ABACC RAG stands for Attribute-Based Access Controlled Context for Retrieval-Augmented Generation fostering an organization-level AI sovereignty.
Most of the retrieval examples usually consider the user to feed the LLM with a series of documents that they deemed relevant to their prompting activity. While this approach has a direct value for a single user, it isn't at an organizational level. However, the most complex challenge with organization-wide knowledge is who can access what, from where, and when.
The ABACC RAG illustrates the ability to leverage enriched metadata, with, in this example, the usage of extended attributes at a Unix/Linux file system level, or managed at scale with a third-party enterprise solution like NetApp BlueXP Classification.
ABACC RAG was prototype to showcase a Copilot for Microsoft Office 365 replacement with an organization-level AI sovereignty architecture.
The components are:
- a Python backend:
- ingesting data and metadata (including extended attributes) from a given directory in ChromaDB
- retrieving an ingested document list via an API endpoint
- retrieving context filtered by the attribute-based access control and submit the context and prompt to the model via an API endpoint
- a web frontend for a classic off-application chatbot
- a Microsoft Word Add-in for in-application chat and insertion
For more details, you can go through the documentation