Documents and memory
Keep policies and notes on a graph, search them, and store memories your AI assistant can recall later.
A graph holds documents as well as numbers. Your AI assistant can search them, quote them, and, if you ask it to, save a policy or a close procedure so it is there the next time.
Knowledge Base
Knowledge Base lists the documents on the selected graph, whether you created them here or your assistant saved them. Each is markdown, with an optional folder and tags. Anyone who can write to the graph can create, edit and delete them. Memories are kept separately, on the Memory page.
Shared repositories such as SEC filings don't have a Knowledge Base of your own. Their filing text is searched through Search and through an MCP client connected to that graph.
Creating and editing documents uses no credits. Neither does search.
Search
Search finds text across the documents in the graph. On the SEC filings graph it searches filing narratives. It matches keywords; turn on Semantic search to also match by meaning.
Memory
Memory is a store on each of your graphs for things your assistant should remember across conversations: a mapping decision, a counterparty you always treat a certain way, a note about a close. You can add, browse, edit and delete memories in the app.
Memory is not available on shared repositories. A subgraph doesn't hold memories; store them on the parent graph.
Your assistant can also remember and recall through MCP. That uses no credits. The Console uses credits because it is RoboSystems' own AI; see Ask questions in the Console.
Try asking
- "Save this as our capitalization policy."
- "What close procedures do we have documented?"
- "Remember that we treat [vendor] as a contractor, not an employee."
- "What do you remember about how we map owner draws?"