fix(transform schema): keep const instead of moving it into the description

[chrikrah]

This carries @pravinireri's #1860 forward against main. It's the same fix. #1860 targeted next, and it closed unreviewed on 2026-10-08 with the other open next pull requests.

transform_schema() has no case for const, so it falls through to the unsupported-keyword fallback in src/anthropic/lib/_parse/_transform.py. It reaches the API as text in description. The structured outputs docs list const as supported. Pydantic emits it for every single-value Literal, including the tag of a discriminated union, so messages.parse() sends that tag as a free string.

Before, on b4b7916, for Pet with pet: Union[Cat, Dog] = Field(discriminator="kind") and kind: Literal["cat"] / Literal["dog"]:

$ python -I probe_const.py   # json.dumps(transform_schema(Pet)["$defs"]["Cat"]["properties"]["kind"])
{"type": "string", "title": "Kind", "description": "{const: cat}"}

After:

$ python -I probe_const.py
{"type": "string", "const": "cat", "title": "Kind"}

const now sits beside enum, which the function already passes through since #1275. It checks for the key, so const: null survives too. No issue exists. Refs #1860.

Verification

$ pytest tests/lib/_parse/test_transform.py -q -p no:cacheprovider
16 passed in 6.52s                # b4b7916: 14 passed
$ pytest tests/lib/_parse tests/lib/test_output_format_helpers.py tests/api_resources/beta/test_output_format_conversion.py -q -n 4
42 passed                         # b4b7916: 40 passed
$ ruff check src/anthropic/lib/_parse tests/lib/_parse && ruff format --check src/anthropic/lib/_parse tests/lib/_parse   # ruff 0.15.6
All checks passed!
6 files already formatted
$ pyright src/anthropic/lib/_parse/_transform.py tests/lib/_parse/test_transform.py   # 1.1.399
0 errors, 0 warnings, 0 informations
# not run: ./scripts/test (mock server, Pydantic v1 matrix) and a live request

Reverting _transform.py and keeping the tests gives 2 failed, 14 passed: test_const_schema and test_const_from_pydantic_literal_discriminator.

@meorphis-anthropic, was the close of #1860 about the next branch or about the change? @craigie-ant, this and my #2015 are both small transform_schema fixes, and I can fold them into one pull request if that's easier to review.