Expand description
Learned-sparse (SPLADE / BGE-M3) vector type for the Uni graph database.
A leaf crate with zero uni-* dependencies — the same layering as
uni-btic. It owns the SparseVector type, its validating constructor,
the lossless binary encode codec, the pure scoring kernels in
ops (notably ops::sparse_dot), and SparseError. All
glue — Value/DataType variants, Arrow lowering, CV tag framing, DDL,
the index — lives in the integration crates that depend on this one.
Re-exports§
pub use error::SparseError;pub use sparse::SparseVector;
Modules§
- encode
- error
- ops
- Pure scoring/transform kernels over
SparseVector. No graph/runtime dependencies — this is the analogue ofuni-btic’s interval math, kept in the type crate so every layer (index, rerank, brute-force oracle) calls one canonical implementation. - sparse