Game publishing operations. Local-first AI workflow tools.
I build practical tools for work that still depends on private files, human review, and repeatable delivery evidence.
A local web studio for game localization teams working with Excel language tables.
Live demo | Workflow guide | Docs
It is built for the part of localization work that is easy to lose in chat history: prompts, glossary decisions, QA findings, artifact history, and final delivery checks.
What is inside:
- Excel workbook flow: source files, target language sheets, glossary tables, and delivery artifacts.
- Review gates: glossary checks, consistency review, and final artifact review before delivery.
- Local data boundary: files, SQLite data, logs, and keys stay on the user's machine.
- Real app surface: React/Vite frontend, FastAPI backend, Playwright coverage, and GitHub Pages docs.
| Area | Project | What it does |
|---|---|---|
| Game localization | Localization Workflow Studio | Web workbench for translation prompts, glossary QA, Excel delivery gates, and reviewable artifacts. |
| Excel QA | Localization Workflow | Earlier strict review and regression harness for localization workbooks. |
| Terminology | Glossary Extraction Workflow | Bilingual glossary extraction and review for Excel language tables. |
| Agent operations | OA CLI | Local monitoring, token cost review, and operations tooling for OpenClaw-style agent sessions. |
| Local files | File Management Assistant | Windows file review and knowledge-base organization without turning private files into a cloud workflow. |
- Make AI-assisted work inspectable instead of hidden inside prompts.
- Keep sensitive workbooks, logs, and databases local by default.
- Ship small workflow tools with clear inputs, outputs, and review points.
Older experiments are being cleaned up or redirected toward the projects above.