Skip to content

CLAUDE.md

Latest commit

 

History

History
40 lines (34 loc) · 2.02 KB

CLAUDE.md

File metadata and controls

40 lines (34 loc) · 2.02 KB
Raw

HawkesPyLib — contributor & agent guide

Python library for simulation and inference of univariate Hawkes processes. The public API is three modules under src/HawkesPyLib/: simulation, inference, and processes. Hot numerical paths live in core/ and are JIT-compiled with numba.

This file is tool-neutral on purpose. The real project knowledge lives in the sources linked below (readable by any human or agent). CLAUDE.md is just an entry point — mirror it to AGENTS.md (cp CLAUDE.md AGENTS.md) for other agentic tools if you use them.

Getting oriented

  • Architecture map (platform-independent): if present, graphify-out/GRAPH_REPORT.md is a knowledge-graph overview of the modules, their communities and relationships (graphify-out/graph.html is the interactive version). The map is produced by the graphify tool — (re)generate it if it is missing or stale.
  • Docs: README.md and https://simbold.github.io/HawkesPyLib/

Working in this repo

  • Tests: tox (Python 3.10–3.13), or in a venv: pip install -e ".[testing]" then pytest.
  • Lint: ruff check src/HawkesPyLib tests (config in pyproject.toml).
  • Supported Python: ≥ 3.10. Runtime deps: numpy, scipy, numba (numpy 2.x compatible).

Releasing

Publishing is tag-driven via GitHub Actions with PyPI Trusted Publishing (OIDC — no tokens). Bump version in pyproject.toml, then push a matching tag; CI checks the tag equals the package version before publishing.

  • Test release → TestPyPI: version X.Y.ZrcN, tag vX.Y.ZrcN (e.g. v0.3.0rc1).
  • Final release → PyPI: version X.Y.Z, tag vX.Y.Z (e.g. v0.3.0).

Conventions that matter

  • Functions in src/HawkesPyLib/core/ are compiled with numba @njit using explicit type signatures, e.g. @njit(float64[:](float64, float64, int32), nogil=True). When you change a function's arguments, update its signature to match or compilation fails at call time.
  • Use np.random.default_rng(...) for randomness (as in inference).