fix: tolerate None text in ResponseOutputText content items

[beveradb]

Summary

ItemHelpers.extract_text does text += content_item.text, which raises TypeError: can only concatenate str (not "NoneType") to str when a ResponseOutputText content item has text=None.

Although the Responses API schema declares text: str, this surfaces in practice when output items are assembled via model_construct during streaming or forwarded through provider gateways (e.g. LiteLLM partial responses). The crash propagates up from inside execute_tools_and_side_effects (called from run_single_turn_streamed), aborting the entire agent turn.

We hit this in production on an agent backed by a LiteLLM → Azure OpenAI Responses gateway: ~57 user-facing failures per day, all surfacing as the same unhelpful "Something went wrong" message. The full traceback ends in:

File ".../agents/run_internal/turn_resolution.py", line 618, in execute_tools_and_side_effects
    ItemHelpers.extract_text(message_items[-1].raw_item) if message_items else None
File ".../agents/items.py", line 686, in extract_text
    text += content_item.text
TypeError: can only concatenate str (not "NoneType") to str

Fix: coerce with or "" so a missing .text contributes nothing instead of crashing. Behavior on well-formed messages is unchanged (covered by the existing test_extract_text_concatenates_all_text_segments).

Test plan

• uv run pytest tests/test_items_helpers.py — all 26 tests pass, including the new regression test.
• New test test_extract_text_tolerates_none_text_content covers:
  • single ResponseOutputText with text=None → returns None (per return text or None)
  • mixed content with None text + real text → returns the real text, both orderings.
• make lint and make format clean.

Issue number

N/A — opening a fresh report.

Checks

• I've added new tests (if relevant)
• I've added/updated the relevant documentation — none needed; method behavior on the documented happy path is unchanged.
• I've run make lint and make format
• I've made sure tests pass

[seratch]

@codex review again

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Breezy!

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".