Skip to content

ider-zh/translation_flow

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

2 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

🌐 Translation Flow - Multi-Agent Document Translation Workflow

FastAPI Next.js Marvin AI Tailwind CSS

Translation Flow is an advanced translation workflow system specifically designed for long documents (.docx). By leveraging Multi-Agent collaboration, Style Restoration technology, and LLM-as-a-Judge (automated quality assessment), it provides precision and formatting retention far beyond traditional translation software.


πŸ“Έ Interface Preview

🏠 Project Workspace

The central hub for your translation project. View and manage original paragraphs, compare multiple AI translations, and monitor the status of your translators and judgers in real-time. Workspace

πŸ€– Translation Fleet

Flexibly configure multiple translation agents to deliver personalized results based on different audiences and styles. Translation Engine

βš–οΈ Quality Assessment (Judger)

Introduce "Judge" agents to automatically compare multiple translation versions and select the best result based on fidelity and naturalness. Judger

🏷️ Terminology Management (NER)

Automatically identify and extract specialized terminology and named entities to ensure consistency across the entire document. Glossary


✨ Core Features

  • πŸ“„ Deep Style Restoration (WordFlux Technology): Translates text while precisely re-aligning original formatting like bold, italic, and underline using LLM-powered tag alignment.
  • 🧠 Intelligent Context Awareness: Uses a sliding window technique to include preceding context during translation, solving the "fragmentation" issue common in long document processing.
  • πŸ” Automated NER Extraction: Concurrent extraction of names, locations, organizations, and technical terms to build a project-specific glossary.
  • 🎭 Customizable Translator Fleet: Deeply customize translations for specific "Audiences" (e.g., General, Technical, Academic) and "Styles" (e.g., Storytelling, Formal, Business).
  • πŸ‘¨β€βš–οΈ AI Judger (LLM-as-a-Judge): Scores versions based on Fidelity, Consistency, and Naturalness, providing detailed reasoning for each selection.
  • ⚑ High-Performance Virtualized UI: Powered by React-Virtuoso to handle documents with tens of thousands of paragraphs smoothly.
  • πŸ’Ύ Versatile Export Options: Export as plain text (.txt) or fully formatted Word documents (.docx).

πŸš€ Quick Start

Backend Setup (Python 3.14+)

  1. Navigate to the backend directory:
    cd backend
  2. Install dependencies (recommended using uv):
    uv sync
  3. Start the development server:
    uv run fastapi dev src/main.py --port 8000

Frontend Setup (Node.js 18+)

  1. Navigate to the frontend directory:
    cd frontend
  2. Install dependencies:
    npm install
  3. Start the development environment:
    npm run dev

Visit http://localhost:3000 to get started.


πŸ› οΈ Tech Stack

  • Backend:
    • FastAPI: High-performance asynchronous API framework.
    • Marvin AI: Core for building structured agents and workflows.
    • SQLAlchemy: Asynchronous Database ORM.
    • python-docx: Word document parsing and generation.
  • Frontend:

🀝 Contributing

We welcome contributions from the community! Whether you are fixing bugs, adding new features, or improving documentation, your help is much appreciated.

  • Fork the Repository: Feel free to fork the project and experiment with your own ideas.
  • Submit a Pull Request: If you have a fix or a feature, please submit a PR! Ensure your code follows the existing style and includes relevant tests.
  • Report Issues: Found a bug? Open an issue and let us know.

πŸ“œ License

This project is licensed under the MIT License. See the LICENSE file for details.


Crafted with ❀️ by the ider

About

🌐 Translation Flow - An advanced multi-agent document translation workflow system for Word documents (.docx). Features AI-powered style restoration, intelligent context awareness, automated terminology extraction, and quality assessment through LLM-based judging.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors