Secure Intelligence for the Solidarity Economy
Rook is a private, Retrieval-Augmented Generation (RAG) platform engineered to address data privacy and client confidentiality.
Unlike public AI tools that process data in a shared cloud environment, Rook operates as a single-tenant deployment. This means your organization's documents are housed in a dedicated digital vault that is not shared with any other users.
By maintaining this boundary, Rook ensures that sensitive case files and privileged communications remain under your organization's control and are never used to train public models.
By hosting this infrastructure on your own hardware, you reduce the risk of external data leaks and maintain an auditable environment that could meet the standards required for handling protected health information (PHI) or sensitive litigation data. Note that Rook is in closed beta and a security audit is still forthcoming.
How it works
The system functions through a four-stage architectural flow designed for transparency.
- File Browser A secure interface where users upload documents, notes, contracts, or research.
- Indexer Rook uses an open model Indexer to convert documents into embeddings (mathematical representations of the information) stored in Qdrant, a vector database.
- Semantic search When a query comes in, Rook performs a semantic search within your private database, retrieving the most relevant passages from your own files.
- Rook response Answers are based on your source documents, significantly reducing hallucinations and providing a link between output and source.