sid-swirl wires the assistant into the firm's filing cabinet

Two new tools let the project assistant search outside its uploaded documents and pull live results back in as first-class files.

searchknowledge-management

The first tool runs federated searches against external sources - case law, regulations, sanctions lists, company filings, and a firm's own Microsoft 365 estate (OneDrive, SharePoint, Outlook, Teams). A short prefix on the query scopes results to a specific connector. Results come back ranked, with titles, snippets, and source names.

The second tool takes any result and imports the underlying PDF or Word file straight into the active project, where the assistant can then read it like any other uploaded document. The plumbing runs through SWIRL, an open-source federated search layer that holds the OAuth tokens for authenticated sources - the AI never sees the credentials. If a firm hasn't set SWIRL up, the tools quietly say so instead of breaking. The work is currently a pull request against willchen96's fork.

So what For firms whose knowledge already lives in M365, this is the difference between an assistant that reads what you upload and one that can go find the document itself.

View this fork on GitHub →

Spotted something wrong? Or know the PR text has fresher detail than the writeup above?