_base_client.py

from __future__ import annotations

import sys
import json
import math
import time
import uuid
import email
import asyncio
import inspect
import logging
import platform
import warnings
import email.utils
from types import TracebackType
from random import random
from typing import (
    TYPE_CHECKING,
    Any,
    Dict,
    Type,
    Union,
    Generic,
    Mapping,
    TypeVar,
    Iterable,
    Iterator,
    Optional,
    Generator,
    AsyncIterator,
    cast,
    overload,
)
from typing_extensions import Unpack, Literal, override, get_origin

import anyio
import httpx2
import pydantic
from httpx2 import URL
from pydantic import PrivateAttr

from . import _exceptions
from ._qs import Querystring
from ._files import to_httpx_files, async_to_httpx_files
from ._types import (
    Body,
    Omit,
    Query,
    Headers,
    Timeout,
    NotGiven,
    ResponseT,
    AnyMapping,
    PostParser,
    BinaryTypes,
    RequestFiles,
    HttpxSendArgs,
    RequestOptions,
    AsyncBinaryTypes,
    HttpxRequestFiles,
    ModelBuilderProtocol,
    not_given,
)
from ._utils import SensitiveHeadersFilter, is_dict, is_list, asyncify, is_given, lru_cache, is_mapping
from ._compat import PYDANTIC_V1, model_copy
from ._httpx2 import (
    status_exceptions,
    request_exceptions,
    timeout_exceptions,
    http_response_types,
    normalize_httpx_url,
    is_httpx2_sync_client,
    normalize_httpx2_auth,
    is_httpx2_async_client,
    normalize_httpx_timeout,
    normalize_httpx2_timeout,
    is_legacy_httpx_sync_client,
    normalize_legacy_httpx_auth,
    is_legacy_httpx_async_client,
    normalize_legacy_httpx_timeout,
)
from ._models import GenericModel, SecurityOptions, FinalRequestOptions, validate_type, construct_type
from ._response import (
    APIResponse,
    BaseAPIResponse,
    AsyncAPIResponse,
    extract_response_type,
)
from ._constants import (
    DEFAULT_TIMEOUT,
    MAX_RETRY_DELAY,
    DEFAULT_MAX_RETRIES,
    INITIAL_RETRY_DELAY,
    RAW_RESPONSE_HEADER,
    MAX_RETRY_AFTER_DELAY,
    OVERRIDE_CAST_TO_HEADER,
    DEFAULT_CONNECTION_LIMITS,
)
from ._streaming import Stream, SSEDecoder, AsyncStream, SSEBytesDecoder
from ._exceptions import (
    OpenAIError,
    APIStatusError,
    APITimeoutError,
    APIConnectionError,
    APIResponseValidationError,
)
from ._utils._json import openapi_dumps
from ._utils._logs import get_http_method_for_logging
from ._legacy_response import LegacyAPIResponse

log: logging.Logger = logging.getLogger(__name__)
log.addFilter(SensitiveHeadersFilter())

# TODO: make base page type vars covariant
SyncPageT = TypeVar("SyncPageT", bound="BaseSyncPage[Any]")
AsyncPageT = TypeVar("AsyncPageT", bound="BaseAsyncPage[Any]")


_T = TypeVar("_T")
_T_co = TypeVar("_T_co", covariant=True)

_StreamT = TypeVar("_StreamT", bound=Stream[Any])
_AsyncStreamT = TypeVar("_AsyncStreamT", bound=AsyncStream[Any])

if TYPE_CHECKING:
    from httpx2._config import (
        DEFAULT_TIMEOUT_CONFIG,  # pyright: ignore[reportPrivateImportUsage]
    )

    HTTPX_DEFAULT_TIMEOUT = DEFAULT_TIMEOUT_CONFIG
else:
    try:
        from httpx2._config import DEFAULT_TIMEOUT_CONFIG as HTTPX_DEFAULT_TIMEOUT
    except ImportError:
        # taken from https://github.com/encode/httpx/blob/3ba5fe0d7ac70222590e759c31442b1cab263791/httpx/_config.py#L366
        HTTPX_DEFAULT_TIMEOUT = Timeout(5.0)


class PageInfo:
    """Stores the necessary information to build the request to retrieve the next page.

    Either `url` or `params` must be set.
    """

    url: URL | NotGiven
    params: Query | NotGiven
    json: Body | NotGiven

    @overload
    def __init__(
        self,
        *,
        url: URL,
    ) -> None: ...

    @overload
    def __init__(
        self,
        *,
        params: Query,
    ) -> None: ...

    @overload
    def __init__(
        self,
        *,
        json: Body,
    ) -> None: ...

    def __init__(
        self,
        *,
        url: URL | NotGiven = not_given,
        json: Body | NotGiven = not_given,
        params: Query | NotGiven = not_given,
    ) -> None:
        self.url = url
        self.json = json
        self.params = params

    @override
    def __repr__(self) -> str:
        if self.url:
            return f"{self.__class__.__name__}(url={self.url})"
        if self.json:
            return f"{self.__class__.__name__}(json={self.json})"
        return f"{self.__class__.__name__}(params={self.params})"


class BasePage(GenericModel, Generic[_T]):
    """
    Defines the core interface for pagination.

    Type Args:
        ModelT: The pydantic model that represents an item in the response.

    Methods:
        has_next_page(): Check if there is another page available
        next_page_info(): Get the necessary information to make a request for the next page
    """

    _options: FinalRequestOptions = PrivateAttr()
    _model: Type[_T] = PrivateAttr()

    def has_next_page(self) -> bool:
        items = self._get_page_items()
        if not items:
            return False
        return self.next_page_info() is not None

    def next_page_info(self) -> Optional[PageInfo]: ...

    def _get_page_items(self) -> Iterable[_T]:  # type: ignore[empty-body]
        ...

    def _params_from_url(self, url: URL) -> httpx2.QueryParams:
        # TODO: do we have to preprocess params here?
        return httpx2.QueryParams(cast(Any, self._options.params)).merge(url.params)

    def _info_to_options(self, info: PageInfo) -> FinalRequestOptions:
        options = model_copy(self._options)
        options._strip_raw_response_header()

        if not isinstance(info.params, NotGiven):
            options.params = {**options.params, **info.params}
            return options

        if not isinstance(info.url, NotGiven):
            params = self._params_from_url(info.url)
            url = info.url.copy_with(params=params)
            options.params = dict(url.params)
            options.url = str(url)
            return options

        if not isinstance(info.json, NotGiven):
            if not is_mapping(info.json):
                raise TypeError("Pagination is only supported with mappings")

            if not options.json_data:
                options.json_data = {**info.json}
            else:
                if not is_mapping(options.json_data):
                    raise TypeError("Pagination is only supported with mappings")

                options.json_data = {**options.json_data, **info.json}
            return options

        raise ValueError("Unexpected PageInfo state")


class BaseSyncPage(BasePage[_T], Generic[_T]):
    _client: SyncAPIClient = pydantic.PrivateAttr()

    def _set_private_attributes(
        self,
        client: SyncAPIClient,
        model: Type[_T],
        options: FinalRequestOptions,
    ) -> None:
        if (not PYDANTIC_V1) and getattr(self, "__pydantic_private__", None) is None:
            self.__pydantic_private__ = {}

        self._model = model
        self._client = client
        self._options = options

