[Shivam Sharma (ShivamSharma43)]

Description

Fixes #3415.

SeedDataset.get_values(first=..., last=...) and get_random_values(number=...) annotated their counts as PositiveInt, but that type was only imported under TYPE_CHECKING, so it was never enforced. A negative count was used directly as a slice bound: first=-2 returned every prompt except the last two, last=-2 every prompt except the first two, and first=2, last=-1 returned duplicates (p0, p1, p1, p2, ...). get_random_values(number=-1) failed inside random.sample with an unrelated message, and True was accepted as 1.

Changes:

• A small _validate_count helper raises ValueError("<name> must be a non-negative integer, got ...") for bools, non-integers and negative values. get_values checks first and last when they are given, and get_random_values checks number.
• 0 stays accepted, because it is already used as "none" (the existing test_seed_dataset_get_values calls get_values(last=0) and get_values(first=2, last=0)). The annotations now say NonNegativeInt to match.
• The docstrings list the new ValueError.

Valid counts behave exactly as before, including the overlap case (first + last >= len) and counts larger than the dataset in get_random_values.

Tests and Documentation

• New tests in tests/unit/models/test_seed_dataset.py: get_values rejects first=-2, last=-2, first=2, last=-1, first=True and last=1.5; get_random_values rejects -1, True, 1.5 and "2"; zero counts still return the expected results.
• On main all 9 new rejection cases fail (the calls return sliced or duplicated prompts, or fail with the random.sample error); the zero-count tests pass on both. With this change they all pass.
• pytest tests/unit/models tests/unit/datasets: 7353 passed.
• ruff check and ruff format pass on the changed files; git diff --check is clean.
• No documentation changes beyond the docstrings.

🤖 Generated with Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.