ci: compare Agents compatibility with PR base

[jbeckwith-oai]

Summary

• check the Agents SDK against both the pull request base and head SDK
• compare normalized mypy JSON diagnostics so only newly introduced incompatibilities fail
• keep abnormal mypy invocations as hard failures

Root cause

The Agents SDK is already incompatible with openai-python main after the HTTPX 2 migration. The existing workflow checks only the pull request SDK, so those pre-existing type errors make unrelated pull requests fail. Simply skipping the head check when the base is red would disable regression detection entirely.

This keeps the regression check meaningful while the upstream incompatibility exists: unchanged or reduced diagnostics pass, and any added diagnostic fails.

Validation

• ./scripts/lint
• parsed .github/workflows/detect-breaking-changes.yml as YAML
• reproduced a red base with eight unchanged head diagnostics and confirmed it passes
• reproduced a compatible base followed by eight head diagnostics and confirmed it fails
• reproduced a red base followed by a fully clean head and confirmed it passes
• completed two-pass correctness review and strict maintainability review

[openai-sdks]

OkTest Summary

✅ 237/237 SDK tests passed in 11.419s for Python SDK PR #3604.

Test results — 42 files

Test	Result	Time
tests/chat-completions-complex-body.test.ts	✅ Passed	209ms
tests/chat-completions-create.test.ts	✅ Passed	229ms
tests/chat-completions-stream.test.ts	✅ Passed	125ms
tests/files-content-binary.test.ts	✅ Passed	204ms
tests/files-create-multipart.test.ts	✅ Passed	229ms
tests/files-list-pagination.test.ts	✅ Passed	171ms
tests/initialize-config.test.ts	✅ Passed	226ms
tests/instance-isolation.test.ts	✅ Passed	252ms
tests/models-list.test.ts	✅ Passed	167ms
tests/responses-background-lifecycle.test.ts	✅ Passed	266ms
tests/responses-body-method-errors.test.ts	✅ Passed	382ms
tests/responses-cancel-timeout.test.ts	✅ Passed	212ms
tests/responses-cancel.test.ts	✅ Passed	217ms
tests/responses-compact-retries.test.ts	✅ Passed	350ms
tests/responses-compact.test.ts	✅ Passed	233ms
tests/responses-create-advanced-stream.test.ts	✅ Passed	140ms
tests/responses-create-advanced.test.ts	✅ Passed	211ms
tests/responses-create-disconnect.test.ts	✅ Passed	1.266s
tests/responses-create-errors.test.ts	✅ Passed	183ms
tests/responses-create-malformed-api-responses.test.ts	✅ Passed	253ms
tests/responses-create-retries.test.ts	✅ Passed	245ms
tests/responses-create-stream-failures.test.ts	✅ Passed	126ms
tests/responses-create-stream-timeout.test.ts	✅ Passed	217ms
tests/responses-create-stream-wire.test.ts	✅ Passed	3.835s
tests/responses-create-stream.test.ts	✅ Passed	272ms
tests/responses-create-terminal-states.test.ts	✅ Passed	271ms
tests/responses-create-timeout.test.ts	✅ Passed	205ms
tests/responses-create.test.ts	✅ Passed	212ms
tests/responses-delete.test.ts	✅ Passed	198ms
tests/responses-input-items-errors.test.ts	✅ Passed	339ms
tests/responses-input-items-list.test.ts	✅ Passed	340ms
tests/responses-input-items-options.test.ts	✅ Passed	292ms
tests/responses-input-tokens-count-timeout.test.ts	✅ Passed	253ms
tests/responses-input-tokens-count.test.ts	✅ Passed	232ms
tests/responses-malformed-inputs.test.ts	✅ Passed	2.948s
tests/responses-not-found-errors.test.ts	✅ Passed	271ms
tests/responses-parse.test.ts	✅ Passed	435ms
tests/responses-retrieve-retries.test.ts	✅ Passed	381ms
tests/responses-retrieve.test.ts	✅ Passed	222ms
tests/responses-stored-method-errors.test.ts	✅ Passed	572ms
tests/retry-behavior.test.ts	✅ Passed	3.154s
tests/sdk-error-shape.test.ts	✅ Passed	292ms

View OkTest run #31630777858

_{SDK merge (c7b525c7f63f) · head (c76a76064670) · base (a1eeab58db02) · OkTest (a845fa206fa4)}