Skip to main content

​
What lives in it

  • Hand-written documents.
  • Imported content from your team’s existing tools (with the right connector).
  • Synthesized rollups — Q&A pairs, retrospectives, summaries.
  • Agent diary entries — decisions and reasoning captured as work happens.

​
How it’s organized

Every document has a scope — a path like cerebrum/components/worker or your-project/overview. Scopes form a tree: the first segment is a domain (a product, a project, a topic), and deeper segments narrow down within that domain. Agents filter retrieval by scope prefix. Asking for context “under cerebrum/” returns only Cerebrum-scoped documents. Asking under another project’s scope returns only that project’s content.

​
How retrieval works

Documents are chunked by heading and embedded into a vector database. When an agent needs context, the platform pulls the most relevant chunks — filtered by sensitivity tier, scope, and time — and includes them in the prompt with citations back to source documents.

​
How content stays current

  • Manual edits through the admin surface. Edits sync to the search index within about a minute.
  • Connector pulls that sync from your team’s other systems on a schedule.
  • Agent diaries that capture decisions automatically as workers execute jobs.
Once a document is in the hub, every agent’s next answer can use it.