Show HN: Loqi, a "local-first" translation tool using Ollama/llama.cpp (github.com)

A new local-first translation tool named Loqi has been announced, utilizing open-weight large language models (LLMs) via Ollama and llama.cpp. This project aims to mitigate reliance on established services like Google Translate and DeepL by allowing users to perform translations on their own machines, ensuring that no personal data is sent to third parties. Loqi supports various backend models and offers three interaction modes: a terminal user interface (TUI), a command-line interface (CLI), and a batch processing mode, all customizable through configuration options like model selection and prediction parameters. The significance of Loqi lies in its promotion of privacy-focused translation solutions while leveraging the contextual understanding capabilities of LLMs. Although it may not be as fast as conventional engines, the tool is designed to produce draft translations that users can refine. Loqi's architecture allows for flexibility by accommodating different backend services, whether leveraging LLMs or more traditional rule-based translation options like Argos-Translate. This hybrid approach not only enhances user control over their data but also broadens the accessibility of language translation technologies in a privacy-conscious manner.