Kapi CLI is an AI-native localization framework designed for processing content across various formats with concurrent pipelines and pluggable tools. It enables efficient handling of content through features such as format-aware parsing, supporting diverse file types like JSON, Markdown, HTML, and more.
Key Features:
Format-Aware Parsing: Supports multiple content formats, ensuring accurate processing.
Concurrent Processing Pipelines: Enhances efficiency with parallel execution capabilities.
AI and Machine Translation Integration: Facilitates seamless integration with AI tools and MT services.
Termbase Management: Provides robust terminology management for consistent localization.
XLIFF/PO Support: Enables effective interchange of translation data.
Extensibility: Offers plugin support for custom workflows and integrations.
Audience & Benefits:
Ideal for developers, localization teams, and content managers seeking efficient localization solutions. Kapi CLI offers increased efficiency in processing tasks, improved quality control through AI-driven checks, and seamless integration with existing workflows, enhancing overall productivity and project management.
Kapi CLI can be installed via winget on Windows, making it accessible for a streamlined setup process.
README
neokapi
> Experimental. Neokapi is an ongoing experiment and not yet recommended for production use.
neokapi is a format-aware content engine in Go: parse any format — JSON, Markdown, HTML, config, office formats — into one unified content model, edit the content inside it, check it, and write it back byte-for-byte. The same engine resolves the context that applies to a piece of content — the terms, voice and rules that hold there — and makes that content work in every language.
The engine carries the Okapi Framework heritage forward — channel-based concurrent processing and pluggable tools — in an AI-native design. Multilingual content is the flagship application: extraction, translation, content memory and a terms store, XLIFF/PO interchange, and an Okapi-parity fidelity story. The same engine also serves on-brand source content, AI ingestion, and programmatic editing.
The bowrain platform — the same context graph held across every project, built on neokapi — lives under bowrain/ with its own README.
Install
brew install neokapi/tap/kapi-cli # macOS/Linux
winget install Neokapi.KapiCli # Windows
Pre-built binaries for Linux, macOS, and Windows (amd64 + arm64) are on the Releases page. Kapi Desktop ships a signed Windows installer and a macOS cask — see the installation guide.
Repository layout
The framework + kapi CLI live at the root. Companion areas are clearly marked.
The build needs ICU development libraries — the SQLite FTS5 ICU tokenizer is
compiled through cgo:
brew bundle # macOS — see Brewfile
sudo apt-get install libicu-dev pkg-config # Debian / Ubuntu
make doctor # report what is still missing
make build # Build kapi CLI → bin/kapi
make test # Run all tests
make check # fmt + vet + lint
go test -tags fts5 ./core/flow/ -run TestExecutorCancellation -v # Single test
Prefer make over a bare go build / go test: it passes -tags fts5 and
locates Homebrew's ICU, without which go reports no such function: fts5 or
a bare [build failed] that never mentions ICU.
For the bowrain platform (server, desktop app, web app), see bowrain/README.md.
Frontend / docs site
A single root pnpm workspace coordinates the kapi side:
vp install # at the repo root
cd web && vp run start # Docusaurus dev server (kapi docs + landing)
make kapi-storybook # Storybook on :6007
Internals (root) — repo-wide testing, interfaces, release process
Make targets
make build Build kapi CLI → bin/kapi
make build-all Build all Go binaries (kapi + bowrain side)
make test Run all tests
make test-unit Unit tests only (-short flag)
make test-race Tests with race detector
make cover Coverage report → coverage/coverage.html
make fmt gofmt -w -s
make vet go vet (all modules)
make lint golangci-lint (all modules)
make check fmt + vet + lint
make pre-push Run checks relevant to your changes (mirrors CI)