    # Pydantic uses a custom `__iter__` method to support casting BaseModels
    # to dictionaries. e.g. dict(model).
    # As we want to support `for item in page`, this is inherently incompatible
    # with the default pydantic behaviour. It is not possible to support both
    # use cases at once. Fortunately, this is not a big deal as all other pydantic
    # methods should continue to work as expected as there is an alternative method
    # to cast a model to a dictionary, model.dict(), which is used internally
    # by pydantic.
    def __iter__(self) -> Iterator[_T]:  # type: ignore
        for page in self.iter_pages():
            for item in page._get_page_items():
                yield item

    def iter_pages(self: SyncPageT) -> Iterator[SyncPageT]:
        page = self
        while True:
            yield page
            if page.has_next_page():
                page = page.get_next_page()
            else:
                return

    def get_next_page(self: SyncPageT) -> SyncPageT:
        info = self.next_page_info()
        if not info:
            raise RuntimeError(
                "No next page expected; please check `.has_next_page()` before calling `.get_next_page()`."
            )

        options = self._info_to_options(info)
        return self._client._request_api_list(self._model, page=self.__class__, options=options)


class AsyncPaginator(Generic[_T, AsyncPageT]):
    def __init__(
        self,
        client: AsyncAPIClient,
        options: FinalRequestOptions,
        page_cls: Type[AsyncPageT],
        model: Type[_T],
    ) -> None:
        self._model = model
        self._client = client
        self._options = options
        self._page_cls = page_cls

    def __await__(self) -> Generator[Any, None, AsyncPageT]:
        return self._get_page().__await__()

    async def _get_page(self) -> AsyncPageT:
        def _parser(resp: AsyncPageT) -> AsyncPageT:
            resp._set_private_attributes(
                model=self._model,
                options=self._options,
                client=self._client,
            )
            return resp

        self._options.post_parser = _parser

        return await self._client.request(self._page_cls, self._options)

    async def __aiter__(self) -> AsyncIterator[_T]:
        # https://github.com/microsoft/pyright/issues/3464
        page = cast(
            AsyncPageT,
            await self,  # type: ignore
        )
        async for item in page:
            yield item


class BaseAsyncPage(BasePage[_T], Generic[_T]):
    _client: AsyncAPIClient = pydantic.PrivateAttr()

    def _set_private_attributes(
        self,
        model: Type[_T],
        client: AsyncAPIClient,
        options: FinalRequestOptions,
    ) -> None:
        if (not PYDANTIC_V1) and getattr(self, "__pydantic_private__", None) is None:
            self.__pydantic_private__ = {}

        self._model = model
        self._client = client
        self._options = options

    async def __aiter__(self) -> AsyncIterator[_T]:
        async for page in self.iter_pages():
            for item in page._get_page_items():
                yield item

    async def iter_pages(self: AsyncPageT) -> AsyncIterator[AsyncPageT]:
        page = self
        while True:
            yield page
            if page.has_next_page():
                page = await page.get_next_page()
            else:
                return

    async def get_next_page(self: AsyncPageT) -> AsyncPageT:
        info = self.next_page_info()
        if not info:
            raise RuntimeError(
                "No next page expected; please check `.has_next_page()` before calling `.get_next_page()`."
            )

        options = self._info_to_options(info)
        return await self._client._request_api_list(self._model, page=self.__class__, options=options)


_HttpxClientT = TypeVar("_HttpxClientT", bound=Union[httpx2.Client, httpx2.AsyncClient])
_DefaultStreamT = TypeVar("_DefaultStreamT", bound=Union[Stream[Any], AsyncStream[Any]])


class BaseClient(Generic[_HttpxClientT, _DefaultStreamT]):
    _client: _HttpxClientT
    _version: str
    _base_url: URL
    max_retries: int
    timeout: Union[float, Timeout, None]
    _strict_response_validation: bool
    _idempotency_header: str | None
    _default_stream_cls: type[_DefaultStreamT] | None = None

    def __init__(
        self,
        *,
        version: str,
        base_url: str | URL,
        _strict_response_validation: bool,
        max_retries: int = DEFAULT_MAX_RETRIES,
        timeout: float | Timeout | None = DEFAULT_TIMEOUT,
        custom_headers: Mapping[str, str] | None = None,
        custom_query: Mapping[str, object] | None = None,
    ) -> None:
        self._version = version
        self._base_url = self._enforce_trailing_slash(normalize_httpx_url(base_url))
        self.max_retries = max_retries
        self.timeout = timeout
        self._custom_headers = custom_headers or {}
        self._custom_query = custom_query or {}
        self._strict_response_validation = _strict_response_validation
        self._idempotency_header = None
        self._platform: Platform | None = None

        self._validate_max_retries(max_retries)

    def _enforce_trailing_slash(self, url: URL) -> URL:
        if url.raw_path.endswith(b"/"):
            return url
        return url.copy_with(raw_path=url.raw_path + b"/")

    def _make_status_error_from_response(
        self,
        response: httpx2.Response,
    ) -> APIStatusError:
        if response.is_closed and not response.is_stream_consumed:
            # We can't read the response body as it has been closed
            # before it was read. This can happen if an event hook
            # raises a status error.
            body = None
            err_msg = f"Error code: {response.status_code}"
        else:
            err_text = response.text.strip()
            body = err_text

            try:
                body = json.loads(err_text)
                err_msg = f"Error code: {response.status_code} - {body}"
            except Exception:
                err_msg = err_text or f"Error code: {response.status_code}"

        return self._make_status_error(err_msg, body=body, response=response)

    def _make_status_error(
        self,
        err_msg: str,
        *,
        body: object,
        response: httpx2.Response,
    ) -> _exceptions.APIStatusError:
        raise NotImplementedError()

    def _auth_headers(
        self,
        security: SecurityOptions,  # noqa: ARG002
    ) -> dict[str, str]:
        return {}

    def _auth_query(
        self,
        security: SecurityOptions,  # noqa: ARG002
    ) -> dict[str, str]:
        return {}

    def _custom_auth(
        self,
        security: SecurityOptions,  # noqa: ARG002
    ) -> httpx2.Auth | None:
        return None

    def _build_headers(self, options: FinalRequestOptions, *, retries_taken: int = 0) -> httpx2.Headers:
        custom_headers = options.headers or {}
        headers_dict = _merge_mappings({**self._auth_headers(options.security), **self.default_headers}, custom_headers)
        self._validate_headers(headers_dict, custom_headers)

        # headers are case-insensitive while dictionaries are not.
        headers = httpx2.Headers(headers_dict)

        idempotency_header = self._idempotency_header
        if idempotency_header and options.idempotency_key and idempotency_header not in headers:
            headers[idempotency_header] = options.idempotency_key

        # Don't set these headers if they were already set or removed by the caller. We check
        # `custom_headers`, which can contain `Omit()`, instead of `headers` to account for the removal case.
        lower_custom_headers = [header.lower() for header in custom_headers]
        if "x-stainless-retry-count" not in lower_custom_headers:
            headers["x-stainless-retry-count"] = str(retries_taken)
        if "x-stainless-read-timeout" not in lower_custom_headers:
            timeout = normalize_httpx_timeout(
                self.timeout if isinstance(options.timeout, NotGiven) else options.timeout
            )
            if isinstance(timeout, Timeout):
                timeout = timeout.read
            if timeout is not None:
                headers["x-stainless-read-timeout"] = str(timeout)

        return headers

    def _prepare_url(self, url: str) -> URL:
        """
        Merge a URL argument together with any 'base_url' on the client,
        to create the URL used for the outgoing request.
        """
        # Copied from httpx's `_merge_url` method.
        merge_url = URL(url)
        if merge_url.is_relative_url:
            merge_raw_path = self.base_url.raw_path + merge_url.raw_path.lstrip(b"/")
            return self.base_url.copy_with(raw_path=merge_raw_path)

        return merge_url

    def _make_sse_decoder(self) -> SSEDecoder | SSEBytesDecoder:
        return SSEDecoder()

    def _build_request(
        self,
        options: FinalRequestOptions,
        *,
        retries_taken: int = 0,
    ) -> httpx2.Request:
        # Request bodies, files, URLs, and custom options can contain private data.
        log.debug(
            "Building HTTP request: method=%s retries_taken=%i",
            get_http_method_for_logging(options.method),
            retries_taken,
        )
        kwargs: dict[str, Any] = {}

        json_data = options.json_data
        if options.extra_json is not None:
            if json_data is None:
                json_data = cast(Body, options.extra_json)
            elif is_mapping(json_data):
                json_data = _merge_mappings(json_data, options.extra_json)
            else:
                raise RuntimeError(f"Unexpected JSON data type, {type(json_data)}, cannot merge with `extra_body`")

        headers = self._build_headers(options, retries_taken=retries_taken)
        params = _merge_mappings({**self._auth_query(options.security), **self.default_query}, options.params)
        content_type = headers.get("Content-Type")
        files = options.files

        # If the given Content-Type header is multipart/form-data then it
        # has to be removed so that httpx can generate the header with
        # additional information for us as it has to be in this form
        # for the server to be able to correctly parse the request:
        # multipart/form-data; boundary=---abc--
        if content_type is not None and content_type.startswith("multipart/form-data"):
            if "boundary" not in content_type:
                # only remove the header if the boundary hasn't been explicitly set
                # as the caller doesn't want httpx to come up with their own boundary
                headers.pop("Content-Type")

            # As we are now sending multipart/form-data instead of application/json
            # we need to tell httpx to use it, https://www.python-httpx.org/advanced/clients/#multipart-file-encoding
            if json_data:
                if not is_dict(json_data):
                    raise TypeError(
                        f"Expected query input to be a dictionary for multipart requests but got {type(json_data)} instead."
                    )
                kwargs["data"] = self._serialize_multipartform(json_data)

