Nucliadb

NucliaDB is the database platform that Progress Agentic RAG uses to store and index data. It can be run on its own, locally or on premises, and connected to Agentic RAG Cloud so that the user keeps ownership of the data while using the cloud service's Understanding API for data extraction, enrichment and inference. Data is organized into Knowledge Boxes, which are containers holding resources. A resource can combine text, files, vectors, labels and annotations. NucliaDB supports hybrid, semantic, keyword, full text and graph searches, makes it easy to compare vectors produced by different models, and lets users access and modify resources efficiently. Data can be exported in formats compatible with most NLP pipelines, including HuggingFace datasets and PyTorch. The documentation walks through a quick start. A PostgreSQL server is required, and NucliaDB runs from a Docker image or through pip with an API key for the Understanding API. A Python SDK creates Knowledge Boxes and uploads resources, and files can also be uploaded over the HTTP API with curl. It is aimed at developers building retrieval augmented generation and search applications who want a self managed data layer for unstructured content. The documentation page shows no prices, so pricing for the software and for the connected cloud service is not published there.