xablin/hf_en-ru-04_me_cli

Purpose and Entry Points

This is a local EN↔RU machine-translation application built on Hugging Face Marian/OPUS models. It supports an interactive console workflow and two desktop GUI implementations. main.py is the primary launcher, selecting mode and translation direction; direct UI entry points also exist in translate_cli.py, translate_gui_tk.py, and translate_gui_qt.py.

Startup and Dependency Gate

main.py coordinates startup: parses launch options, invokes dependency validation, initializes settings, verifies/downloads models, then starts the selected UI. dependency_checker.py checks required Python packages and Tkinter availability before application execution. Runtime dependencies are declared in requirements.txt.

Translation Core

translator.py is the central application service. It owns direction-aware model selection, Hugging Face tokenizer/model loading, device selection, and text translation. The supported model boundary is Hugging Face model IDs for Helsinki-NLP OPUS Marian EN→RU and RU→EN models, cached locally by the Hugging Face stack.

Model Lifecycle and External Services

The translator checks whether required models are locally available and can interactively download missing artifacts from Hugging Face Hub. Authentication is optional but supported through the HF_TOKEN environment variable. Model files are stored in the standard Hugging Face cache (~/.cache/huggingface/); inference uses PyTorch and may use CUDA when available, otherwise CPU.

Interface Adapters

The UI layer is separated by runtime:

• Console interactive translation loop: translate_cli.py

• Tkinter desktop UI: translate_gui_tk.py

• Qt desktop UI: translate_gui_qt.py

• Shared/help-oriented dialog surface: help_dialog.py

These adapters collect source text and direction choices, invoke the translator service, and present results. They should not own model lifecycle or inference policy.

Configuration and Persistent User State

config.py defines application-level defaults and constants, including translation directions and likely model/UI settings. settings_manager.py manages persisted user preferences and translation-history-related state, providing a boundary between UI interactions and local storage behavior. Relevant paths: config.py, settings_manager.py, translator.py.

Primary Control Flow

1. Launch through main.py with optional direction/UI arguments.

2. Validate local dependencies through dependency_checker.py.

3. Resolve settings and translation direction via settings_manager.py and config.py.

4. Check/download and load the selected Marian model in translator.py.

5. Run a CLI or GUI adapter, repeatedly passing text to the loaded translator.

6. Return translated text to the active interface; retain user state/history through the settings layer where enabled.