fix(transform schema): preserve const constraints

[pravinireri]

Summary

• Preserve JSON Schema const constraints in transform_schema() instead of moving them into description text.
• Add regression coverage for scalar and null const values.

Problem

Pydantic v2 emits const for single-value Literal fields. transform_schema() currently has explicit handling for enum but not const, so the keyword reaches the unsupported-property fallback and is moved into description. The request schema therefore loses a supported grammar-level constraint (const is listed as supported for structured outputs), while SDK-side parsing still validates against the original Pydantic model.

Related to the enum handling added in #1275.

class Job(pydantic.BaseModel):
    status: Literal["ok"]

# property schema emitted by pydantic:   {"const": "ok", "title": "Status", "type": "string"}
# transform_schema() before this change: {"type": "string", "title": "Status", "description": "{const: ok}"}
# transform_schema() after this change:  {"type": "string", "const": "ok", "title": "Status"}

Fix

Preserve const next to the existing enum handling. Use a key-presence check rather than pop("const", None) so a valid const: null constraint is not confused with an absent key.

Tests

Both new tests fail on unmodified upstream/next because const is demoted into description ({"type": "string", "const": "ok"} becomes {"type": "string", "description": "{const: ok}"}).

Commands run (all exited zero):

• uv run --isolated --all-extras pytest tests/lib/_parse/test_transform.py -n0 — 16 passed (Python 3.9, pydantic 2.12.5)
• uv run --isolated --all-extras --no-extra=mcp --group=pydantic-v1 pytest tests/lib/_parse/test_transform.py -n0 — 16 passed (Python 3.9, pydantic v1)
• same test file with UV_PYTHON=">=3.14.0" — 16 passed (Python 3.14, pydantic 2.12)
• uv run ruff check on the two touched files — passed
• uv run ruff format --check on the two touched files — already formatted
• uv run pyright on the two touched files — 0 errors, 0 warnings

Scope

Hand-maintained schema-transformer logic (src/anthropic/lib/_parse/_transform.py) and its focused regression tests only.

[Copilot]

Pull request overview

This pull request updates the SDK’s JSON Schema transformer (transform_schema()) to preserve JSON Schema const constraints (notably emitted by Pydantic v2 for single-value Literal fields) so that request schemas retain grammar-level constraints rather than demoting them into descriptive text.

Changes:

• Preserve const in transform_schema() similarly to existing enum handling.
• Implement const handling via key-presence check to correctly retain const: null.
• Add regression tests covering scalar and null const values.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
src/anthropic/lib/_parse/_transform.py	Adds explicit const preservation logic (including correct handling for const: null).
tests/lib/_parse/test_transform.py	Adds regression tests verifying const is preserved for both scalar and null values.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.