notebooklm-mode

Source-grounded research involves collecting and indexing verbatim source extracts to answer questions based strictly on those specific sources, ensuring responses are well-reasoned and verifiable.

Yes, you can build a searchable knowledge base by storing verbatim source extracts and search indexes in a customizable vault, utilizing an optional semantic search engine for querying large volumes of content.

Semantic search over a large research vault works by indexing stored verbatim source extracts, allowing the system to retrieve information based on conceptual meaning rather than exact keyword matches.

Yes, it supports parallel task execution through a multi-mode design: direct mode runs all roles inline, while subagent mode spawns separate agents for parallel research processing.

The best way to synthesize knowledge from multiple web sources is using augmented grounding, which combines strict source grounding with labeled background knowledge to generate comprehensive, well-reasoned responses.

You should use subagent mode instead of direct mode when handling parallel research tasks or multiple user requests, whereas direct mode is suitable for running all roles inline within a single session.