revert: nested BaseModel coercion (#295)

[jbeckwith-oai]

Summary

• Reverts fix: coerce nested BaseModel fields #295 as a standalone mechanical revert.
• Adds focused characterization tests for the published raw-request / parsed-response BaseModel contract.
• Does not introduce a replacement union or discriminator design.

Commit structure

1. b269a43e is the literal Git revert of merge commit e5152a92962251ce76ceb3d89ec0b6b37fa4efdb.
2. 6e8650e1 adds only the compatibility test suite.

Keeping those changes separate makes the rollback itself directly auditable and the additional contract coverage independently reviewable.

Why

#295 globally coerced nested request-model values, changing the documented contract that BaseModel#[] and #to_h expose the raw values callers supplied. The broader conversion changes also interact unsafely with optional discriminators and schemas where multiple variants share the same discriminator.

The previous forward-fix PR #373 was closed rather than repurposed. Draft #370 was also closed; issue #287 remains open for a narrower streaming-scoped solution after the model system is redesigned from first principles.

Compatibility coverage

• raw request identity through accessors, #[], #to_h, and #deep_to_h
• idiomatic nested fetch chains such as params.to_h.fetch(:input).fetch(0).fetch(:type)
• request serialization after in-place mutation of caller-owned hashes
• parsed-model symbol access, equality, and hashing
• response coercion retaining expected nested model objects

Generator scope

No Castiron change is required. Castiron intentionally excludes the SDK-owned lib/openai/internal/ runtime, its signatures, and the handwritten test tree, so generated output is unchanged.

Validation

• focused BaseModel suite: 17 runs, 261 assertions
• focused compatibility suite: 9 runs, 39 assertions
• full suite via the canonical scripts/test entry point on Ruby 3.3.12, 3.4.10, and 4.0.6: 613 runs, 2,256 assertions on each
• bundle exec rake lint: RuboCop, Sorbet, and 1,212 RBS files green
• bundle exec rake build:gem
• thermonuclear code-quality review: no findings

[openai-sdks]

OkTest Summary

✅ 237/237 SDK tests passed in 8.914s for Ruby SDK PR #375.

Test results — 42 files

Test	Result	Time
tests/chat-completions-complex-body.test.ts	✅ Passed	170ms
tests/chat-completions-create.test.ts	✅ Passed	256ms
tests/chat-completions-stream.test.ts	✅ Passed	141ms
tests/files-content-binary.test.ts	✅ Passed	235ms
tests/files-create-multipart.test.ts	✅ Passed	249ms
tests/files-list-pagination.test.ts	✅ Passed	193ms
tests/initialize-config.test.ts	✅ Passed	188ms
tests/instance-isolation.test.ts	✅ Passed	279ms
tests/models-list.test.ts	✅ Passed	203ms
tests/responses-background-lifecycle.test.ts	✅ Passed	242ms
tests/responses-body-method-errors.test.ts	✅ Passed	455ms
tests/responses-cancel-timeout.test.ts	✅ Passed	207ms
tests/responses-cancel.test.ts	✅ Passed	253ms
tests/responses-compact-retries.test.ts	✅ Passed	284ms
tests/responses-compact.test.ts	✅ Passed	212ms
tests/responses-create-advanced-stream.test.ts	✅ Passed	135ms
tests/responses-create-advanced.test.ts	✅ Passed	162ms
tests/responses-create-disconnect.test.ts	✅ Passed	157ms
tests/responses-create-errors.test.ts	✅ Passed	293ms
tests/responses-create-malformed-api-responses.test.ts	✅ Passed	132ms
tests/responses-create-retries.test.ts	✅ Passed	282ms
tests/responses-create-stream-failures.test.ts	✅ Passed	134ms
tests/responses-create-stream-timeout.test.ts	✅ Passed	236ms
tests/responses-create-stream-wire.test.ts	✅ Passed	2.064s
tests/responses-create-stream.test.ts	✅ Passed	106ms
tests/responses-create-terminal-states.test.ts	✅ Passed	264ms
tests/responses-create-timeout.test.ts	✅ Passed	213ms
tests/responses-create.test.ts	✅ Passed	348ms
tests/responses-delete.test.ts	✅ Passed	301ms
tests/responses-input-items-errors.test.ts	✅ Passed	189ms
tests/responses-input-items-list.test.ts	✅ Passed	181ms
tests/responses-input-items-options.test.ts	✅ Passed	190ms
tests/responses-input-tokens-count-timeout.test.ts	✅ Passed	285ms
tests/responses-input-tokens-count.test.ts	✅ Passed	269ms
tests/responses-malformed-inputs.test.ts	✅ Passed	2.23s
tests/responses-not-found-errors.test.ts	✅ Passed	352ms
tests/responses-parse.test.ts	✅ Passed	209ms
tests/responses-retrieve-retries.test.ts	✅ Passed	287ms
tests/responses-retrieve.test.ts	✅ Passed	243ms
tests/responses-stored-method-errors.test.ts	✅ Passed	741ms
tests/retry-behavior.test.ts	✅ Passed	1.988s
tests/sdk-error-shape.test.ts	✅ Passed	381ms

View OkTest run #31648868996

_{SDK merge (c4d144d8860b) · head (6e8650e1408d) · base (919977919667) · OkTest (a845fa206fa4)}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 6e8650e140

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".

[jbeckwith-oai]

Automated review identified six pre-#295 converter behaviors that this literal rollback intentionally restores. They are now tracked together in #376 with explicit regression-coverage requirements: union error isolation, nullable fields, strictness preservation, existing-model identity, collection error retention, and nullable collection matchers.

I replied to and resolved each inline thread as separately tracked follow-up work. No code was added to this PR because doing so would make #375 a partial model-conversion redesign rather than the standalone rollback requested here.