Game Localizer v0.1.0

Game Localizer v0.1.0

v0.1.0 is the first public release baseline of Game Localizer: a traceable game-localization pipeline that combines translation memory, model-assisted translation, deterministic QA, human review, artifact building, and governed publishing.

Highlights

• Multiple localization formats — Gettext PO/MO, ParaTranz JSON, and Paradox YAML.
• Translation memory — SQLite TM with stable coordinates, source fingerprints, bootstrap/migration support, and stale-entry recovery.
• Model-assisted workflows — OpenAI-compatible providers with configurable concurrency, rate limiting, and tokenizer support.
• Deterministic quality gates — placeholder, terminology, residual source-language, filtering, normalization, and release-readiness checks remain authoritative.
• Human-in-the-loop review — auditable revision, checkpoints, recovery, and incremental rebuilds.
• Governed artifacts — separate preview and QualityGate-protected release builds.
• Publishing — local directories, GitHub Releases, Cloudflare R2, and Alibaba Cloud OSS.
• Local dashboard — observability and targeted single-user revision without duplicating the Python business kernel.
• Bilingual documentation — Chinese and English documentation with setup, usage, configuration, TM, and roadmap guides.

Installation

Python 3.10 or later is required.

From a source checkout:

python -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install -e ".[all]"
localizer dashboard projects/example/project.yaml --host 127.0.0.1 --port 8080

This GitHub Release also attaches a built Python wheel and source distribution. PyPI publication is not part of this release workflow.

Validation

The release candidate is expected to pass:

• Windows and Ubuntu test jobs;
• Pydantic 1.x and Pydantic 2.x compatibility jobs;
• full-history gitleaks scanning;
• strict documentation build on the release PR.

Current boundaries

• Community-platform workflows do not yet include an online API client.
• The dashboard is local/single-user rather than a multi-user approval platform.
• Remote publishing requires optional dependencies and explicit credentials/governance configuration.
• M8 controlled agent workflows and the Tauri client are roadmap work and are not included in v0.1.0.

Roadmap

M8 work is tracked through public issues covering the shared AgentRun runtime, TM auditing, reproducible i18n Prompt evaluation, Prompt foundation workflows, governed build orchestration, and a Windows Tauri client. Contributor-sized evaluation tasks are also available under good first issue / help wanted.

See CHANGELOG.md for the complete release baseline.