Response.output_text crashes when output_text.content.text is None

[casabre]

Confirm this is an issue with the Python library and not an underlying OpenAI API

• This is an issue with the Python library

Describe the bug

Accessing Response.output_text can raise a TypeError when one or more output_text content blocks contain text=None.

This occurs with the Responses API in openai-python and causes downstream frameworks (e.g. LangChain / LangGraph streaming agents) to crash during normal execution.

The failure violates the implied contract of output_text being a safe, aggregated string representation of model output.

Observed error:

TypeError: sequence item 0: expected str instance, NoneType found

Traceback (most recent call last):
  File "agent/agent.py", line 255, in stream
    async for item in self.graph.astream(...)

  File "langgraph/pregel/main.py", line 2971, in astream
    async for _ in runner.atick(...)

  File "langgraph/pregel/_runner.py", line 304, in atick
    await arun_with_retry(...)

  File "langgraph/pregel/_retry.py", line 137, in arun_with_retry
    return await task.proc.ainvoke(task.input, config)

  File "langgraph/_internal/_runnable.py", line 705, in ainvoke
    input = await asyncio.create_task(...)

  File "langgraph/_internal/_runnable.py", line 473, in ainvoke
    ret = await self.afunc(*args, **kwargs)

  File "langchain/agents/factory.py", line 1188, in amodel_node
    response = await awrap_model_call_handler(...)

  File "langchain/agents/factory.py", line 277, in final_normalized
    final_result = await result(request, handler)

  File "langchain/agents/factory.py", line 261, in composed
    outer_result = await outer(request, inner_handler)

  File "langchain/agents/middleware/todo.py", line 228, in awrap_model_call
    return await handler(request.override(...))

  File "langchain/agents/factory.py", line 257, in inner_handler
    inner_result = await inner(req, handler)

  File "deepagents/middleware/filesystem.py", line 975, in awrap_model_call
    return await handler(request)

  File "deepagents/middleware/subagents.py", line 490, in awrap_model_call
    return await handler(request.override(...))

  File "langchain_anthropic/middleware/prompt_caching.py", line 140, in awrap_model_call
    return await handler(request)

  File "langchain/agents/factory.py", line 1156, in _execute_model_async
    output = await model_.ainvoke(messages)

  File "langchain_core/runnables/base.py", line 5570, in ainvoke
    return await self.bound.ainvoke(...)

  File "langchain_core/language_models/chat_models.py", line 425, in ainvoke
    llm_result = await self.agenerate_prompt(...)

  File "langchain_core/language_models/chat_models.py", line 1132, in agenerate_prompt
    return await self.agenerate(...)

  File "langchain_core/language_models/chat_models.py", line 1090, in agenerate
    raise exceptions[0]

  File "langchain_core/language_models/chat_models.py", line 1343, in _agenerate_with_cache
    result = await self._agenerate(...)

  File "langchain_openai/chat_models/base.py", line 1631, in _agenerate
    return _construct_lc_result_from_responses_api(...)

  File "langchain_openai/chat_models/base.py", line 4338, in _construct_lc_result_from_responses_api
    output_text = _get_output_text(response)

  File "langchain_openai/chat_models/base.py", line 4192, in _get_output_text
    if hasattr(response, "output_text"):

  File "openai/types/responses/response.py", line 316, in output_text
    return "".join(texts)

TypeError: sequence item 0: expected str instance, NoneType found

To Reproduce

1. Run the provided minimal async reproduction script using:

• langchain-openai with the Responses API
• a self-hosted gpt-oss-120b OpenAI-compatible model via litellm

2. Initialize a deep agent with:

• tool support
• in-memory checkpointing and store
• reasoning enabled

3. Stream a single agent turn via graph.astream(..., stream_mode="updates")
4. During streaming, response.output_text is accessed internally
5. Execution fails with:

TypeError: sequence item 0: expected str instance, NoneType found

Code snippets

"""Minimal reproduction script for initializing a deep agent with memory and reasoning."""

import asyncio
import os
import uuid

from deepagents import create_deep_agent
from dotenv import load_dotenv
from langchain.tools import tool
from langchain_core.messages import HumanMessage
from langchain_openai import ChatOpenAI
from langgraph.checkpoint.memory import MemorySaver
from langgraph.store.memory import InMemoryStore
from pydantic import SecretStr

load_dotenv()


@tool
def fun_fact_about_satellites() -> str:
    """Return a fun fact about satellites."""
    return "Satellites can be used to monitor environmental changes on Earth in real-time."


async def main() -> None:
    """Create a deep agent, attach in-memory checkpointing/store, and run a single turn."""
    api_key = os.getenv("OPENAI_API_KEY")
    if not api_key:
        raise ValueError("OPENAI_API_KEY environment variable is not set.")
    base_url = os.getenv("OPENAI_API_BASE_URL")
    model = ChatOpenAI(
        model="gpt-oss-120b",
        temperature=0,
        api_key=SecretStr(api_key),
        base_url=base_url,
        reasoning={
            "effort": "medium",
            "summary": "auto",
        },
    )

    graph = create_deep_agent(
        tools=[fun_fact_about_satellites],
        model=model,
        system_prompt="You are a helpful assistant that provides concise and accurate information.",
        checkpointer=MemorySaver(),
        store=InMemoryStore(),
    )

    thread_id = str(uuid.uuid4())
    prompt = "Give me a one-line fun fact about satellites."
    inputs = {"messages": [HumanMessage(content=prompt)]}

    async for chunk in graph.astream(
        input=inputs,
        config={"configurable": {"thread_id": thread_id}},
        stream_mode="updates",
    ):
        print(chunk)


if __name__ == "__main__":
    asyncio.run(main())

OS

Windows

Python version

Python 3.12.3

Library version

openai==2.14.0, langchain-openai==1.1.6, langchain==1.2.3, langgraph==1.0.5, pydantic==2.12.5, deepagents==0.3.4, python-dotenv==1.2.1

[RobertCraigie]

a self-hosted gpt-oss-120b OpenAI-compatible model via litellm

If they're returning null, then they are not OpenAI compatible. Can you please file an issue with the upstream server?

[casabre]

re-raised under litellm #19032