Product guide

How it works

What the assistant is, how it turns your documents into cited answers, and why those answers can be trusted.

First-run setup & the guided tour

A fresh install greets you with a short wizard before the console opens: system language, the administrator account, workspace branding and email, document sources (with “Import your team” where the platform knows your people), team accounts and groups, and chat channels.

Every step except the administrator account can be skipped — the wizard configures the same runtime settings the console edits, so anything skipped is done later under Documents, Users & groups, Chat channels and Settings. “Skip setup for now” (bottom left) jumps straight to the console.

The guided tour

When setup finishes, the console opens with a guided tour: each step spotlights one part of the screen and explains it in a small card beside it — the Overview signals, sources, channels, people and access, the Settings tabs, and this help centre. Use the arrow keys or the buttons; Esc leaves at any point.

Rerun it whenever you like — “Guided tour” sits in the sidebar, just under Help.

How Naxis Assistant works

Naxis Assistant answers your team's questions using only your company's own documents and data, and always shows the exact passages behind each answer. As the administrator you control two things: where knowledge comes from (Documents), and who may see what (Users & groups). Everything else — syncing, indexing, retention — runs itself.

Everything a person asks is checked against what their groups allow. The assistant never answers from a document the asker isn't permitted to see, and when the answer isn't in the documents it can reach, it says so instead of guessing. How that works mechanically is described in “How answers are made”.

The Overview page, tile by tile

How answers are made (and why they can be trusted)

When someone asks a question, the assistant does not simply look up similar text and paraphrase it. It runs a short investigation over the knowledge base — restricted, at every step, to documents the asker is allowed to read.

1. Access first: the asker's groups, personal grants and email grants are resolved into an access list. Every search that follows is filtered by it — passages the person may not read never even reach the model.
2. A first search runs immediately, combining meaning-based (semantic) and exact-word (keyword) matching, tuned by who is asking and any dates the question mentions.
3. The assistant reads the results and decides whether it can answer. If not, it investigates: it searches again with different wordings, opens full documents when an excerpt isn't enough (contracts, agreements), and lists everything about an entity or a period for timeline and “all of the …” questions.
4. It writes the answer with numbered citations [1], [2] … — each number pointing at a specific passage it was actually shown.
5. A mechanical grounding check runs last: every claim must carry a valid citation. Invalid citations are stripped; if none survive, the user gets an honest “I don't have that in the knowledge base” instead of a confident guess.

Rules the assistant answers by

• Evidence has rank: a signed instrument beats a report, a final report beats meeting notes, and everything beats drafts and stale copies. Between comparable documents, the newer governs.
• Dates are computed, not guessed: every date in the evidence is annotated with exact, machine-computed distances (deadlines passed, renewal windows), because language models are bad at calendar arithmetic and this system doesn't let them do any.
• Premises get corrected: if the question asserts something the documents contradict, the answer says so explicitly.
• It answers in the asker's language, whatever language the documents are in.

What users see

Under each answer: the numbered sources. Clicking one opens the exact passage, so anyone can verify any claim in seconds. Follow-up questions continue the same conversation — the assistant remembers the thread and re-checks it against the documents rather than trusting its own memory.

How documents become searchable

Between “file in a drive” and “cited in an answer” every document passes the same pipeline, fully automatic:

1. Sync (sweep): each source is compared against the knowledge base — new, changed and removed items are detected using the provider's own change markers, so only what actually changed is re-processed.
2. Conversion: Word, PDF, Excel, PowerPoint and the rest are converted to clean text; scanned PDFs are OCR'd automatically.
3. Chunking: the text is split into passages (“chunks”) that follow the document's own structure — headings, clauses, tables — because those are the units answers cite.
4. Indexing: every chunk gets a meaning vector (embedding) plus a keyword index entry, so search matches both ideas and exact terms like invoice numbers.
5. Analysis: each document gets a card — its date, kind (contract, invoice, report…), the people and companies in it, a one-line summary, and an authority rank used when documents conflict. This is what powers timeline and “everything about X” questions.

Reading the Documents table

Documents

The Documents page is the knowledge base seen by source: connected sources first (each card shows its live health and document count), then the catalog of everything you can connect. Click a source to open it — its setup steps while it's new, then the documents themselves.

• Upload adds a file to the built-in Uploaded documents source — only the uploaded file is indexed, not the whole source. Markdown and text ingest directly; Word, PDF, Excel, PowerPoint and scans are converted (scans are OCR'd).
• Access changes who can read a single document — uploaded documents right here in their list, and ANY document (cloud too) from a person's file map: Users → Edit → View file map.
• A document that couldn't be read shows a red reason — replace or remove it.
• Sync all re-checks every source now; each source also has its own Sync in its ⋯ menu. Syncs run in the background — watch them under Settings → Background activity.
• Removing a source removes its documents from the knowledge base; the originals in the connected account are never touched.

What the source card signals mean