Schema-first build

Implementation & Testing

Sweepster is designed for AI-assisted implementation, but not by asking an AI coding tool to build everything at once.

Build order

  1. Define schemas for every module.
  2. Build the golden set from labelled vacancies.
  3. Build the loader and empty pipeline harness.
  4. Implement modules in pipeline order.
  5. Test each module against examples and edge cases.
  6. Run seam tests between modules.
  7. Implement scoring in isolation.
  8. Add open questions and feedback loop once scored output exists.

Suggested stack

Python Pydantic pytest JSON fixtures OpenAI API Claude API