Your personal operating system for career transition.

Implement a Quality Threshold Gate for CI

An eval harness only protects your system if it blocks bad deployments. This exercise builds the gate that translates eval results into a deploy/reject decision.

What to build

Implement QualityGate:

1. Constructor — accepts a thresholds: dict with any subset of these keys: min_pass_rate (float, default 0.80), max_regression_count (int, default 0), min_avg_score (float, default 0.70), max_failure_rate_by_category: dict[str, float] (per-category max failure rate).

2. check(report: dict) -> GateResult — report is the dict returned by a GoldenSuite.report() call. Returns a GateResult dataclass with: passed: bool, violations: list[str] (human-readable strings per violated threshold), summary: str (one-sentence deploy/reject verdict).

3. assert_passes(report: dict) -> None — Calls check() and raises RuntimeError with the violations if passed == False. Designed for use in CI scripts.

Constraints

• Standard library only. Each violation message should name the threshold, the actual value, and the threshold value.