Knowledge & memory
Knowledge base
Point Anvil at your documents and build a durable, queryable library that never leaves your Anvil.
Your organisation sits on knowledge — PDFs, Word docs, spreadsheets, notes. The Library turns that into a durable, searchable knowledge base Anvil can draw on, held entirely on your Anvil.
What it does
Add a source — a local folder or a remote store — and Anvil scans and indexes it. From then on, that knowledge is queryable with citations: ask a question, and Anvil answers from your documents and shows where each answer came from.
- Scan & index — PDFs, Office docs, text, and more, read and made searchable.
- Search by meaning — find documents that say the same thing in other words, not just keyword matches.
- Citations — every answer points back to the source document.
- Classifications — mark a collection Public, Internal, Confidential, or Restricted; the host answers for the person asking.
[!NOTE] The Library lives on your Anvil — documents never leave it. Add a collection from your command centre, point it at your sources, and start querying.
Adding a source
Create a collection, then add sources to it:
- A local folder on the machine your Anvil runs on.
- A remote store — WebDAV, S3, SFTP, or anything rclone reaches (which covers Google Drive, OneDrive, Dropbox, Box, and more).
Give it a moment to scan, and the documents become part of the knowledge base.
Sovereignty
The whole point is data sovereignty: the documents, the index, and the answers stay on your Anvil. The hub never sees a document, a snippet, or a source path.
[!NOTE] The hub never sees your documents. The Library is scanned, indexed, and queried entirely on your Anvil.
Related
- Memory — the agent's own durable memory, distinct from the document library.
- Security & vault — where credentials for remote sources are kept.