fix: handle null text in output_text items

src/openai/_base_client.py

@@ -833,6 +833,9 @@ def _idempotency_key(self) -> str:
 
 class _DefaultHttpxClient(httpx.Client):
     def __init__(self, **kwargs: Any) -> None:
+        from ._utils import sanitize_proxy_env_vars
+
+        sanitize_proxy_env_vars()
         kwargs.setdefault("timeout", DEFAULT_TIMEOUT)
         kwargs.setdefault("limits", DEFAULT_CONNECTION_LIMITS)
         kwargs.setdefault("follow_redirects", True)
@@ -1420,6 +1423,9 @@ def get_api_list(
 
 class _DefaultAsyncHttpxClient(httpx.AsyncClient):
     def __init__(self, **kwargs: Any) -> None:
+        from ._utils import sanitize_proxy_env_vars
+
+        sanitize_proxy_env_vars()
         kwargs.setdefault("timeout", DEFAULT_TIMEOUT)
         kwargs.setdefault("limits", DEFAULT_CONNECTION_LIMITS)
         kwargs.setdefault("follow_redirects", True)
@@ -1437,6 +1443,9 @@ def __init__(self, **_kwargs: Any) -> None:
 
     class _DefaultAioHttpClient(httpx_aiohttp.HttpxAiohttpClient):  # type: ignore
         def __init__(self, **kwargs: Any) -> None:
+            from ._utils import sanitize_proxy_env_vars
+
+            sanitize_proxy_env_vars()
             kwargs.setdefault("timeout", DEFAULT_TIMEOUT)
             kwargs.setdefault("limits", DEFAULT_CONNECTION_LIMITS)
             kwargs.setdefault("follow_redirects", True)

src/openai/_models.py

@@ -657,8 +657,11 @@ def construct_type(*, value: object, type_: object, metadata: Optional[List[Any]
         if not is_mapping(value):
             return value
 
-        _, items_type = get_args(type_)  # Dict[_, items_type]
-        return {key: construct_type(value=item, type_=items_type) for key, item in value.items()}
+        dict_args = get_args(type_)
+        if len(dict_args) >= 2:
+            items_type = dict_args[1]
+            return {key: construct_type(value=item, type_=items_type) for key, item in value.items()}
+        return dict(value)
 
     if (
         not is_literal_type(type_)
@@ -678,8 +681,10 @@ def construct_type(*, value: object, type_: object, metadata: Optional[List[Any]
         if not is_list(value):
             return value
 
-        inner_type = args[0]  # List[inner_type]
-        return [construct_type(value=entry, type_=inner_type) for entry in value]
+        if args:
+            inner_type = args[0]  # List[inner_type]
+            return [construct_type(value=entry, type_=inner_type) for entry in value]
+        return list(value)
 
     if origin == float:
         if isinstance(value, int):

src/openai/_streaming.py

@@ -63,11 +63,10 @@ def __stream__(self) -> Iterator[_T]:
                 if sse.data.startswith("[DONE]"):
                     break
 
-                # we have to special case the Assistants `thread.` events since we won't have an "event" key in the data
-                if sse.event and sse.event.startswith("thread."):
+                # Handle error events first - these can occur outside of thread.* events
+                if sse.event == "error":
                     data = sse.json()
-
-                    if sse.event == "error" and is_mapping(data) and data.get("error"):
+                    if is_mapping(data) and data.get("error"):
                         message = None
                         error = data.get("error")
                         if is_mapping(error):
@@ -80,7 +79,16 @@ def __stream__(self) -> Iterator[_T]:
                             request=self.response.request,
                             body=data["error"],
                         )
+                    # If error event doesn't have expected structure, still raise
+                    raise APIError(
+                        message="An error occurred during streaming",
+                        request=self.response.request,
+                        body=data,
+                    )
 
+                # we have to special case the Assistants `thread.` events since we won't have an "event" key in the data
+                if sse.event and sse.event.startswith("thread."):
+                    data = sse.json()
                     yield process_data(data={"data": data, "event": sse.event}, cast_to=cast_to, response=response)
                 else:
                     data = sse.json()
@@ -173,11 +181,10 @@ async def __stream__(self) -> AsyncIterator[_T]:
                 if sse.data.startswith("[DONE]"):
                     break
 
