feat: v0.1.2 api update

CONTRIBUTING.md

@@ -17,8 +17,7 @@ $ rye sync --all-features
 You can then run scripts using `rye run python script.py` or by activating the virtual environment:
 
 ```sh
-$ rye shell
-# or manually activate - https://docs.python.org/3/library/venv.html#how-venvs-work
+# Activate the virtual environment - https://docs.python.org/3/library/venv.html#how-venvs-work
 $ source .venv/bin/activate
 
 # now you can omit the `rye run` prefix

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "llama_api_client"
-version = "0.1.1"
+version = "0.1.2"
 description = "The official Python library for the llama-api-client API"
 dynamic = ["readme"]
 license = "MIT"
@@ -37,6 +37,8 @@ classifiers = [
 Homepage = "https://github.com/meta-llama/llama-api-python"
 Repository = "https://github.com/meta-llama/llama-api-python"
 
+[project.optional-dependencies]
+aiohttp = ["aiohttp", "httpx_aiohttp>=0.1.6"]
 
 [tool.rye]
 managed = true
@@ -54,6 +56,7 @@ dev-dependencies = [
     "importlib-metadata>=6.7.0",
     "rich>=13.7.1",
     "nest_asyncio==1.6.0",
+    "pytest-xdist>=3.6.1",
 ]
 
 [tool.rye.scripts]
@@ -125,7 +128,7 @@ replacement = '[\1](https://github.com/meta-llama/llama-api-python/tree/main/\g<
 
 [tool.pytest.ini_options]
 testpaths = ["tests"]
-addopts = "--tb=short"
+addopts = "--tb=short -n auto"
 xfail_strict = true
 asyncio_mode = "auto"
 asyncio_default_fixture_loop_scope = "session"

requirements-dev.lock

@@ -10,13 +10,24 @@
 #   universal: false
 
 -e file:.
+aiohappyeyeballs==2.6.1
+    # via aiohttp
+aiohttp==3.12.8
+    # via httpx-aiohttp
+    # via llama-api-client
+aiosignal==1.3.2
+    # via aiohttp
 annotated-types==0.6.0
     # via pydantic
 anyio==4.4.0
     # via httpx
     # via llama-api-client
 argcomplete==3.1.2
     # via nox
+async-timeout==5.0.1
+    # via aiohttp
+attrs==25.3.0
+    # via aiohttp
 certifi==2023.7.22
     # via httpcore
     # via httpx
@@ -30,25 +41,37 @@ distro==1.8.0
 exceptiongroup==1.2.2
     # via anyio
     # via pytest
+execnet==2.1.1
+    # via pytest-xdist
 filelock==3.12.4
     # via virtualenv
+frozenlist==1.6.2
+    # via aiohttp
+    # via aiosignal
 h11==0.14.0
     # via httpcore
 httpcore==1.0.2
     # via httpx
 httpx==0.28.1
+    # via httpx-aiohttp
     # via llama-api-client
     # via respx
+httpx-aiohttp==0.1.6
+    # via llama-api-client
 idna==3.4
     # via anyio
     # via httpx
+    # via yarl
 importlib-metadata==7.0.0
 iniconfig==2.0.0
     # via pytest
 markdown-it-py==3.0.0
     # via rich
 mdurl==0.1.2
     # via markdown-it-py
+multidict==6.4.4
+    # via aiohttp
+    # via yarl
 mypy==1.14.1
 mypy-extensions==1.0.0
     # via mypy
@@ -63,6 +86,9 @@ platformdirs==3.11.0
     # via virtualenv
 pluggy==1.5.0
     # via pytest
+propcache==0.3.1
+    # via aiohttp
+    # via yarl
 pydantic==2.10.3
     # via llama-api-client
 pydantic-core==2.27.1
@@ -72,7 +98,9 @@ pygments==2.18.0
 pyright==1.1.399
 pytest==8.3.3
     # via pytest-asyncio
+    # via pytest-xdist
 pytest-asyncio==0.24.0
+pytest-xdist==3.7.0
 python-dateutil==2.8.2
     # via time-machine
 pytz==2023.3.post1
@@ -94,11 +122,14 @@ tomli==2.0.2
 typing-extensions==4.12.2
     # via anyio
     # via llama-api-client
+    # via multidict
     # via mypy
     # via pydantic
     # via pydantic-core
     # via pyright
 virtualenv==20.24.5
     # via nox
+yarl==1.20.0
+    # via aiohttp
 zipp==3.17.0
     # via importlib-metadata

requirements.lock

@@ -10,27 +10,51 @@
 #   universal: false
 
 -e file:.
