What problem does it solve?
This Skill solves the problem of catching Swift regressions and ensuring your speech model pipeline still works after code changes.
Core Features & Use Cases
- Unit testing: Runs quick, CI-safe tests without downloading models by default.
- End-to-end testing (E2E): Runs full pipeline tests that require cached weights and a metallib setup for model execution.
- Targeted test filtering: Passes any custom filter string through to swift test for running specific test suites or cases.
Use case: After changing MLX/CoreML inference logic or test harnesses, run unit tests first to catch fast failures, then run E2E tests to confirm the full speech pipeline works with real models.
Quick Start
Run the unit test suite by executing the test Skill with the argument unit.