Shells
One ecosystem, one tensor graph
Shells are reef packages that ride the Chelis runtime. Numeric work composes with the compiler's tensor graph instead of sitting beside it, so automatic differentiation flows through every one of them.
Core runtime
The runtime that ships in the compiler
Every shell rides the same tensor DAG, so automatic differentiation flows through all of them.
Numerical and data
Numbers and tables, typed end to end
Machine learning
Models on a checkable substrate
Bridges
Bring code in from where it already lives
octant
LaTeX-to-Chelis translation that tracks provenance, so a compiled trace points back to the original formula.
View on GitHub →calcify
Python-to-Chelis translation, carrying numeric code onto the typed tensor graph.
View on GitHub →hydronnx
The ONNX shell, reading exported models into Chelis Deep.
View on GitHub →c-earchin
EARS-to-Chelis translation that attaches property-witness metadata to the requirements it carries across.
View on GitHub →Domain shells
Built for the fields that need the proof
shoals
Quantitative finance on typed tensors, with derivatives that the compiler differentiates for you.
View on GitHub →whale
A reusable betting shell for modeling odds and stakes on the same checked substrate.
View on GitHub →hull
The compiler-validation and language-spec shell.
View on GitHub →