AutoGluon-RAG
A framework that follows the AutoML philosophy to let a developer build a customized retrieval-augmented generation pipeline in a few lines of code, without handling each module separately.
AutoGluon-RAG streamlines building retrieval-augmented generation pipelines, which normally require navigating retrievers, generators, vector database construction, fast semantic search and long-context handling as separate pieces. Following the AutoML philosophy used elsewhere in the AutoGluon project, it abstracts those modules behind a simple interface so a user can build a working RAG pipeline over their own documents with minimal code, focusing on domain requirements instead of the underlying implementation.
It can be used from the command line through its own tool, pointed at a configuration file, or called directly from Python code, and installs by cloning the repository and installing it into a virtual environment. A related data-pruning framework for efficient retrieval-model fine-tuning, developed alongside it, was accepted at COLM 2026.