            # httpx determines whether or not to send a "multipart/form-data"
            # request based on the truthiness of the "files" argument.
            # This gets around that issue by generating a dict value that
            # evaluates to true.
            #
            # https://github.com/encode/httpx/discussions/2399#discussioncomment-3814186
            if not files:
                files = cast(HttpxRequestFiles, ForceMultipartDict())

        prepared_url = self._prepare_url(options.url)
        # preserve hard-coded query params from the url
        if params and prepared_url.query:
            params = {**dict(prepared_url.params.items()), **params}
            prepared_url = prepared_url.copy_with(raw_path=prepared_url.raw_path.split(b"?", 1)[0])

        is_body_allowed = options.method.lower() != "get"

        if is_body_allowed:
            if options.content is not None and json_data is not None:
                raise TypeError("Passing both `content` and `json_data` is not supported")
            if options.content is not None and files is not None:
                raise TypeError("Passing both `content` and `files` is not supported")
            if options.content is not None:
                kwargs["content"] = options.content
            elif isinstance(json_data, bytes):
                kwargs["content"] = json_data
            elif not files:
                # Don't set content when JSON is sent as multipart/form-data,
                # since httpx's content param overrides other body arguments
                kwargs["content"] = openapi_dumps(json_data) if is_given(json_data) and json_data is not None else None
            kwargs["files"] = files
        else:
            headers.pop("Content-Type", None)
            kwargs.pop("data", None)

        timeout = self.timeout if isinstance(options.timeout, NotGiven) else options.timeout
        request_url = str(prepared_url)
        request_headers = list(headers.multi_items())
        if is_legacy_httpx_sync_client(self._client) or is_legacy_httpx_async_client(self._client):
            timeout = normalize_legacy_httpx_timeout(timeout)
        else:
            timeout = normalize_httpx2_timeout(timeout)

        # TODO: report this error to httpx
        return self._client.build_request(  # pyright: ignore[reportUnknownMemberType]
            headers=request_headers,
            timeout=timeout,
            method=options.method,
            url=request_url,
            # the `Query` type that we use is incompatible with qs'
            # `Params` type as it needs to be typed as `Mapping[str, object]`
            # so that passing a `TypedDict` doesn't cause an error.
            # https://github.com/microsoft/pyright/issues/3526#event-6715453066
            params=self.qs.stringify(cast(Mapping[str, Any], params)) if params else None,
            **kwargs,
        )

    def _serialize_multipartform(self, data: Mapping[object, object]) -> dict[str, object]:
        items = self.qs.stringify_items(
            # TODO: type ignore is required as stringify_items is well typed but we can't be
            # well typed without heavy validation.
            data,  # type: ignore
            array_format="brackets",
        )
        serialized: dict[str, object] = {}
        for key, value in items:
            existing = serialized.get(key)

            if not existing:
                serialized[key] = value
                continue

            # If a value has already been set for this key then that
            # means we're sending data like `array[]=[1, 2, 3]` and we
            # need to tell httpx that we want to send multiple values with
            # the same key which is done by using a list or a tuple.
            #
            # Note: 2d arrays should never result in the same key at both
            # levels so it's safe to assume that if the value is a list,
            # it was because we changed it to be a list.
            if is_list(existing):
                existing.append(value)
            else:
                serialized[key] = [existing, value]

        return serialized

    def _maybe_override_cast_to(self, cast_to: type[ResponseT], options: FinalRequestOptions) -> type[ResponseT]:
        if not is_given(options.headers):
            return cast_to

        # make a copy of the headers so we don't mutate user-input
        headers = dict(options.headers)

        # we internally support defining a temporary header to override the
        # default `cast_to` type for use with `.with_raw_response` and `.with_streaming_response`
        # see _response.py for implementation details
        override_cast_to = headers.pop(OVERRIDE_CAST_TO_HEADER, not_given)
        if is_given(override_cast_to):
            options.headers = headers
            return cast(Type[ResponseT], override_cast_to)

        return cast_to

    def _should_stream_response_body(self, request: httpx2.Request) -> bool:
        return request.headers.get(RAW_RESPONSE_HEADER) == "stream"  # type: ignore[no-any-return]

    def _process_response_data(
        self,
        *,
        data: object,
        cast_to: type[ResponseT],
        response: httpx2.Response,
    ) -> ResponseT:
        if data is None:
            return cast(ResponseT, None)

        if cast_to is object:
            return cast(ResponseT, data)

        try:
            if inspect.isclass(cast_to) and issubclass(cast_to, ModelBuilderProtocol):
                return cast(ResponseT, cast_to.build(response=response, data=data))

            if self._strict_response_validation:
                return cast(ResponseT, validate_type(type_=cast_to, value=data))

            return cast(ResponseT, construct_type(type_=cast_to, value=data))
        except pydantic.ValidationError as err:
            raise APIResponseValidationError(response=response, body=data) from err

    @property
    def qs(self) -> Querystring:
        return Querystring()

    @property
    def custom_auth(self) -> httpx2.Auth | None:
        return None

    @property
    def auth_headers(self) -> dict[str, str]:
        return {}

    @property
    def default_headers(self) -> dict[str, str | Omit]:
        return {
            "Accept": "application/json",
            "Content-Type": "application/json",
            "User-Agent": self.user_agent,
            **self.platform_headers(),
            **self._custom_headers,
        }

    @property
    def default_query(self) -> dict[str, object]:
        return {
            **self._custom_query,
        }

    def _validate_headers(
        self,
        headers: Headers,  # noqa: ARG002
        custom_headers: Headers,  # noqa: ARG002
    ) -> None:
        """Validate the given default headers and custom headers.

        Does nothing by default.
        """
        return

    @property
    def user_agent(self) -> str:
        return f"{self.__class__.__name__}/Python {self._version}"

    @property
    def base_url(self) -> URL:
        return self._base_url

    @base_url.setter
    def base_url(self, url: URL | str) -> None:
        self._base_url = self._enforce_trailing_slash(normalize_httpx_url(url))

    def platform_headers(self) -> Dict[str, str]:
        # the actual implementation is in a separate `lru_cache` decorated
        # function because adding `lru_cache` to methods will leak memory
        # https://github.com/python/cpython/issues/88476
        return platform_headers(self._version, platform=self._platform)

    def _parse_retry_after_header(self, response_headers: Optional[httpx2.Headers] = None) -> float | None:
        """Returns a float of the number of seconds (not milliseconds) to wait after retrying, or None if unspecified.

        About the Retry-After header: https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Retry-After
        See also  https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Retry-After#syntax
        """
        if response_headers is None:
            return None

        # Prefer milliseconds, then seconds (allowing nonstandard floats).
        for header, divisor in (("retry-after-ms", 1000), ("retry-after", 1)):
            value = response_headers.get(header)
            if value is None:
                continue
            try:
                delay = float(value)
            except ValueError:
                continue
            if delay == math.inf and value.strip().lower() not in ("inf", "+inf", "infinity", "+infinity"):
                # Numeric overflow is an excessive server delay, not a malformed
                # infinity literal. Keep a finite sentinel so retry eligibility
                # refuses it instead of falling back to a shorter wait.
                return MAX_RETRY_AFTER_DELAY + 1
            return delay / divisor

        # Last, try parsing `retry-after` as a date.
        retry_header = response_headers.get("retry-after")
        try:
            retry_date_tuple = email.utils.parsedate_tz(retry_header)
            if retry_date_tuple is None:
                return None

            retry_date = email.utils.mktime_tz(retry_date_tuple)
        except (TypeError, ValueError, OverflowError, OSError):
            return None

        return float(retry_date - time.time())

    @staticmethod
    def _validate_max_retries(value: object) -> None:
        if value is None:
            raise TypeError(
                "max_retries cannot be None. Use 0 to disable retries or a large integer for a larger retry budget."
            )
        if not isinstance(value, int):
            raise TypeError("max_retries must be a non-negative integer")
        if value < 0:
            raise ValueError("max_retries must be a non-negative integer")

    def _calculate_retry_timeout(
        self,
        remaining_retries: int,
        options: FinalRequestOptions,
        response_headers: Optional[httpx2.Headers] = None,
    ) -> float:
        max_retries = options.get_max_retries(self.max_retries)

