DeltaCore is a Python backend for derivatives pricing, Greeks, volatility calibration, and market-risk analytics. Keep dependency flow directed from the API through services into pure domain, model, numerical, calibration, and risk modules. Domain code must not depend on FastAPI.
Backend code must remain Python-only. Libraries with native internals are welcome when they serve a documented numerical purpose, but custom native extensions are out of scope.
- Use typed Pydantic models at API boundaries.
- Keep pricing kernels pure and deterministic: no hidden network calls, mutable globals, or wall-clock dependencies.
- Use explicit seeds for stochastic tests and examples.
- Document public functions with inputs, units, conventions, and numerical assumptions.
- Surface numerical failures through structured errors or domain-specific exceptions.
- Keep tests independent of live market data by using small, documented, synthetic fixtures.
- Update public documentation when behavior or model coverage changes.
Every pricing model needs at least one independent validation, such as a closed-form reference, limiting case, put-call parity check, finite-difference comparison, or regression fixture.
Greeks must state bump convention, units, annualization, and sign. Calibration results must expose convergence status, objective value, parameter bounds, and failure reason. Monte Carlo outputs exposed to users must include standard error or a confidence interval. Do not describe unvalidated or toy models as production-accurate.
uv sync --extra dev
uv run ruff format --check .
uv run ruff check .
uv run mypy src tests
uv run pytestChanges are ready when typing, tests, numerical invariants, and affected documentation all agree.