perf(lib): stop pulling openai.types.beta into every import openai

[aviseth]

• I understand that this repository is auto-generated and my pull request may not be merged

Changes being requested

openai/__init__.py imports openai.lib.streaming, and openai/lib/streaming/_assistants.py imported openai.types.beta at module scope. That loaded 318 modules on every import openai — about 30% of the total import cost — so that anyone importing the package paid for the Assistants API whether or not they used it.

_assistants.py already has from __future__ import annotations, so all but two of those names were needed only by type checkers. This moves them under TYPE_CHECKING and imports the two runtime exceptions locally: RunStep, used by construct_type(type_=RunStep, ...) in accumulate_run_step, and MessageContent, used twice in accumulate_event.

Measured from a clean checkout with the repo venv, 25 cold runs each on Python 3.13:

	sys.modules	openai.types.beta	wall, median
main	1234	318	359.2 ms
this branch	916	0	256.7 ms

1.40x faster, −102 ms, −318 modules.

There is no public API change. openai.AssistantEventHandler and openai.AsyncAssistantEventHandler remain eagerly exported from openai/__init__.py, from openai import * is unchanged, and import openai.lib.streaming still works. This deliberately avoids a module-level __getattr__ on the package.

The whole change is confined to src/openai/lib/, which CONTRIBUTING.md says the generator never modifies and which carries no File generated from our OpenAPI spec header. It therefore does not touch generated output, does not consume Castiron custom-code budget, and cannot conflict with a regen — which is the concern that closed #2950.

tests/lib/test_streaming_lazy_types.py asserts in a fresh interpreter that import openai loads no openai.types.beta module, and that both handler names are still eagerly exported. It fails on main and passes here.

./scripts/lint is clean (ruff, pyright strict, mypy 1573 files) and ./scripts/test passes both lanes: 9127 passed / 144 skipped on Pydantic v2, 9113 passed / 158 skipped on Pydantic v1.

Additional context & links

Fixes the bulk of #2819. Supersedes the approach in #2950, which deferred openai.lib itself — that is only 0.9% of import cost, which is why it measured ~4%. The cost was in what one lib module pulled in behind it, not in lib.

The remaining openai.types.* non-beta cost (~144 ms across 434 modules) is reachable from openai/__init__.py directly and would need a change to generated code, so it is out of scope here.