[pull] main from openai:main - #142
Merged
Merged
Conversation
## Summary Refine documentation for the image 'background' parameter across the Images resource and related type definitions. These changes only update descriptions and wording; there are no changes to function signatures, types, or runtime behavior. ## Changes - Rewrite and clarify the 'background' parameter description for image generation and editing to state it must be one of 'transparent', 'opaque', or 'auto' and that 'auto' lets the model choose the background. - Document that transparent backgrounds are available for supported GPT Image models and explicitly mark gpt-image-2 and gpt-image-2-2026-04-21 support as preview-level, rather than stating transparent backgrounds are not supported. - Add/clarify guidance that when requesting a transparent background the returned image format should support transparency (e.g., 'png' or 'webp'). - Apply these documentation updates consistently in runtime resource docstrings and in typed models/typed dicts describing image parameters and responses. - No changes to public types, method signatures, parameter names, accepted values, or return types; no customer-visible behavior or API contract changes. Co-authored-by: apcha-oai <228803254+apcha-oai@users.noreply.github.com>
## Summary Set a generous 10,000-line ceiling for the Python SDK's remaining customization of generated files. The current verified patch is 6,558 lines (+6,010 / −548), so normal development has headroom while unusually large patches need an explicit decision. This PR intentionally changes only `.castiron-ratchet.json`. Keeping the policy separate means an SDK change cannot raise its own allowance. A human approving review is required; the automated technical review is not policy approval. Companion tooling PR #3715 adds enforcement using the budget already on `main`. Land this policy first. This seed alone does not enable a required check, and it does not change SDK behavior or CODEOWNERS.
## Summary A small SDK diff can hide a large patch against generated output. Add a custom-code gate that counts custom additions and deletions together, using the budget and checker already on `main`. Extend the existing two custom-code workflows and reuse their trusted report—no new workflows or compiler changes. Keep the report comment available when the gate fails, preserve default CODEOWNERS, and require a separate justified, human-reviewed PR for budget increases. Budget-only #3714 is merged; this PR is rebased onto `main`. Enable `Castiron / budget-only change` and `Castiron / custom-code budget` only after the tooling is on `main` and the rollout checks in `scripts/castiron/CUSTOM_CODE.md` pass, retaining the required merge queue. `AGENTS.md` and `CONTRIBUTING.md` contain the agent rules and contributor policy and link to those technical details. This PR does not activate repository rules or enable auto-merge.
- [x] I understand that this repository is auto-generated and my pull request may not be merged ## Changes being requested Move the existing sync/async vector-store file and file-batch polling bodies into SDK-owned `lib/_vector_stores.py`, sharing their identical `openai-poll-after-ms` lookup, integer conversion, and one-second fallback. The public `poll` methods remain thin delegates with the same signatures, defaults, docstrings, and return types. Request headers, response parsing, interval handling, terminal conditions, return identity, and error/cancellation propagation are preserved. Upload orchestration, create-and-poll methods, other requests, and response wrappers are untouched. This adds no generic polling framework, timeout behavior, schema/compiler changes, or generation-metadata changes. Review pointers: - Original public polling bodies: [Files](https://github.com/openai/openai-python/blob/bedb9a7b8839e193107e88b92f7cc166f08ac83d/src/openai/resources/vector_stores/files.py#L360) and [FileBatches](https://github.com/openai/openai-python/blob/bedb9a7b8839e193107e88b92f7cc166f08ac83d/src/openai/resources/vector_stores/file_batches.py#L331), with the corresponding async methods in the same files. - [SDK-owned helper](https://github.com/openai/openai-python/blob/25e95b31c1402b3a927cedb9fff5d3d7a785bbeb/src/openai/lib/_vector_stores.py#L16): the four bodies only rename their receiver and replace the repeated interval block with the shared lookup. Source/AST comparison against the base verifies those edits and the unchanged remainder of both resources. - [Focused tests](https://github.com/openai/openai-python/blob/25e95b31c1402b3a927cedb9fff5d3d7a785bbeb/tests/lib/test_vector_store_polling.py#L47): 106 lines / 22 cases covering the four public entry points, terminal results, default/server/explicit intervals and headers, retrieval errors, and async cancellation. Existing API and helper-signature tests are unchanged. ## Additional context & links Validation: - Commands: `.venv/bin/python -m pytest tests/lib/test_vector_store_polling.py -q -n 0` and the same command with `.venv-pydantic-v1/bin/python` passed all **22 cases against the original implementation and again after extraction**. - Commands: `TEST_API_BASE_URL=http://127.0.0.1:4142 .venv/bin/python -m pytest tests/api_resources/vector_stores tests/lib/test_vector_store_files.py tests/lib/test_vector_store_file_batches.py tests/lib/test_vector_store_polling.py -q -n 0` and the same command with `.venv-pydantic-v1/bin/python` passed **248 tests in each Pydantic mode**. - Commands: `./scripts/format` and `./scripts/lint` passed, including Ruff, Pyright, mypy, and import checks. Unrelated reporter-formatting edits were excluded. - Command: `./scripts/build` passed. Both distributions include the helper, and a fresh import from the built wheel loads the helper and all four resource classes. The verified custom-code report keeps **36 mixed files**, with only the two vector-store resource customizations changed: - `files.py`: **+222/-4 to +179/-1**. - `file_batches.py`: **+275/-4 to +242/-3**. All other 34 customizations, generation metadata, API reference, dependencies, and workflows are unchanged. Reproduction command: ```sh python3 scripts/castiron/custom_code_report.py report \ --base bedb9a7 \ --head 25e95b3 \ --fetch --require-head-hash --public \ --out /tmp/castiron-vector-polling ```
…3712) - [x] I understand that this repository is auto-generated and my pull request may not be merged ## Changes being requested Move the existing sync and async file-processing polling bodies into SDK-owned `lib/_files.py`. `Files.wait_for_processing` and `AsyncFiles.wait_for_processing` keep their public signatures, defaults, docstrings, and return types, and delegate to the helpers. No schema, compiler, dependency, or generation-metadata changes are needed. The polling bodies are unchanged apart from renaming the receiver from `self` to `files`. This preserves the `processed`/`error`/`deleted` terminal states, wall-clock timing, dynamic `retrieve` and `_sleep` hooks, return identity, and exact errors. In particular, the strict `>` timeout check still happens after a repeated retrieval, even if that retrieval returns a terminal state. This does not introduce a generic polling framework or change existing timeout behavior. Review pointers: - Original public methods: [sync](https://github.com/openai/openai-python/blob/bedb9a7b8839e193107e88b92f7cc166f08ac83d/src/openai/resources/files.py#L366) and [async](https://github.com/openai/openai-python/blob/bedb9a7b8839e193107e88b92f7cc166f08ac83d/src/openai/resources/files.py#L720). - [Extracted bodies](https://github.com/openai/openai-python/blob/3dfdf85e4d33a3274270aadccda6278c41ac8ede/src/openai/lib/_files.py#L12). - [Focused regression tests](https://github.com/openai/openai-python/blob/3dfdf85e4d33a3274270aadccda6278c41ac8ede/tests/lib/test_file_processing.py#L36): 13 cases covering sync/async terminal results, polling, timeout and retrieval-error propagation, plus async cancellation. Following the [review discussion](#3712 (comment)), the new handwritten suite is limited to 87 lines rather than an exhaustive matrix of clock, timeout, and event-order details. Existing Files API tests are unchanged. The review revision changes only this new test file; runtime source remains byte-identical to the original PR head. ## Additional context & links Validation: - Commands: `TEST_API_BASE_URL=http://127.0.0.1:4141 .venv/bin/python -m pytest tests/lib/test_file_processing.py tests/api_resources/test_files.py -q -n 0` and the same command using `.venv-pydantic-v1/bin/python`: **130 passed, 3 skipped** in each Pydantic mode. The skips are the existing aiohttp/respx2-incompatible `test_method_content`, `test_raw_response_content`, and `test_streaming_response_content` cases. - Commands: `.venv/bin/ruff format tests/lib/test_file_processing.py` and `./scripts/lint` passed on the review revision, including Ruff, Pyright, mypy, and import checks. The initial full `./scripts/format` also passed; unrelated reporter-formatting changes were excluded. - Command: `./scripts/build` passed. Both the wheel and source distribution contain `openai/lib/_files.py`; importing the helper and resource classes from the built wheel also passed. - Exact source/AST comparison against the base verified both polling bodies, all other resource statements, public signatures, exports, request construction, and response wrappers. The verified custom-code report keeps 36 mixed files and changes only the Files resource customization, shrinking it from **+47/-0 to +26/-0**. The other 35 customizations are unchanged. Reproduction command: ```sh python3 scripts/castiron/custom_code_report.py report \ --base bedb9a7 \ --head 3dfdf85 \ --fetch --require-head-hash --public \ --out /tmp/castiron-file-processing ```
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
See Commits and Changes for more details.
Created by
pull[bot] (v2.0.0-alpha.4)
Can you help keep this open source service alive? 💖 Please sponsor : )