-                # we have to special case the Assistants `thread.` events since we won't have an "event" key in the data
-                if sse.event and sse.event.startswith("thread."):
+                # Handle error events first - these can occur outside of thread.* events
+                if sse.event == "error":
                     data = sse.json()
-
-                    if sse.event == "error" and is_mapping(data) and data.get("error"):
+                    if is_mapping(data) and data.get("error"):
                         message = None
                         error = data.get("error")
                         if is_mapping(error):
@@ -190,7 +197,16 @@ async def __stream__(self) -> AsyncIterator[_T]:
                             request=self.response.request,
                             body=data["error"],
                         )
+                    # If error event doesn't have expected structure, still raise
+                    raise APIError(
+                        message="An error occurred during streaming",
+                        request=self.response.request,
+                        body=data,
+                    )
 
+                # we have to special case the Assistants `thread.` events since we won't have an "event" key in the data
+                if sse.event and sse.event.startswith("thread."):
+                    data = sse.json()
                     yield process_data(data={"data": data, "event": sse.event}, cast_to=cast_to, response=response)
                 else:
                     data = sse.json()

src/openai/_utils/__init__.py

@@ -29,6 +29,7 @@
     get_async_library as get_async_library,
     maybe_coerce_float as maybe_coerce_float,
     get_required_header as get_required_header,
+    sanitize_proxy_env_vars as sanitize_proxy_env_vars,
     maybe_coerce_boolean as maybe_coerce_boolean,
     maybe_coerce_integer as maybe_coerce_integer,
     is_async_azure_client as is_async_azure_client,

src/openai/_utils/_transform.py

@@ -180,8 +180,11 @@ def _transform_recursive(
         return _transform_typeddict(data, stripped_type)
 
     if origin == dict and is_mapping(data):
-        items_type = get_args(stripped_type)[1]
-        return {key: _transform_recursive(value, annotation=items_type) for key, value in data.items()}
+        args = get_args(stripped_type)
+        if len(args) >= 2:
+            items_type = args[1]
+            return {key: _transform_recursive(value, annotation=items_type) for key, value in data.items()}
+        return dict(data)
 
     if (
         # List[T]
@@ -196,6 +199,12 @@ def _transform_recursive(
         if isinstance(data, dict):
             return cast(object, data)
 
+        # bare list/iterable/sequence with no type args - return as-is
+        if not get_args(stripped_type):
+            if is_list(data):
+                return data
+            return list(data)
+
         inner_type = extract_type_arg(stripped_type, 0)
         if _no_transform_needed(inner_type):
             # for some types there is no need to transform anything, so we can get a small
@@ -346,8 +355,11 @@ async def _async_transform_recursive(
         return await _async_transform_typeddict(data, stripped_type)
 
     if origin == dict and is_mapping(data):
-        items_type = get_args(stripped_type)[1]
-        return {key: _transform_recursive(value, annotation=items_type) for key, value in data.items()}
+        args = get_args(stripped_type)
+        if len(args) >= 2:
+            items_type = args[1]
+            return {key: _transform_recursive(value, annotation=items_type) for key, value in data.items()}
+        return dict(data)
 
     if (
         # List[T]
@@ -362,6 +374,12 @@ async def _async_transform_recursive(
         if isinstance(data, dict):
             return cast(object, data)
 
+        # bare list/iterable/sequence with no type args - return as-is
+        if not get_args(stripped_type):
+            if is_list(data):
+                return data
+            return list(data)
+
         inner_type = extract_type_arg(stripped_type, 0)
         if _no_transform_needed(inner_type):
             # for some types there is no need to transform anything, so we can get a small

src/openai/_utils/_utils.py

@@ -447,3 +447,32 @@ def is_async_azure_client(client: object) -> TypeGuard[AsyncAzureOpenAI]:
     from ..lib.azure import AsyncAzureOpenAI
 
