How It Works

Overview

Remoki is a three-part system: mobile apps for capturing text, a cloud server for queuing, and a desktop app for processing entries into Anki cards.

┌─────────┐     ┌──────────────┐     ┌─────────────┐     ┌─────────┐
│  Phone  │ ──→ │ Cloud Server │ ──→ │ Desktop App │ ──→ │  Anki   │
│ (share) │     │   (queue)    │     │  (process)  │     │ (cards) │
└─────────┘     └──────────────┘     └─────────────┘     └─────────┘
                                           │
                                     ┌─────┴─────┐
                                     │  Yomitan  │
                                     │   (API)   │
                                     └───────────┘

Components

Cloud Server

• Runs on Cloudflare Workers with a D1 (SQLite) database
• Stores pending mining entries per user
• Handles device registration and authentication
• Globally distributed — low latency from anywhere
• No personal data stored beyond mining entries and auth tokens

Desktop App

• Built with Tauri (Rust backend, React frontend)
• Polls the server every few seconds for new entries
• Sends text to the local Yomitan API for tokenization and dictionary lookup
• Creates Anki cards via AnkiConnect with your configured field mappings
• Manages settings, field mappings, and pairing

Mobile Apps

• Android: Kotlin with Jetpack Compose, integrates as a share sheet extension and text selection action
• iOS: SwiftUI with a share extension
• Both use QR code scanning to pair with the desktop app
• Minimal footprint — their only job is to send selected text to the server

Data Flow

1. Registration

When the desktop app launches for the first time, it generates a unique user ID and registers with the cloud server. The server returns an authentication token. No email or sign-up required.

2. Pairing

The desktop app encodes the server URL, user ID, and auth token into a QR code. When a mobile app scans this code, it stores those credentials locally. All subsequent shares from that phone are authenticated with those credentials.

Caution

The QR code contains your auth credentials. Treat it like a password — don’t share screenshots of it publicly.

3. Mining Cycle

When you share text from your phone:

1. Capture — The mobile app sends a POST /add request to the server with the selected text and your auth headers (X-User-Id, X-Auth-Token).

2. Queue — The server stores the entry in the database with added_to_anki = false.

3. Poll — The desktop app polls GET /pending every few seconds and receives any unprocessed entries.

4. Tokenize — For each entry, the desktop app sends the text to the local Yomitan API’s /tokenize endpoint, which breaks it into individual words/tokens.

5. Enrich — For each token, it calls Yomitan’s /ankiFields endpoint to get dictionary data: definitions, readings, pitch accents, audio URLs, etc.

6. Create Card — The desktop app sends an addNote request to AnkiConnect with the enriched data mapped to your configured Anki fields.

7. Mark Complete — On success, the desktop app updates the server entry to added_to_anki = true. On failure, the error is logged and the entry can be retried.

Security

• Per-user authentication — Every API request includes X-User-Id and X-Auth-Token headers. Entries are isolated per user.
• QR pairing — Credentials are transferred via QR code scan, avoiding manual entry of long tokens.
• Local processing — All text enrichment (dictionary lookups, definitions) happens on your machine through Yomitan. The cloud server only sees the raw text you shared.
• Edge hosting — The server runs on Cloudflare’s edge network with built-in DDoS protection.