        # Honor server-directed delays up to two minutes.
        retry_after = self._parse_retry_after_header(response_headers)
        if retry_after is not None and math.isfinite(retry_after) and 0 < retry_after <= MAX_RETRY_AFTER_DELAY:
            return retry_after

        # Also cap retry count to 1000 to avoid any potential overflows with `pow`
        nb_retries = min(max_retries - remaining_retries, 1000)

        # Apply exponential backoff, but not more than the max.
        sleep_seconds = min(INITIAL_RETRY_DELAY * pow(2.0, nb_retries), MAX_RETRY_DELAY)

        # Reduce the calculated timeout by a random range between 0-25%
        jitter = 1 - 0.25 * random()
        timeout = sleep_seconds * jitter
        return timeout if timeout >= 0 else 0

    def _should_retry(self, response: httpx2.Response) -> bool:
        retry_after = self._parse_retry_after_header(response.headers)
        if retry_after is not None and math.isfinite(retry_after) and retry_after > MAX_RETRY_AFTER_DELAY:
            log.debug(
                "Not retrying because `Retry-After` of %s seconds exceeds the maximum of %s seconds",
                retry_after,
                MAX_RETRY_AFTER_DELAY,
            )
            return False

        # Note: this is not a standard header
        should_retry_header = response.headers.get("x-should-retry")

        # If the server explicitly says whether or not to retry, obey.
        if should_retry_header == "true":
            log.debug("Retrying as header `x-should-retry` is set to `true`")
            return True
        if should_retry_header == "false":
            log.debug("Not retrying as header `x-should-retry` is set to `false`")
            return False

        # Retry on request timeouts.
        if response.status_code == 408:
            log.debug("Retrying due to status code %i", response.status_code)
            return True

        # Retry on lock timeouts.
        if response.status_code == 409:
            log.debug("Retrying due to status code %i", response.status_code)
            return True

        # Retry on rate limits.
        if response.status_code == 429:
            log.debug("Retrying due to status code %i", response.status_code)
            return True

        # Retry internal errors.
        if response.status_code >= 500:
            log.debug("Retrying due to status code %i", response.status_code)
            return True

        log.debug("Not retrying")
        return False

    def _idempotency_key(self) -> str:
        return f"stainless-python-retry-{uuid.uuid4()}"


class _DefaultHttpxClient(httpx2.Client):
    def __init__(self, **kwargs: Any) -> None:
        kwargs.setdefault("timeout", DEFAULT_TIMEOUT)
        kwargs.setdefault("limits", DEFAULT_CONNECTION_LIMITS)
        kwargs.setdefault("follow_redirects", True)
        super().__init__(**kwargs)


if TYPE_CHECKING:
    DefaultHttpxClient = httpx2.Client
    """An alias to `httpx2.Client` that provides the same defaults that this SDK
    uses internally.

    This is useful because overriding the `http_client` with your own instance of
    `httpx2.Client` will result in HTTPX2's defaults being used, not ours.
    """
else:
    DefaultHttpxClient = _DefaultHttpxClient


class SyncHttpxClientWrapper(DefaultHttpxClient):
    def __del__(self) -> None:
        if self.is_closed:
            return

        try:
            self.close()
        except Exception:
            pass


class SyncAPIClient(BaseClient[httpx2.Client, Stream[Any]]):
    _client: httpx2.Client
    _default_stream_cls: type[Stream[Any]] | None = None

    def __init__(
        self,
        *,
        version: str,
        base_url: str | URL,
        max_retries: int = DEFAULT_MAX_RETRIES,
        timeout: float | Timeout | None | NotGiven = not_given,
        http_client: httpx2.Client | None = None,
        custom_headers: Mapping[str, str] | None = None,
        custom_query: Mapping[str, object] | None = None,
        _strict_response_validation: bool,
    ) -> None:
        if not is_given(timeout):
            # if the user passed in a custom http client with a non-default
            # timeout set then we use that timeout.
            #
            # note: there is an edge case here where the user passes in a client
            # where they've explicitly set the timeout to match the default timeout
            # as this check is structural, meaning that we'll think they didn't
            # pass in a timeout and will ignore it
            client_timeout = normalize_httpx_timeout(http_client.timeout) if http_client else None
            if http_client and client_timeout != HTTPX_DEFAULT_TIMEOUT:
                timeout = client_timeout
            else:
                timeout = DEFAULT_TIMEOUT

        if (
            http_client is not None
            and not is_httpx2_sync_client(http_client)
            and not is_legacy_httpx_sync_client(http_client)
        ):
            raise TypeError(
                "Invalid `http_client` argument; Expected an instance of `httpx.Client` or `httpx2.Client` "
                f"but got {type(http_client)}"
            )

        super().__init__(
            version=version,
            # cast to a valid type because mypy doesn't understand our type narrowing
            timeout=cast(Timeout, timeout),
            base_url=base_url,
            max_retries=max_retries,
            custom_query=custom_query,
            custom_headers=custom_headers,
            _strict_response_validation=_strict_response_validation,
        )
        self._client = http_client or SyncHttpxClientWrapper(
            base_url=base_url,
            # cast to a valid type because mypy doesn't understand our type narrowing
            timeout=cast(Timeout, timeout),
        )

    def is_closed(self) -> bool:
        return self._client.is_closed

    def close(self) -> None:
        """Close the underlying HTTPX client.

        The client will *not* be usable after this.
        """
        # If an error is thrown while constructing a client, self._client
        # may not be present
        if hasattr(self, "_client"):
            self._client.close()

    def __enter__(self: _T) -> _T:
        return self

    def __exit__(
        self,
        exc_type: type[BaseException] | None,
        exc: BaseException | None,
        exc_tb: TracebackType | None,
    ) -> None:
        self.close()

    def _prepare_options(
        self,
        options: FinalRequestOptions,  # noqa: ARG002
    ) -> FinalRequestOptions:
        """Hook for mutating the given options"""
        return options

    def _prepare_request(
        self,
        request: httpx2.Request,  # noqa: ARG002
    ) -> None:
        """This method is used as a callback for mutating the `Request` object
        after it has been constructed.
        This is useful for cases where you want to add certain headers based off of
        the request properties, e.g. `url`, `method` etc.
        """
        return None

    def _send_request(
        self,
        request: httpx2.Request,
        *,
        stream: bool,
        **kwargs: Unpack[HttpxSendArgs],
    ) -> httpx2.Response:
        return self._client.send(request, stream=stream, **kwargs)

    @overload
    def request(
        self,
        cast_to: Type[ResponseT],
        options: FinalRequestOptions,
        *,
        stream: Literal[True],
        stream_cls: Type[_StreamT],
    ) -> _StreamT: ...

    @overload
    def request(
        self,
        cast_to: Type[ResponseT],
        options: FinalRequestOptions,
        *,
        stream: Literal[False] = False,
    ) -> ResponseT: ...

    @overload
    def request(
        self,
        cast_to: Type[ResponseT],
        options: FinalRequestOptions,
        *,
        stream: bool = False,
        stream_cls: Type[_StreamT] | None = None,
    ) -> ResponseT | _StreamT: ...

    def request(
        self,
        cast_to: Type[ResponseT],
        options: FinalRequestOptions,
        *,
        stream: bool = False,
        stream_cls: type[_StreamT] | None = None,
    ) -> ResponseT | _StreamT:
        cast_to = self._maybe_override_cast_to(cast_to, options)