     return isinstance(client, AsyncAzureOpenAI)
+
+
+_PROXY_ENV_VARS = (
+    "HTTP_PROXY",
+    "http_proxy",
+    "HTTPS_PROXY",
+    "https_proxy",
+    "ALL_PROXY",
+    "all_proxy",
+    "NO_PROXY",
+    "no_proxy",
+)
+
+
+def sanitize_proxy_env_vars() -> None:
+    """Sanitize proxy-related environment variables by removing newline characters.
+
+    This works around an issue where httpx's proxy parsing only splits by comma
+    and fails with InvalidURL when newlines are present in the value.
+    """
+    for var in _PROXY_ENV_VARS:
+        value = os.environ.get(var)
+        if value and ("\n" in value or "\r" in value):
+            sanitized = ",".join(
+                part.strip()
+                for part in value.replace("\n", ",").replace("\r", ",").split(",")
+                if part.strip()
+            )
+            os.environ[var] = sanitized

src/openai/lib/_parsing/_responses.py

@@ -58,7 +58,7 @@ def parse_response(
 ) -> ParsedResponse[TextFormatT]:
     output_list: List[ParsedResponseOutputItem[TextFormatT]] = []
 
-    for output in response.output:
+    for output in (response.output or []):
         if output.type == "message":
             content_list: List[ParsedContent[TextFormatT]] = []
             for item in output.content:
@@ -71,7 +71,7 @@ def parse_response(
                         type_=ParsedResponseOutputText[TextFormatT],
                         value={
                             **item.to_dict(),
-                            "parsed": parse_text(item.text, text_format=text_format),
+                            "parsed": parse_text(item.text, text_format=text_format) if item.text is not None else None,
                         },
                     )
                 )
@@ -140,7 +140,7 @@ def parse_response(
 
 
 def parse_text(text: str, text_format: type[TextFormatT] | Omit) -> TextFormatT | None:
-    if not is_given(text_format):
+    if not is_given(text_format) or text_format is None:
         return None
 
     if is_basemodel_type(text_format):

src/openai/types/responses/response.py

@@ -447,7 +447,7 @@ def output_text(self) -> str:
         for output in self.output:
             if output.type == "message":
                 for content in output.content:
-                    if content.type == "output_text":
+                    if content.type == "output_text" and content.text is not None:
                         texts.append(content.text)
 
         return "".join(texts)

src/openai/types/responses/response_function_web_search.py

@@ -19,11 +19,15 @@
 class ActionSearchSource(BaseModel):
     """A source used in the search."""
 
-    type: Literal["url"]
-    """The type of source. Always `url`."""
+    type: Literal["url", "api"]
+    """The type of source. `url` for a web page, `api` for a built-in OpenAI data source."""
 
-    url: str
-    """The URL of the source."""
+    url: Optional[str] = None
+    """The URL of the source. Present when type is `url`."""
+
+    name: Optional[str] = None
+    """The name of the built-in data source (e.g. `oai-weather`, `oai-sports`, `oai-finance`).
+    Present when type is `api`."""
 
 
 class ActionSearch(BaseModel):
@@ -78,7 +82,7 @@ class ResponseFunctionWebSearch(BaseModel):
     id: str
     """The unique ID of the web search tool call."""
 
-    action: Action
+    action: Optional[Action] = None
     """
     An object describing the specific action taken in this web search call. Includes
     details on how the model used the web (search, open_page, find_in_page).

src/openai/types/responses/response_function_web_search_param.py

@@ -20,11 +20,15 @@
 class ActionSearchSource(TypedDict, total=False):
     """A source used in the search."""
 
-    type: Required[Literal["url"]]
-    """The type of source. Always `url`."""
+    type: Required[Literal["url", "api"]]
+    """The type of source. `url` for a web page, `api` for a built-in OpenAI data source."""
 
-    url: Required[str]
-    """The URL of the source."""
+    url: str
+    """The URL of the source. Required when type is `url`."""
+
+    name: str
+    """The name of the built-in data source (e.g. `oai-weather`, `oai-sports`, `oai-finance`).
+    Required when type is `api`."""
 
 
 class ActionSearch(TypedDict, total=False):