Guides · 09

Files and documents

Your agents can only reason about what they can see. The Files page is where you hand them documents - and where answers start coming with sources.

narrated · sound on

Why this matters

An answer with a source is worth ten without one

When Cronus answers from your documents, it names the file it used. That single habit turns AI from a guessing machine into something you can check - and checking is what the disclaimer on this site keeps telling you to do.

The walkthrough

Drop a file in

Drag any .txt, .md, or .pdf onto the Files page, or click to choose one. Contracts, reports, invoices, notes - once in, they are searchable by the whole stack.

Extract facts from a file

The built-in extractor turns a messy document into clean, structured facts - feed it a scanned invoice and get the vendor, amount, and due date, ready to use.

Ask about a folder

Point it at a directory and ask questions about what is inside - useful for a project folder, a client folder, or that pile of PDFs you have been meaning to read.

Ask in chat, get cited answers

Once files are in, just talk to Oracle. When a question touches your documents, Cronus searches them and answers with the filename it drew from - "according to invoice_march.pdf" - so you can verify in one click.

Local, as always

Files live on your machine. Content only leaves if you route a question about it to a cloud model - your choice, per question, and local models exist for exactly the files that should never leave.

← Previous guide Next guide →