        # create a copy of the options we were given so that if the
        # options are mutated later & we then retry, the retries are
        # given the original options
        input_options = model_copy(options)
        if input_options.idempotency_key is None and input_options.method.lower() != "get":
            # ensure the idempotency key is reused between requests
            input_options.idempotency_key = self._idempotency_key()

        response: httpx2.Response | None = None
        max_retries = input_options.get_max_retries(self.max_retries)
        self._validate_max_retries(max_retries)

        retries_taken = 0
        for retries_taken in range(max_retries + 1):
            options = model_copy(input_options)
            options = self._prepare_options(options)

            remaining_retries = max_retries - retries_taken
            request = self._build_request(options, retries_taken=retries_taken)
            self._prepare_request(request)

            kwargs: HttpxSendArgs = {}
            custom_auth = self._custom_auth(options.security)
            if custom_auth is not None:
                kwargs["auth"] = (
                    normalize_httpx2_auth(custom_auth)
                    if is_httpx2_sync_client(self._client)
                    else normalize_legacy_httpx_auth(custom_auth)
                )

            if options.follow_redirects is not None:
                kwargs["follow_redirects"] = options.follow_redirects

            log.debug("Sending HTTP Request: %s", get_http_method_for_logging(request.method))

            response = None
            try:
                response = self._send_request(
                    request,
                    stream=stream or self._should_stream_response_body(request=request),
                    **kwargs,
                )
            except timeout_exceptions() as err:
                log.debug("Encountered a timeout exception: %s", type(err).__name__)

                if remaining_retries > 0:
                    self._sleep_for_retry(
                        retries_taken=retries_taken,
                        max_retries=max_retries,
                        options=input_options,
                        response=None,
                    )
                    continue

                log.debug("Raising timeout error")
                raise APITimeoutError(request=request) from err
            except OpenAIError as err:
                # Propagate OpenAIErrors as-is, without retrying or wrapping in APIConnectionError
                raise err
            except request_exceptions() as err:
                log.debug("Encountered exception: %s", type(err).__name__)

                if remaining_retries > 0:
                    self._sleep_for_retry(
                        retries_taken=retries_taken,
                        max_retries=max_retries,
                        options=input_options,
                        response=None,
                    )
                    continue

                log.debug("Raising connection error")
                raise APIConnectionError(request=request) from err

            log.debug(
                "HTTP Response: %s %i",
                get_http_method_for_logging(request.method),
                response.status_code,
            )
            log.debug("request_id: %s", response.headers.get("x-request-id"))

            try:
                response.raise_for_status()
            except status_exceptions() as err:  # thrown on 4xx and 5xx status code
                log.debug("Encountered an HTTP status error: %i", response.status_code)

                if remaining_retries > 0 and self._should_retry(err.response):
                    err.response.close()
                    self._sleep_for_retry(
                        retries_taken=retries_taken,
                        max_retries=max_retries,
                        options=input_options,
                        response=response,
                    )
                    continue

                # If the response is streamed then we need to explicitly read the response
                # to completion before attempting to access the response text.
                if not err.response.is_closed:
                    err.response.read()

                log.debug("Re-raising status error")
                raise self._make_status_error_from_response(err.response) from None

            break

        assert response is not None, "could not resolve response (should never happen)"
        return self._process_response(
            cast_to=cast_to,
            options=options,
            response=response,
            stream=stream,
            stream_cls=stream_cls,
            retries_taken=retries_taken,
        )

    def _sleep_for_retry(
        self, *, retries_taken: int, max_retries: int, options: FinalRequestOptions, response: httpx2.Response | None
    ) -> None:
        remaining_retries = max_retries - retries_taken
        if remaining_retries == 1:
            log.debug("1 retry left")
        else:
            log.debug("%i retries left", remaining_retries)

        timeout = self._calculate_retry_timeout(remaining_retries, options, response.headers if response else None)
        log.info("Retrying request in %f seconds (retry %i of %s)", timeout, retries_taken + 1, max_retries)

        time.sleep(timeout)

    def _process_response(
        self,
        *,
        cast_to: Type[ResponseT],
        options: FinalRequestOptions,
        response: httpx2.Response,
        stream: bool,
        stream_cls: type[Stream[Any]] | type[AsyncStream[Any]] | None,
        retries_taken: int = 0,
    ) -> ResponseT:
        if response.request.headers.get(RAW_RESPONSE_HEADER) == "true":
            return cast(
                ResponseT,
                LegacyAPIResponse(
                    raw=response,
                    client=self,
                    cast_to=cast_to,
                    stream=stream,
                    stream_cls=stream_cls,
                    options=options,
                    retries_taken=retries_taken,
                ),
            )

        origin = get_origin(cast_to) or cast_to

        if (
            inspect.isclass(origin)
            and issubclass(origin, BaseAPIResponse)
            # we only want to actually return the custom BaseAPIResponse class if we're
            # returning the raw response, or if we're not streaming SSE, as if we're streaming
            # SSE then `cast_to` doesn't actively reflect the type we need to parse into
            and (not stream or bool(response.request.headers.get(RAW_RESPONSE_HEADER)))
        ):
            if not issubclass(origin, APIResponse):
                raise TypeError(f"API Response types must subclass {APIResponse}; Received {origin}")

            response_cls = cast("type[BaseAPIResponse[Any]]", cast_to)
            return cast(
                ResponseT,
                response_cls(
                    raw=response,
                    client=self,
                    cast_to=extract_response_type(response_cls),
                    stream=stream,
                    stream_cls=stream_cls,
                    options=options,
                    retries_taken=retries_taken,
                ),
            )

        if cast_to in http_response_types():
            return cast(ResponseT, response)

        api_response = APIResponse(
            raw=response,
            client=self,
            cast_to=cast("type[ResponseT]", cast_to),  # pyright: ignore[reportUnnecessaryCast]
            stream=stream,
            stream_cls=stream_cls,
            options=options,
            retries_taken=retries_taken,
        )
        if bool(response.request.headers.get(RAW_RESPONSE_HEADER)):
            return cast(ResponseT, api_response)

        return api_response.parse()

    def _request_api_list(
        self,
        model: Type[object],
        page: Type[SyncPageT],
        options: FinalRequestOptions,
    ) -> SyncPageT:
        def _parser(resp: SyncPageT) -> SyncPageT:
            resp._set_private_attributes(
                client=self,
                model=model,
                options=options,
            )
            return resp

        options.post_parser = _parser

        return self.request(page, options, stream=False)

    @overload
    def get(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        options: RequestOptions = {},
        stream: Literal[False] = False,
    ) -> ResponseT: ...

    @overload
    def get(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        options: RequestOptions = {},
        stream: Literal[True],
        stream_cls: type[_StreamT],
    ) -> _StreamT: ...

    @overload
    def get(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        options: RequestOptions = {},
        stream: bool,
        stream_cls: type[_StreamT] | None = None,
    ) -> ResponseT | _StreamT: ...

    def get(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        options: RequestOptions = {},
        stream: bool = False,
        stream_cls: type[_StreamT] | None = None,
    ) -> ResponseT | _StreamT:
        opts = FinalRequestOptions.construct(method="get", url=path, **options)
        # cast is required because mypy complains about returning Any even though
        # it understands the type variables
        return cast(ResponseT, self.request(cast_to, opts, stream=stream, stream_cls=stream_cls))

    @overload
    def post(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        body: Body | None = None,
        content: BinaryTypes | None = None,
        options: RequestOptions = {},
        files: RequestFiles | None = None,
        stream: Literal[False] = False,
    ) -> ResponseT: ...

    @overload
    def post(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        body: Body | None = None,
        content: BinaryTypes | None = None,
        options: RequestOptions = {},
        files: RequestFiles | None = None,
        stream: Literal[True],
        stream_cls: type[_StreamT],
    ) -> _StreamT: ...

    @overload
    def post(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        body: Body | None = None,
        content: BinaryTypes | None = None,
        options: RequestOptions = {},
        files: RequestFiles | None = None,
        stream: bool,
        stream_cls: type[_StreamT] | None = None,
    ) -> ResponseT | _StreamT: ...

    def post(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        body: Body | None = None,
        content: BinaryTypes | None = None,
        options: RequestOptions = {},
        files: RequestFiles | None = None,
        stream: bool = False,
        stream_cls: type[_StreamT] | None = None,
    ) -> ResponseT | _StreamT:
        if body is not None and content is not None:
            raise TypeError("Passing both `body` and `content` is not supported")
        if files is not None and content is not None:
            raise TypeError("Passing both `files` and `content` is not supported")
        if isinstance(body, bytes):
            warnings.warn(
                "Passing raw bytes as `body` is deprecated and will be removed in a future version. "
                "Please pass raw bytes via the `content` parameter instead.",
                DeprecationWarning,
                stacklevel=2,
            )
        opts = FinalRequestOptions.construct(
            method="post", url=path, json_data=body, content=content, files=to_httpx_files(files), **options
        )
        return cast(ResponseT, self.request(cast_to, opts, stream=stream, stream_cls=stream_cls))

    def patch(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        body: Body | None = None,
        content: BinaryTypes | None = None,
        files: RequestFiles | None = None,
        options: RequestOptions = {},
    ) -> ResponseT:
        if body is not None and content is not None:
            raise TypeError("Passing both `body` and `content` is not supported")
        if files is not None and content is not None:
            raise TypeError("Passing both `files` and `content` is not supported")
        if isinstance(body, bytes):
            warnings.warn(
                "Passing raw bytes as `body` is deprecated and will be removed in a future version. "
                "Please pass raw bytes via the `content` parameter instead.",
                DeprecationWarning,
                stacklevel=2,
            )
        opts = FinalRequestOptions.construct(
            method="patch", url=path, json_data=body, content=content, files=to_httpx_files(files), **options
        )
        return self.request(cast_to, opts)

