descriptor, which typically confuses typing tools
+    as they don't know when they should use ``__get__`` or not when referring
+    to the descriptor assignment as an instance variable. See
+    sqlalchemy.orm.interfaces.PropComparator.prop
+
+    """
+
+    if TYPE_CHECKING:
+
+        def __get__(self, instance: object, owner: Any) -> _DESC: ...
+
+        def __set__(self, instance: Any, value: _DESC) -> None: ...
+
+        def __delete__(self, instance: Any) -> None: ...
+
+
+_DESC_co = TypeVar("_DESC_co", bound=DescriptorProto, covariant=True)
+
+
+class RODescriptorReference(Generic[_DESC_co]):
+    """a descriptor that refers to a descriptor.
+
+    same as :class:`.DescriptorReference` but is read-only, so that subclasses
+    can define a subtype as the generically contained element
+
+    """
+
+    if TYPE_CHECKING:
+
+        def __get__(self, instance: object, owner: Any) -> _DESC_co: ...
+
+        def __set__(self, instance: Any, value: Any) -> NoReturn: ...
+
+        def __delete__(self, instance: Any) -> NoReturn: ...
+
+
+_FN = TypeVar("_FN", bound=Optional[Callable[..., Any]])
+
+
+class CallableReference(Generic[_FN]):
+    """a descriptor that refers to a callable.
+
+    works around mypy's limitation of not allowing callables assigned
+    as instance variables
+
+
+    """
+
+    if TYPE_CHECKING:
+
+        def __get__(self, instance: object, owner: Any) -> _FN: ...
+
+        def __set__(self, instance: Any, value: _FN) -> None: ...
+
+        def __delete__(self, instance: Any) -> None: ...
+
+
+class _TypingInstances:
+    def __getattr__(self, key: str) -> tuple[type, ...]:
+        types = tuple(
+            {
+                t
+                for t in [
+                    getattr(typing, key, None),
+                    getattr(typing_extensions, key, None),
+                ]
+                if t is not None
+            }
+        )
+        if not types:
+            raise AttributeError(key)
+        self.__dict__[key] = types
+        return types
+
+
+_type_tuples = _TypingInstances()
+if TYPE_CHECKING:
+    _type_instances = typing_extensions
+else:
+    _type_instances = _type_tuples
+
+LITERAL_TYPES = _type_tuples.Literal