The collections
The repositories and portfolios the system draws on.
How it works
The system is a retrieval augmented generation (RAG) system. It finds the most relevant passages in an indexed knowledge base, then composes its answer from them and points back to the source.
Ingest and chunk
Documents are parsed with standardized metadata and split into passages, so answers can point to the exact section.
Embed and index
Passages become vector embeddings stored in a semantic index, so search works by meaning rather than exact wording.
Retrieve (semantic search)
Your question is embedded the same way and matched to the most relevant passages and their metadata.
Generate with grounding
A frontier large language model writes the answer from the retrieved passages and aims to cite where each result comes from.
Good to know. Documents remain the property of their original publishers and are subject to their own licences. The assistant summarizes and links to them, and it can make mistakes, so please check important details in the original source.