    def put(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        body: Body | None = None,
        content: BinaryTypes | None = None,
        files: RequestFiles | None = None,
        options: RequestOptions = {},
    ) -> ResponseT:
        if body is not None and content is not None:
            raise TypeError("Passing both `body` and `content` is not supported")
        if files is not None and content is not None:
            raise TypeError("Passing both `files` and `content` is not supported")
        if isinstance(body, bytes):
            warnings.warn(
                "Passing raw bytes as `body` is deprecated and will be removed in a future version. "
                "Please pass raw bytes via the `content` parameter instead.",
                DeprecationWarning,
                stacklevel=2,
            )
        opts = FinalRequestOptions.construct(
            method="put", url=path, json_data=body, content=content, files=to_httpx_files(files), **options
        )
        return self.request(cast_to, opts)

    def delete(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        body: Body | None = None,
        content: BinaryTypes | None = None,
        options: RequestOptions = {},
    ) -> ResponseT:
        if body is not None and content is not None:
            raise TypeError("Passing both `body` and `content` is not supported")
        if isinstance(body, bytes):
            warnings.warn(
                "Passing raw bytes as `body` is deprecated and will be removed in a future version. "
                "Please pass raw bytes via the `content` parameter instead.",
                DeprecationWarning,
                stacklevel=2,
            )
        opts = FinalRequestOptions.construct(method="delete", url=path, json_data=body, content=content, **options)
        return self.request(cast_to, opts)

    def get_api_list(
        self,
        path: str,
        *,
        model: Type[object],
        page: Type[SyncPageT],
        body: Body | None = None,
        options: RequestOptions = {},
        method: str = "get",
    ) -> SyncPageT:
        opts = FinalRequestOptions.construct(method=method, url=path, json_data=body, **options)
        return self._request_api_list(model, page, opts)


class _DefaultAsyncHttpxClient(httpx2.AsyncClient):
    def __init__(self, **kwargs: Any) -> None:
        kwargs.setdefault("timeout", DEFAULT_TIMEOUT)
        kwargs.setdefault("limits", DEFAULT_CONNECTION_LIMITS)
        kwargs.setdefault("follow_redirects", True)
        super().__init__(**kwargs)


_DefaultAioHttpClient: type[httpx2.AsyncClient]

try:
    from ._vendor.httpx_aiohttp import Httpx2AiohttpClient
except ImportError:

    class _MissingAioHttpClient(httpx2.AsyncClient):
        def __init__(self, **_kwargs: Any) -> None:
            raise RuntimeError("To use the aiohttp client you must have installed the package with the `aiohttp` extra")

    _DefaultAioHttpClient = _MissingAioHttpClient
else:

    class _InstalledAioHttpClient(Httpx2AiohttpClient):
        def __init__(self, **kwargs: Any) -> None:
            kwargs.setdefault("timeout", DEFAULT_TIMEOUT)
            kwargs.setdefault("limits", DEFAULT_CONNECTION_LIMITS)
            kwargs.setdefault("follow_redirects", True)
            super().__init__(**kwargs)

    _DefaultAioHttpClient = _InstalledAioHttpClient


if TYPE_CHECKING:
    DefaultAsyncHttpxClient = httpx2.AsyncClient
    """An alias to `httpx2.AsyncClient` that provides the same defaults that this SDK
    uses internally.

    This is useful because overriding the `http_client` with your own instance of
    `httpx2.AsyncClient` will result in HTTPX2's defaults being used, not ours.
    """

    DefaultAioHttpClient = httpx2.AsyncClient
    """An alias to `httpx2.AsyncClient` that changes the default HTTP transport to `aiohttp`."""
else:
    DefaultAsyncHttpxClient = _DefaultAsyncHttpxClient
    DefaultAioHttpClient = _DefaultAioHttpClient


class AsyncHttpxClientWrapper(DefaultAsyncHttpxClient):
    def __del__(self) -> None:
        if self.is_closed:
            return

        try:
            # TODO(someday): support non asyncio runtimes here
            asyncio.get_running_loop().create_task(self.aclose())
        except Exception:
            pass


class AsyncAPIClient(BaseClient[httpx2.AsyncClient, AsyncStream[Any]]):
    _client: httpx2.AsyncClient
    _default_stream_cls: type[AsyncStream[Any]] | None = None

    def __init__(
        self,
        *,
        version: str,
        base_url: str | URL,
        _strict_response_validation: bool,
        max_retries: int = DEFAULT_MAX_RETRIES,
        timeout: float | Timeout | None | NotGiven = not_given,
        http_client: httpx2.AsyncClient | None = None,
        custom_headers: Mapping[str, str] | None = None,
        custom_query: Mapping[str, object] | None = None,
    ) -> None:
        if not is_given(timeout):
            # if the user passed in a custom http client with a non-default
            # timeout set then we use that timeout.
            #
            # note: there is an edge case here where the user passes in a client
            # where they've explicitly set the timeout to match the default timeout
            # as this check is structural, meaning that we'll think they didn't
            # pass in a timeout and will ignore it
            client_timeout = normalize_httpx_timeout(http_client.timeout) if http_client else None
            if http_client and client_timeout != HTTPX_DEFAULT_TIMEOUT:
                timeout = client_timeout
            else:
                timeout = DEFAULT_TIMEOUT

        if (
            http_client is not None
            and not is_httpx2_async_client(http_client)
            and not is_legacy_httpx_async_client(http_client)
        ):
            raise TypeError(
                "Invalid `http_client` argument; Expected an instance of `httpx.AsyncClient` or "
                f"`httpx2.AsyncClient` but got {type(http_client)}"
            )

        super().__init__(
            version=version,
            base_url=base_url,
            # cast to a valid type because mypy doesn't understand our type narrowing
            timeout=cast(Timeout, timeout),
            max_retries=max_retries,
            custom_query=custom_query,
            custom_headers=custom_headers,
            _strict_response_validation=_strict_response_validation,
        )
        self._client = http_client or AsyncHttpxClientWrapper(
            base_url=base_url,
            # cast to a valid type because mypy doesn't understand our type narrowing
            timeout=cast(Timeout, timeout),
        )

    def is_closed(self) -> bool:
        return self._client.is_closed

    async def close(self) -> None:
        """Close the underlying HTTPX client.

        The client will *not* be usable after this.
        """
        await self._client.aclose()

    async def __aenter__(self: _T) -> _T:
        return self

    async def __aexit__(
        self,
        exc_type: type[BaseException] | None,
        exc: BaseException | None,
        exc_tb: TracebackType | None,
    ) -> None:
        await self.close()

    async def _prepare_options(
        self,
        options: FinalRequestOptions,  # noqa: ARG002
    ) -> FinalRequestOptions:
        """Hook for mutating the given options"""
        return options

    async def _prepare_request(
        self,
        request: httpx2.Request,  # noqa: ARG002
    ) -> None:
        """This method is used as a callback for mutating the `Request` object
        after it has been constructed.
        This is useful for cases where you want to add certain headers based off of
        the request properties, e.g. `url`, `method` etc.
        """
        return None

    async def _send_request(
        self,
        request: httpx2.Request,
        *,
        stream: bool,
        **kwargs: Unpack[HttpxSendArgs],
    ) -> httpx2.Response:
        return await self._client.send(request, stream=stream, **kwargs)

    @overload
    async def request(
        self,
        cast_to: Type[ResponseT],
        options: FinalRequestOptions,
        *,
        stream: Literal[False] = False,
    ) -> ResponseT: ...

    @overload
    async def request(
        self,
        cast_to: Type[ResponseT],
        options: FinalRequestOptions,
        *,
        stream: Literal[True],
        stream_cls: type[_AsyncStreamT],
    ) -> _AsyncStreamT: ...

