Repository map

The repository separates reusable runtime libraries from runnable examples. Only the six artifacts in the upper part of this map are published; the three examples are CI smoke applications and learning material.

Published artifacts

Artifact	Responsibility	Depends on another Text artifact?
bluetape4k-text-bom	aligns Text artifact versions	constrains all five runtime modules
tokenizer-core	request/response models, options, severity, dictionaries, compact character collections	no
tokenizer-korean	Korean normalization, tokenization, stemming, phrase extraction, sentence splitting, blockwords	tokenizer-core
tokenizer-japanese	Kuromoji IPAdic tokenization, POS filtering, blockwords	tokenizer-core
lingua	Lingua detector factories, mixed-language extension, Unicode script filters	no Text runtime dependency
text-search	immutable Aho-Corasick automaton, DSL, replacement, Flow matching	no required Text runtime dependency

The BOM is metadata. It belongs in dependency management, not in application code. tokenizer-core is a foundation for processor implementations and safe request boundaries; most applications should depend on a Korean or Japanese processor rather than core alone.

Runnable examples

Gradle project	What it proves
:examples:text-search-examples	builder and DSL produce equivalent risk matches; replacement and first Flow alert work
:examples:lingua-examples	a detector can be reused, restricted to a language subset, and configured for low accuracy
:examples:tokenizer-safety-examples	blank, oversized, and processor failures map to sanitized service responses

Run all three:

./gradlew :examples:text-search-examples:run \
  :examples:lingua-examples:run \
  :examples:tokenizer-safety-examples:run

The examples are not published to Maven Central. Copy the demonstrated boundary or composition into your application; do not add an example project as a dependency.

Common compositions

Route by language

Use lingua to identify likely languages, then send supported text to tokenizer-korean or tokenizer-japanese. Keep an explicit fallback for unknown or ambiguous input. The mixed-language guide explains why detection should guide routing rather than act as unquestionable truth.

Tokenize and filter

Use a language processor for morphology and its blockword facilities for dictionary-backed filtering. Use tokenizer-core request models at the external boundary. The dictionary guide covers runtime updates and ownership.

Search without morphology

Use text-search when you already know the exact patterns and need to find many of them in one pass. It can search raw text independently of the tokenizers. Apply explicit normalization and word-boundary options when the input contract requires them.

Source evidence

• Release project registration
• Release module overview
• Examples overview

Release diagrams

These diagrams are loaded directly from README assets published with the 0.2.1 release and pinned to its immutable commit. They describe this manual’s released structure and runtime flows, not later Snapshot changes. Select a preview to open the SVG at the same release commit.

text Architecture diagram

Release README: README.md

Bluetape4k Text module composition chart

Release README: README.md

Bluetape4k Text overview diagram

Release README: README.md