nucliadb

Why we included this project

NucliaDB is a self-hosted search database aimed at the storage and retrieval side of RAG pipelines over unstructured content. It combines vector, full-text, and graph indexes in one system, so semantic similarity lookups and plain keyword queries hit the same corpus without you maintaining separate stores. The Rust and Python codebase targets large datasets and multi-tenant deployments, and it keeps original files, extracted text, and vector embeddings together in one place. It can also export indexed data in formats that feed directly into common NLP training pipelines, which is handy when you want to reuse content beyond search. If you're building a search backend that needs to grow with a real RAG application rather than a demo, this is worth a look.