    @overload
    async def request(
        self,
        cast_to: Type[ResponseT],
        options: FinalRequestOptions,
        *,
        stream: bool,
        stream_cls: type[_AsyncStreamT] | None = None,
    ) -> ResponseT | _AsyncStreamT: ...

    async def request(
        self,
        cast_to: Type[ResponseT],
        options: FinalRequestOptions,
        *,
        stream: bool = False,
        stream_cls: type[_AsyncStreamT] | None = None,
    ) -> ResponseT | _AsyncStreamT:
        if self._platform is None:
            # `get_platform` can make blocking IO calls so we
            # execute it earlier while we are in an async context
            self._platform = await asyncify(get_platform)()

        cast_to = self._maybe_override_cast_to(cast_to, options)

        # create a copy of the options we were given so that if the
        # options are mutated later & we then retry, the retries are
        # given the original options
        input_options = model_copy(options)
        if input_options.idempotency_key is None and input_options.method.lower() != "get":
            # ensure the idempotency key is reused between requests
            input_options.idempotency_key = self._idempotency_key()

        response: httpx2.Response | None = None
        max_retries = input_options.get_max_retries(self.max_retries)
        self._validate_max_retries(max_retries)

        retries_taken = 0
        for retries_taken in range(max_retries + 1):
            options = model_copy(input_options)
            options = await self._prepare_options(options)

            remaining_retries = max_retries - retries_taken
            request = self._build_request(options, retries_taken=retries_taken)
            await self._prepare_request(request)

            kwargs: HttpxSendArgs = {}
            if self.custom_auth is not None:
                kwargs["auth"] = (
                    normalize_httpx2_auth(self.custom_auth)
                    if is_httpx2_async_client(self._client)
                    else normalize_legacy_httpx_auth(self.custom_auth)
                )

            if options.follow_redirects is not None:
                kwargs["follow_redirects"] = options.follow_redirects

            log.debug("Sending HTTP Request: %s", get_http_method_for_logging(request.method))

            response = None
            try:
                response = await self._send_request(
                    request,
                    stream=stream or self._should_stream_response_body(request=request),
                    **kwargs,
                )
            except timeout_exceptions() as err:
                log.debug("Encountered a timeout exception: %s", type(err).__name__)

                if remaining_retries > 0:
                    await self._sleep_for_retry(
                        retries_taken=retries_taken,
                        max_retries=max_retries,
                        options=input_options,
                        response=None,
                    )
                    continue

                log.debug("Raising timeout error")
                raise APITimeoutError(request=request) from err
            except OpenAIError as err:
                # Propagate OpenAIErrors as-is, without retrying or wrapping in APIConnectionError
                raise err
            except request_exceptions() as err:
                log.debug("Encountered exception: %s", type(err).__name__)

                if remaining_retries > 0:
                    await self._sleep_for_retry(
                        retries_taken=retries_taken,
                        max_retries=max_retries,
                        options=input_options,
                        response=None,
                    )
                    continue

                log.debug("Raising connection error")
                raise APIConnectionError(request=request) from err

            log.debug(
                "HTTP Response: %s %i",
                get_http_method_for_logging(request.method),
                response.status_code,
            )
            log.debug("request_id: %s", response.headers.get("x-request-id"))

            try:
                response.raise_for_status()
            except status_exceptions() as err:  # thrown on 4xx and 5xx status code
                log.debug("Encountered an HTTP status error: %i", response.status_code)

                if remaining_retries > 0 and self._should_retry(err.response):
                    await err.response.aclose()
                    await self._sleep_for_retry(
                        retries_taken=retries_taken,
                        max_retries=max_retries,
                        options=input_options,
                        response=response,
                    )
                    continue

                # If the response is streamed then we need to explicitly read the response
                # to completion before attempting to access the response text.
                if not err.response.is_closed:
                    await err.response.aread()

                log.debug("Re-raising status error")
                raise self._make_status_error_from_response(err.response) from None

            break

        assert response is not None, "could not resolve response (should never happen)"
        return await self._process_response(
            cast_to=cast_to,
            options=options,
            response=response,
            stream=stream,
            stream_cls=stream_cls,
            retries_taken=retries_taken,
        )

    async def _sleep_for_retry(
        self, *, retries_taken: int, max_retries: int, options: FinalRequestOptions, response: httpx2.Response | None
    ) -> None:
        remaining_retries = max_retries - retries_taken
        if remaining_retries == 1:
            log.debug("1 retry left")
        else:
            log.debug("%i retries left", remaining_retries)

        timeout = self._calculate_retry_timeout(remaining_retries, options, response.headers if response else None)
        log.info("Retrying request in %f seconds (retry %i of %s)", timeout, retries_taken + 1, max_retries)

        await anyio.sleep(timeout)

    async def _process_response(
        self,
        *,
        cast_to: Type[ResponseT],
        options: FinalRequestOptions,
        response: httpx2.Response,
        stream: bool,
        stream_cls: type[Stream[Any]] | type[AsyncStream[Any]] | None,
        retries_taken: int = 0,
    ) -> ResponseT:
        if response.request.headers.get(RAW_RESPONSE_HEADER) == "true":
            return cast(
                ResponseT,
                LegacyAPIResponse(
                    raw=response,
                    client=self,
                    cast_to=cast_to,
                    stream=stream,
                    stream_cls=stream_cls,
                    options=options,
                    retries_taken=retries_taken,
                ),
            )

        origin = get_origin(cast_to) or cast_to

        if (
            inspect.isclass(origin)
            and issubclass(origin, BaseAPIResponse)
            # we only want to actually return the custom BaseAPIResponse class if we're
            # returning the raw response, or if we're not streaming SSE, as if we're streaming
            # SSE then `cast_to` doesn't actively reflect the type we need to parse into
            and (not stream or bool(response.request.headers.get(RAW_RESPONSE_HEADER)))
        ):
            if not issubclass(origin, AsyncAPIResponse):
                raise TypeError(f"API Response types must subclass {AsyncAPIResponse}; Received {origin}")

            response_cls = cast("type[BaseAPIResponse[Any]]", cast_to)
            return cast(
                "ResponseT",
                response_cls(
                    raw=response,
                    client=self,
                    cast_to=extract_response_type(response_cls),
                    stream=stream,
                    stream_cls=stream_cls,
                    options=options,
                    retries_taken=retries_taken,
                ),
            )

        if cast_to in http_response_types():
            return cast(ResponseT, response)

        api_response = AsyncAPIResponse(
            raw=response,
            client=self,
            cast_to=cast("type[ResponseT]", cast_to),  # pyright: ignore[reportUnnecessaryCast]
            stream=stream,
            stream_cls=stream_cls,
            options=options,
            retries_taken=retries_taken,
        )
        if bool(response.request.headers.get(RAW_RESPONSE_HEADER)):
            return cast(ResponseT, api_response)

        return await api_response.parse()

    def _request_api_list(
        self,
        model: Type[_T],
        page: Type[AsyncPageT],
        options: FinalRequestOptions,
    ) -> AsyncPaginator[_T, AsyncPageT]:
        return AsyncPaginator(client=self, options=options, page_cls=page, model=model)

    @overload
    async def get(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        options: RequestOptions = {},
        stream: Literal[False] = False,
    ) -> ResponseT: ...

    @overload
    async def get(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        options: RequestOptions = {},
        stream: Literal[True],
        stream_cls: type[_AsyncStreamT],
    ) -> _AsyncStreamT: ...

    @overload
    async def get(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        options: RequestOptions = {},
        stream: bool,
        stream_cls: type[_AsyncStreamT] | None = None,
    ) -> ResponseT | _AsyncStreamT: ...

    async def get(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        options: RequestOptions = {},
        stream: bool = False,
        stream_cls: type[_AsyncStreamT] | None = None,
    ) -> ResponseT | _AsyncStreamT:
        opts = FinalRequestOptions.construct(method="get", url=path, **options)
        return await self.request(cast_to, opts, stream=stream, stream_cls=stream_cls)

    @overload
    async def post(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        body: Body | None = None,
        content: AsyncBinaryTypes | None = None,
        files: RequestFiles | None = None,
        options: RequestOptions = {},
        stream: Literal[False] = False,
    ) -> ResponseT: ...

    @overload
    async def post(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        body: Body | None = None,
        content: AsyncBinaryTypes | None = None,
        files: RequestFiles | None = None,
        options: RequestOptions = {},
        stream: Literal[True],
        stream_cls: type[_AsyncStreamT],
    ) -> _AsyncStreamT: ...

