PRD: GemmaTrans Desktop Client

1. Project Overview

GemmaTrans is a lightweight, high-performance desktop translation client built with Tauri and Vue 3. It is designed to act as a modern frontend for the TranslateGemma model running on a local or remote Ollama instance (including over private networks like Tailscale). The application prioritizes speed, privacy, and a professional "Google Translate" aesthetic.

2. Technical Stack

Core Framework: Tauri (Rust backend for system-level API access).
Frontend Framework: Vue 3 (Composition API).
Styling: Tailwind CSS (Minimalist, modern utility-first CSS).
Package Manager: pnpm.
State Management: Pinia (To persist user settings and prompt templates).
Icons: Lucide Vue Next.

3. User Interface (UI) Requirements

The UI should be Modern, Minimalist, and Functional, mimicking the clean layout of Google Translate.

3.1 Main Translation Interface

Split-Pane Layout:
- Left (Source): Input text area with a header for language selection.
- Right (Target): Read-only text area for the translation result, with a header for the target language selection.
Language Selection: Dropdowns supporting both Full Name (e.g., English) and ISO Code (e.g., en).
Actions: * "Translate" button (with loading state).
- "Copy" button on the target pane.
- "Clear" button on the source pane.

3.2 Settings View

API Configuration: A text input for the Ollama API Address (e.g., http://100.x.y.z:11434).
Output Control: A toggle switch for Enable Streaming (renders text as it is generated).
Prompt Engineering: A large text area for the System Prompt Template.

4. Functional Requirements

4.1 Prompt Engineering Engine

The application must dynamically inject variables into the user-provided template before sending the request to the Ollama API.

Default Template:

You are a professional {SOURCE_LANG} ({SOURCE_CODE}) to {TARGET_LANG} ({TARGET_CODE}) translator. Your goal is to accurately convey the meaning and nuances of the original {SOURCE_LANG} text while adhering to {TARGET_LANG} grammar, vocabulary, and cultural sensitivities. Produce only the {TARGET_LANG} translation, without any additional explanations or commentary. Please translate the following {SOURCE_LANG} text into {TARGET_LANG}:

{TEXT}

4.2 Template Variables

The engine must support and correctly replace the following tokens:

{SOURCE_LANG}: Full name of the input language (e.g., "Chinese").
{SOURCE_CODE}: Short code of the input language (e.g., "zh-Hans").
{TARGET_LANG}: Full name of the output language (e.g., "English").
{TARGET_CODE}: Short code of the output language (e.g., "en").
{TEXT}: The actual content provided by the user.

4.3 Ollama Integration

Support for the /api/generate endpoint.
Ability to handle both stream: true (progressive UI updates) and stream: false (batch updates).
Support for connecting to any valid URL provided in Settings, ensuring compatibility with Tailscale IPs.

5. Technical Implementation Details

5.1 Language Mapping (Default Configuration)

The application will provide a default list of languages:

Language	Code
English	en
Chinese	zh-Hans
Japanese	ja
Spanish	es
French	fr
German	de

5.2 Persistence

All settings (API Address, Stream Toggle, and Custom Prompt Template) must be stored in the user's local application data folder to ensure they are remembered across restarts.

6. Success Metrics

Connectivity: Seamless communication with an Ollama instance over a Tailscale network.
UX: The "Streaming" mode should feel responsive with no UI lag during high-token-per-second generation.

3.9 KiB Raw Blame History