+aiohappyeyeballs==2.6.1
+    # via aiohttp
+aiohttp==3.12.8
+    # via httpx-aiohttp
+    # via llama-api-client
+aiosignal==1.3.2
+    # via aiohttp
 annotated-types==0.6.0
     # via pydantic
 anyio==4.4.0
     # via httpx
     # via llama-api-client
+async-timeout==5.0.1
+    # via aiohttp
+attrs==25.3.0
+    # via aiohttp
 certifi==2023.7.22
     # via httpcore
     # via httpx
 distro==1.8.0
     # via llama-api-client
 exceptiongroup==1.2.2
     # via anyio
+frozenlist==1.6.2
+    # via aiohttp
+    # via aiosignal
 h11==0.14.0
     # via httpcore
 httpcore==1.0.2
     # via httpx
 httpx==0.28.1
+    # via httpx-aiohttp
+    # via llama-api-client
+httpx-aiohttp==0.1.6
     # via llama-api-client
 idna==3.4
     # via anyio
     # via httpx
+    # via yarl
+multidict==6.4.4
+    # via aiohttp
+    # via yarl
+propcache==0.3.1
+    # via aiohttp
+    # via yarl
 pydantic==2.10.3
     # via llama-api-client
 pydantic-core==2.27.1
@@ -41,5 +65,8 @@ sniffio==1.3.0
 typing-extensions==4.12.2
     # via anyio
     # via llama-api-client
+    # via multidict
     # via pydantic
     # via pydantic-core
+yarl==1.20.0
+    # via aiohttp

src/llama_api_client/__init__.py

@@ -36,7 +36,7 @@
     UnprocessableEntityError,
     APIResponseValidationError,
 )
-from ._base_client import DefaultHttpxClient, DefaultAsyncHttpxClient
+from ._base_client import DefaultHttpxClient, DefaultAioHttpClient, DefaultAsyncHttpxClient
 from ._utils._logs import setup_logging as _setup_logging
 
 __all__ = [
@@ -78,6 +78,7 @@
     "DEFAULT_CONNECTION_LIMITS",
     "DefaultHttpxClient",
     "DefaultAsyncHttpxClient",
+    "DefaultAioHttpClient",
 ]
 
 if not _t.TYPE_CHECKING:

src/llama_api_client/_base_client.py

@@ -960,6 +960,9 @@ def request(
             if self.custom_auth is not None:
                 kwargs["auth"] = self.custom_auth
 
+            if options.follow_redirects is not None:
+                kwargs["follow_redirects"] = options.follow_redirects
+
             log.debug("Sending HTTP Request: %s %s", request.method, request.url)
 
             response = None
@@ -1068,7 +1071,14 @@ def _process_response(
     ) -> ResponseT:
         origin = get_origin(cast_to) or cast_to
 
-        if inspect.isclass(origin) and issubclass(origin, BaseAPIResponse):
+        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}")
 
@@ -1279,6 +1289,24 @@ def __init__(self, **kwargs: Any) -> None:
         super().__init__(**kwargs)
 
 
+try:
+    import httpx_aiohttp
+except ImportError:
+
+    class _DefaultAioHttpClient(httpx.AsyncClient):
+        def __init__(self, **_kwargs: Any) -> None:
+            raise RuntimeError("To use the aiohttp client you must have installed the package with the `aiohttp` extra")
+else:
+
+    class _DefaultAioHttpClient(httpx_aiohttp.HttpxAiohttpClient):  # type: ignore
+        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:
     DefaultAsyncHttpxClient = httpx.AsyncClient
     """An alias to `httpx.AsyncClient` that provides the same defaults that this SDK
@@ -1287,8 +1315,12 @@ def __init__(self, **kwargs: Any) -> None:
     This is useful because overriding the `http_client` with your own instance of
     `httpx.AsyncClient` will result in httpx's defaults being used, not ours.
