Reads your actual documents

Hand Nia a PDF, a Word document, an Excel spreadsheet, or a PowerPoint deck, and it doesn't just see a filename — the real content gets extracted server-side into text Nia can actually read, search, and reason about.

That's a meaningfully different thing from a model that can only describe a file by its name and size. Nia can answer a question about what's on page 12, pull a number out of a specific sheet, or summarize what a 40-slide deck actually argues.

Extracted content doesn't disappear once you close that conversation. It becomes part of a searchable knowledge base, so a document you shared weeks ago can still surface in a completely different chat, when it's actually relevant — not something you have to re-upload every time you need it.

Files you give Nia live in the Vault — storage on your own hardware, not an upload to a third-party file service. It's the same storage whether it's a document for Nia to read, or a bot's own working files.