    @overload
    async def post(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        body: Body | None = None,
        content: AsyncBinaryTypes | None = None,
        files: RequestFiles | None = None,
        options: RequestOptions = {},
        stream: bool,
        stream_cls: type[_AsyncStreamT] | None = None,
    ) -> ResponseT | _AsyncStreamT: ...

    async def post(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        body: Body | None = None,
        content: AsyncBinaryTypes | None = None,
        files: RequestFiles | None = None,
        options: RequestOptions = {},
        stream: bool = False,
        stream_cls: type[_AsyncStreamT] | None = None,
    ) -> ResponseT | _AsyncStreamT:
        if body is not None and content is not None:
            raise TypeError("Passing both `body` and `content` is not supported")
        if files is not None and content is not None:
            raise TypeError("Passing both `files` and `content` is not supported")
        if isinstance(body, bytes):
            warnings.warn(
                "Passing raw bytes as `body` is deprecated and will be removed in a future version. "
                "Please pass raw bytes via the `content` parameter instead.",
                DeprecationWarning,
                stacklevel=2,
            )
        opts = FinalRequestOptions.construct(
            method="post", url=path, json_data=body, content=content, files=await async_to_httpx_files(files), **options
        )
        return await self.request(cast_to, opts, stream=stream, stream_cls=stream_cls)

    async def patch(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        body: Body | None = None,
        content: AsyncBinaryTypes | None = None,
        files: RequestFiles | None = None,
        options: RequestOptions = {},
    ) -> ResponseT:
        if body is not None and content is not None:
            raise TypeError("Passing both `body` and `content` is not supported")
        if files is not None and content is not None:
            raise TypeError("Passing both `files` and `content` is not supported")
        if isinstance(body, bytes):
            warnings.warn(
                "Passing raw bytes as `body` is deprecated and will be removed in a future version. "
                "Please pass raw bytes via the `content` parameter instead.",
                DeprecationWarning,
                stacklevel=2,
            )
        opts = FinalRequestOptions.construct(
            method="patch",
            url=path,
            json_data=body,
            content=content,
            files=await async_to_httpx_files(files),
            **options,
        )
        return await self.request(cast_to, opts)

    async def put(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        body: Body | None = None,
        content: AsyncBinaryTypes | None = None,
        files: RequestFiles | None = None,
        options: RequestOptions = {},
    ) -> ResponseT:
        if body is not None and content is not None:
            raise TypeError("Passing both `body` and `content` is not supported")
        if files is not None and content is not None:
            raise TypeError("Passing both `files` and `content` is not supported")
        if isinstance(body, bytes):
            warnings.warn(
                "Passing raw bytes as `body` is deprecated and will be removed in a future version. "
                "Please pass raw bytes via the `content` parameter instead.",
                DeprecationWarning,
                stacklevel=2,
            )
        opts = FinalRequestOptions.construct(
            method="put", url=path, json_data=body, content=content, files=await async_to_httpx_files(files), **options
        )
        return await self.request(cast_to, opts)

    async def delete(
        self,
        path: str,
        *,
        cast_to: Type[ResponseT],
        body: Body | None = None,
        content: AsyncBinaryTypes | None = None,
        options: RequestOptions = {},
    ) -> ResponseT:
        if body is not None and content is not None:
            raise TypeError("Passing both `body` and `content` is not supported")
        if isinstance(body, bytes):
            warnings.warn(
                "Passing raw bytes as `body` is deprecated and will be removed in a future version. "
                "Please pass raw bytes via the `content` parameter instead.",
                DeprecationWarning,
                stacklevel=2,
            )
        opts = FinalRequestOptions.construct(method="delete", url=path, json_data=body, content=content, **options)
        return await self.request(cast_to, opts)

    def get_api_list(
        self,
        path: str,
        *,
        model: Type[_T],
        page: Type[AsyncPageT],
        body: Body | None = None,
        options: RequestOptions = {},
        method: str = "get",
    ) -> AsyncPaginator[_T, AsyncPageT]:
        opts = FinalRequestOptions.construct(method=method, url=path, json_data=body, **options)
        return self._request_api_list(model, page, opts)


def make_request_options(
    *,
    query: Query | None = None,
    extra_headers: Headers | None = None,
    extra_query: Query | None = None,
    extra_body: Body | None = None,
    idempotency_key: str | None = None,
    timeout: float | httpx2.Timeout | None | NotGiven = not_given,
    post_parser: PostParser | NotGiven = not_given,
    security: SecurityOptions | None = None,
    synthesize_event_and_data: bool | None = None,
) -> RequestOptions:
    """Create a dict of type RequestOptions without keys of NotGiven values."""
    options: RequestOptions = {}
    if extra_headers is not None:
        options["headers"] = extra_headers

    if extra_body is not None:
        options["extra_json"] = cast(AnyMapping, extra_body)

    if query is not None:
        options["params"] = query

    if extra_query is not None:
        options["params"] = {**options.get("params", {}), **extra_query}

    if not isinstance(timeout, NotGiven):
        options["timeout"] = timeout

    if idempotency_key is not None:
        options["idempotency_key"] = idempotency_key

    if is_given(post_parser):
        # internal
        options["post_parser"] = post_parser  # type: ignore

    if security is not None:
        options["security"] = security

    if synthesize_event_and_data is not None:
        options["synthesize_event_and_data"] = synthesize_event_and_data

    return options


class ForceMultipartDict(Dict[str, None]):
    def __bool__(self) -> bool:
        return True


class OtherPlatform:
    def __init__(self, name: str) -> None:
        self.name = name

    @override
    def __str__(self) -> str:
        return f"Other:{self.name}"


Platform = Union[
    OtherPlatform,
    Literal[
        "MacOS",
        "Linux",
        "Windows",
        "FreeBSD",
        "OpenBSD",
        "iOS",
        "Android",
        "Unknown",
    ],
]


def get_platform() -> Platform:
    try:
        system = platform.system().lower()
        platform_name = platform.platform().lower()
    except Exception:
        return "Unknown"

    if "iphone" in platform_name or "ipad" in platform_name:
        # Tested using Python3IDE on an iPhone 11 and Pythonista on an iPad 7
        # system is Darwin and platform_name is a string like:
        # - Darwin-21.6.0-iPhone12,1-64bit
        # - Darwin-21.6.0-iPad7,11-64bit
        return "iOS"

    if system == "darwin":
        return "MacOS"

    if system == "windows":
        return "Windows"

    if "android" in platform_name:
        # Tested using Pydroid 3
        # system is Linux and platform_name is a string like 'Linux-5.10.81-android12-9-00001-geba40aecb3b7-ab8534902-aarch64-with-libc'
        return "Android"

    if system == "freebsd":
        return "FreeBSD"

    if system == "openbsd":
        return "OpenBSD"

    if system == "linux":
        try:
            distro_id = platform.freedesktop_os_release().get("ID", "").lower()
        except OSError:
            # This diagnostic header does not require an os-release file.
            distro_id = ""
        if distro_id == "freebsd":
            return "FreeBSD"

        if distro_id == "openbsd":
            return "OpenBSD"

        return "Linux"

    if platform_name:
        return OtherPlatform(platform_name)

    return "Unknown"


@lru_cache(maxsize=None)
def platform_headers(version: str, *, platform: Platform | None) -> Dict[str, str]:
    return {
        "X-Stainless-Lang": "python",
        "X-Stainless-Package-Version": version,
        "X-Stainless-OS": str(platform or get_platform()),
        "X-Stainless-Arch": str(get_architecture()),
        "X-Stainless-Runtime": get_python_runtime(),
        "X-Stainless-Runtime-Version": get_python_version(),
    }


class OtherArch:
    def __init__(self, name: str) -> None:
        self.name = name

    @override
    def __str__(self) -> str:
        return f"other:{self.name}"


Arch = Union[OtherArch, Literal["x32", "x64", "arm", "arm64", "unknown"]]


def get_python_runtime() -> str:
    try:
        return platform.python_implementation()
    except Exception:
        return "unknown"


def get_python_version() -> str:
    try:
        return platform.python_version()
    except Exception:
        return "unknown"


def get_architecture() -> Arch:
    try:
        machine = platform.machine().lower()
    except Exception:
        return "unknown"

    if machine in ("arm64", "aarch64"):
        return "arm64"

    # TODO: untested
    if machine == "arm":
        return "arm"

    if machine == "x86_64":
        return "x64"

    # TODO: untested
    if sys.maxsize <= 2**32:
        return "x32"

    if machine:
        return OtherArch(machine)

    return "unknown"


def _merge_mappings(
    obj1: Mapping[_T_co, Union[_T, Omit]],
    obj2: Mapping[_T_co, Union[_T, Omit]],
) -> Dict[_T_co, _T]:
    """Merge two mappings of the same type, removing any values that are instances of `Omit`.

    In cases with duplicate keys the second mapping takes precedence.
    """
    merged = {**obj1, **obj2}
    return {key: value for key, value in merged.items() if not isinstance(value, Omit)}