Crate sakurs_core 

Delta-Stack Monoid algorithm for parallel sentence boundary detection

This crate implements a mathematically sound parallel approach to sentence boundary detection using monoid algebra. The core innovation lies in representing parsing state as a monoid, enabling associative operations that can be computed in parallel while maintaining perfect accuracy.

Stability Notice

This crate is in pre-release (v0.1.0-dev). The API is not yet stable and may change significantly in future releases. Use with caution in production code.

We recommend pinning to exact versions in your Cargo.toml:

sakurs-core = "=0.1.0"

Architecture

The crate follows a hexagonal architecture pattern:

• Domain layer: Pure mathematical algorithms and monoid operations
• Application layer: Orchestration and parallel processing logic
• Adapter layer: Interfaces for different use cases (CLI, Python, etc.)

Example

use sakurs_core::{SentenceProcessor, Input};

// Create processor with default configuration
let processor = SentenceProcessor::new();

// Process text
let text = "Hello world. This is a test.";
let result = processor.process(Input::from_text(text)).unwrap();

// Check boundaries
assert!(!result.boundaries.is_empty());
// Note: SentenceProcessor may detect different boundary counts than expected

Re-exports

pub use api::Boundary;

pub use api::Config;

pub use api::ConfigBuilder;

pub use api::Error as ApiError;

pub use api::Input;

pub use api::Language;

pub use api::Output;

pub use api::ProcessingMetadata;

pub use api::ProcessingStats;

pub use api::SentenceProcessor;

pub use application::DeltaStackProcessor;

pub use application::DeltaStackResult;

pub use application::ExecutionMode;

pub use application::ProcessorConfig;

pub use domain::*;

Modules

api

New unified API for sakurs-core

application

Application layer for orchestrating text processing

domain

Domain layer for the Delta-Stack Monoid algorithm