     """
+
+    DefaultAioHttpClient = httpx.AsyncClient
+    """An alias to `httpx.AsyncClient` that changes the default HTTP transport to `aiohttp`."""
 else:
     DefaultAsyncHttpxClient = _DefaultAsyncHttpxClient
+    DefaultAioHttpClient = _DefaultAioHttpClient
 
 
 class AsyncHttpxClientWrapper(DefaultAsyncHttpxClient):
@@ -1460,6 +1492,9 @@ async def request(
             if self.custom_auth is not None:
                 kwargs["auth"] = self.custom_auth
 
+            if options.follow_redirects is not None:
+                kwargs["follow_redirects"] = options.follow_redirects
+
             log.debug("Sending HTTP Request: %s %s", request.method, request.url)
 
             response = None
@@ -1568,7 +1603,14 @@ async def _process_response(
     ) -> ResponseT:
         origin = get_origin(cast_to) or cast_to
 
-        if inspect.isclass(origin) and issubclass(origin, BaseAPIResponse):
+        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}")
 

src/llama_api_client/_models.py

@@ -737,6 +737,7 @@ class FinalRequestOptionsInput(TypedDict, total=False):
     idempotency_key: str
     json_data: Body
     extra_json: AnyMapping
+    follow_redirects: bool
 
 
 @final
@@ -750,6 +751,7 @@ class FinalRequestOptions(pydantic.BaseModel):
     files: Union[HttpxRequestFiles, None] = None
     idempotency_key: Union[str, None] = None
     post_parser: Union[Callable[[Any], Any], NotGiven] = NotGiven()
+    follow_redirects: Union[bool, None] = None
 
     # It should be noted that we cannot use `json` here as that would override
     # a BaseModel method in an incompatible fashion.

src/llama_api_client/_types.py

@@ -100,6 +100,7 @@ class RequestOptions(TypedDict, total=False):
     params: Query
     extra_json: AnyMapping
     idempotency_key: str
+    follow_redirects: bool
 
 
 # Sentinel class used until PEP 0661 is accepted
@@ -215,3 +216,4 @@ class _GenericAlias(Protocol):
 
 class HttpxSendArgs(TypedDict, total=False):
     auth: httpx.Auth
+    follow_redirects: bool

src/llama_api_client/_version.py

@@ -1,4 +1,4 @@
 # File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
 
 __title__ = "llama_api_client"
-__version__ = "0.1.1"
+__version__ = "0.1.2"

tests/api_resources/chat/test_completions.py

@@ -205,7 +205,9 @@ def test_streaming_response_create_overload_2(self, client: LlamaAPIClient) -> N
 
 
 class TestAsyncCompletions:
-    parametrize = pytest.mark.parametrize("async_client", [False, True], indirect=True, ids=["loose", "strict"])
+    parametrize = pytest.mark.parametrize(
+        "async_client", [False, True, {"http_client": "aiohttp"}], indirect=True, ids=["loose", "strict", "aiohttp"]
+    )
 
     @pytest.mark.skip()
     @parametrize

tests/api_resources/test_models.py

@@ -89,7 +89,9 @@ def test_streaming_response_list(self, client: LlamaAPIClient) -> None:
 
 
 class TestAsyncModels:
-    parametrize = pytest.mark.parametrize("async_client", [False, True], indirect=True, ids=["loose", "strict"])
+    parametrize = pytest.mark.parametrize(
+        "async_client", [False, True, {"http_client": "aiohttp"}], indirect=True, ids=["loose", "strict", "aiohttp"]
+    )
 
     @pytest.mark.skip()
     @parametrize

tests/api_resources/test_moderations.py

@@ -82,7 +82,9 @@ def test_streaming_response_create(self, client: LlamaAPIClient) -> None:
 
 
 class TestAsyncModerations:
-    parametrize = pytest.mark.parametrize("async_client", [False, True], indirect=True, ids=["loose", "strict"])
+    parametrize = pytest.mark.parametrize(
+        "async_client", [False, True, {"http_client": "aiohttp"}], indirect=True, ids=["loose", "strict", "aiohttp"]
+    )
 
     @pytest.mark.skip()
     @parametrize