Book Translator: Two-pass local translation with self-reflection via Ollama (github.com)

The newly launched Book Translator app utilizes an Ollama-powered architecture to facilitate local, long-form document translation through a two-stage workflow. Initially, the software generates a draft translation before executing a second pass to enhance fluency, consistency, and overall style. This approach is tailored for users seeking a local-first interface that prioritizes tracking, history retention, and the ability to download refined translations without the need for intricate prompt setups. Significantly, this project addresses common challenges in long-document translation, such as the need for chapter chunking, retries, and visibility into progress. The modular Python application runs a Flask API that handles uploads and job management, while a static frontend provides an intuitive user interface. With features like translation history, cache management, and performance metrics, the Book Translator fulfills a key need in the AI/ML community for accessible, efficient translation tools, enabling users to produce polished translations even from complex, lengthy texts.