fix(client): treat empty OPENAI_BASE_URL env var as unset

src/openai/_client.py

@@ -248,7 +248,7 @@ def __init__(
         if provider_runtime is not None:
             base_url = provider_runtime.base_url
         elif base_url is None:
-            base_url = os.environ.get("OPENAI_BASE_URL")
+            base_url = os.environ.get("OPENAI_BASE_URL") or None
         if base_url is None:
             base_url = f"https://api.openai.com/v1"
 
@@ -844,7 +844,7 @@ def __init__(
         if provider_runtime is not None:
             base_url = provider_runtime.base_url
         elif base_url is None:
-            base_url = os.environ.get("OPENAI_BASE_URL")
+            base_url = os.environ.get("OPENAI_BASE_URL") or None
         if base_url is None:
             base_url = f"https://api.openai.com/v1"
 

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_input_item_param.py

@@ -193,10 +193,10 @@ class ImageGenerationCall(TypedDict, total=False):
     id: Required[str]
     """The unique ID of the image generation call."""
 
-    result: Required[Optional[str]]
+    result: Optional[str]
     """The generated image encoded in base64."""
 
-    status: Required[Literal["in_progress", "completed", "generating", "failed"]]
+    status: Literal["in_progress", "completed", "generating", "failed"]
     """The status of the image generation call."""
 
     type: Required[Literal["image_generation_call"]]

tests/lib/responses/test_responses.py

@@ -8,6 +8,9 @@
 
 from openai import OpenAI, AsyncOpenAI
 from openai._utils import assert_signatures_in_sync
+from openai.types.responses.response import Response
+from openai.types.responses.response_output_message import ResponseOutputMessage
+from openai.types.responses.response_output_text import ResponseOutputText
 
 from ...conftest import base_url
 from ..snapshots import make_snapshot_request
@@ -41,6 +44,77 @@ def test_output_text(client: OpenAI, respx_mock: MockRouter) -> None:
     )
 
 
+def _make_response_with_text(text: str | None) -> Response:
+    """Build a minimal Response via model_construct, bypassing validation to simulate
+    raw API payloads where `text` may be null."""
+    content_block = ResponseOutputText.model_construct(
+        type="output_text",
+        annotations=[],
+        logprobs=None,
+        text=text,
+    )
+    message = ResponseOutputMessage.model_construct(
+        id="msg_test",
+        type="message",
+        status="completed",
+        role="assistant",
+        content=[content_block],
+    )
+    return Response.model_construct(
+        id="resp_test",
+        object="response",
+        created_at=0,
+        status="completed",
+        model="gpt-4o-mini",
+        output=[message],
+        parallel_tool_calls=True,
+        text=None,
+        tool_choice="auto",
+        tools=[],
+        truncation="disabled",
+    )
+
+
+def test_output_text_null_guard() -> None:
+    """output_text must not crash when content items have text=None (issue #3063)."""
+    # Single null text block -> empty string
+    response = _make_response_with_text(None)
+    assert response.output_text == ""
+
+    # Valid text block -> the text is returned
+    response = _make_response_with_text("hello")
+    assert response.output_text == "hello"
+
+    # Mixed: null and valid in the same message -> only valid text joined
+    null_block = ResponseOutputText.model_construct(
+        type="output_text", annotations=[], logprobs=None, text=None
+    )
+    valid_block = ResponseOutputText.model_construct(
+        type="output_text", annotations=[], logprobs=None, text="world"
+    )
+    message = ResponseOutputMessage.model_construct(
+        id="msg_mixed",
+        type="message",
+        status="completed",
+        role="assistant",
+        content=[null_block, valid_block],
+    )
+    response = Response.model_construct(
+        id="resp_mixed",
+        object="response",
+        created_at=0,
+        status="completed",
+        model="gpt-4o-mini",
+        output=[message],
+        parallel_tool_calls=True,
+        text=None,
+        tool_choice="auto",
+        tools=[],
+        truncation="disabled",
+    )
+    assert response.output_text == "world"
+
+
 @pytest.mark.parametrize("sync", [True, False], ids=["sync", "async"])
 def test_stream_method_definition_in_sync(sync: bool, client: OpenAI, async_client: AsyncOpenAI) -> None:
     checking_client: OpenAI | AsyncOpenAI = client if sync else async_client