Local Translator GUI

GUI Application for Translation Running on Your Local PC

English | 日本語

This repository provides a GUI application that allows you to select translation engines (such as OpenAI, Ollama, etc.) and perform text translation.
In addition to the main translation functionality, it features the ability to translate the same text with multiple engines and display the results for comparison.
The UI uses Streamlit, allowing you to operate it in your web browser.

Main Features

1. Model Management via YAML File
   Engine names selected by the user are mapped to the actual model IDs via configs/avail_models.yaml.
   Example:

   OpenAI/GPT-4o-mini: gpt-4o-mini
   OpenAI/GPT-4o: gpt-4o
   Ollama/gemma-3-4b-it-gguf: gemma3:4b-it-q4_K_M

   This allows you to add or modify models without editing the code.

2. Single Engine Translation & Multi-Engine Comparison

   • At the top of the interface, you can select a "Translation Engine" and a "Target Language", input text, and press "Run Translation" to display the result.
   • At the bottom of the screen, you can select multiple engines and toggle between Vertical or Horizontal layout to compare translation results.

3. Support for Multiple Engines

   • Selecting an OpenAI engine sends a request to the OpenAI API.
   • Selecting an Ollama engine sends a request to the local Ollama server.

Setup

1. Clone or Download the Repository

   git clone git@github.com:Onely7/local_translator.git
   cd local_translator

2. Install Required Libraries

   uv python install 3.12
   uv python pin 3.12
   uv sync --no-dev

3. Check the Model Configuration File
   In configs/avail_models.yaml, define the model mappings you want to use.
   Example:

   OpenAI/GPT-4o-mini: gpt-4o-mini
   OpenAI/GPT-4o: gpt-4o
   Ollama/gemma-3-4b-it-gguf: gemma3:4b-it-q4_K_M

4. OpenAI API Key (Optional)
   To use OpenAI, set your API key as an environment variable, like:
   export OPENAI_API_KEY="sk-xxxxxx"

5. Install Ollama (Optional)
   If you want to use Ollama, install it separately.
   You have two options for installing Ollama:

   • Download the installer from the Ollama official website
   • If you're using Homebrew, download it via Homebrew:

     brew install ollama

6. Install Local LLMs
   To use LLMs via Ollama, you need to download the models locally in advance.

   1. Find a model you want to use from the Ollama model list
      
      
      
      In this example, we will download gemma3:4b-it-q4_K_M.

   2. Download the Model

      ollama pull gemma3:4b-it-q4_K_M

How to Run

uv run streamlit run translator.py

After execution, you should see the following message in your terminal:

You can now view your Streamlit app in your browser.

  Local URL:  http://localhost:8501
  Network URL: ...

A browser window will open displaying the translation app.
If it doesn't appear, open the URL shown above in your browser manually to view the interface.

How to Use

1. Main Translation (Single Engine)

   • Select a Translation Engine from the dropdown at the top of the screen
   • Select a Target Language from the dropdown on the right
   • Enter the text you want to translate in the left column and click the "Run Translation" button below
   • The translation result will appear in the right column.
   

2. Compare Engines (Multi-Engine Comparison)

   • In the Select multiple engines to compare section at the bottom, select multiple engines you want to compare
   • Choose either Horizontal or Vertical under Comparison Layout
   • Click the "Compare" button, and you’ll see the translation results from each engine for the same input text and target language displayed together.
   

Notes

• The source language of the text is automatically detected.
  Simply specify the target language to perform translations such as Japanese → English or English → Japanese.
  Adjust system prompts or pre-instructions as needed.
• OpenAI API Costs
  Using OpenAI will incur usage fees.
  Please check the OpenAI pricing page before using it in production.
• Ollama Model Sizes
  Models run locally with Ollama can be large in size.
  Be mindful of download size and memory usage.