artitw/text2text

Provides unified APIs for multilingual text processing tasks including LLM inference (with streaming and structured output support), embeddings, semantic search via TF-IDF/BM25 indexing, machine translation, and data augmentation through back-translation. Built on transformer-based models with pluggable pretrained backends (defaulting to Facebook's M2M-100), operating efficiently on commodity hardware and free Colab resources. Exposes sub-word tokenization, language identification, and edit-distance calculations alongside a conversational assistant interface compatible with OpenAI's chat completion schema.

303 stars and 606 monthly downloads. No commits in the last 6 months. Available on PyPI.