Skip to main content
This guide walks you through starting the app from source. When you finish, you can enter a Japanese sentence in your browser and see the analysis.

​
Prerequisites

Before you start, make sure you have:
  • Node.js 22 and npm
  • Git
  • A DeepSeek API key or a Gemini API key
We recommend setting up DeepSeek first. It is the default provider and handles analysis, translation, definitions, and image recognition. Only Gemini TTS requires a Gemini API key.

​
Start the app

1

Clone the repository

2

Install dependencies

3

Create the environment file

4

Add an API key

Open .env.local and fill in at least one key:
.env.local
All other variables are optional. For the full list, see Configuration.
5

Start the development server

Open http://localhost:3000.
Enter a Japanese sentence, such as 「天気がいいから、散歩しましょう」, and click Analyze. If you see word-by-word readings and a translation in your interface language, your setup works. You can also click Sentence of the day on the home page to analyze it directly.

​
Access from your phone

To open the development server from a phone on the same local network, make the server listen on all network interfaces:
Then open http://<your-computer-IP>:3000 in your phone’s browser.

​
Only a Gemini key is configured

The app uses DeepSeek by default. If you only filled in GEMINI_API_KEY, open the page and click Settings in the top-right corner. Switch AI provider to Gemini, then click Save settings.

​
Run a production build

To try the production build on your machine, run:
npm start listens on port 3000 by default.

​
What’s next

User guide

Learn how to use analysis, word details, and multi-word selection.

Deployment

Publish to Vercel, or run it on a server with Docker.