Product guide
Settings, operations & compliance
Settings, branding, email, background jobs, the audit trail, and the security & privacy facts.
Settings, tab by tab
General
| Branding | Your company name and logo — see Branding. |
|---|---|
| Language | The system language for menus, sign-in pages, emails and system messages — see “System language”. The assistant itself always answers in the asker's language. |
| Email (SMTP) | The outgoing mail server used for invitations and password resets — see Email. |
| Public address | The base URL this system is reached on. It is what emailed links point at and what platform webhooks (WhatsApp, Slack…) are built from — set it to the address your users actually use, especially behind a reverse proxy. |
The page saves as one unit — the save bar appears once anything is edited.
Security & privacy
Measured facts about who can read what, accounts and sign-in, the tamper-evident audit chain, retention, and where data lives — plus GDPR tools (erasure, subject export) and the live compliance manifest. See “Security & privacy” and “Audit log”.
Background activity
A read-only feed of the background jobs (syncs, document analysis, retention). Jobs run and repair themselves — nothing there needs attention. See “Background activity (jobs)”.
Deployment-level configuration (database, retention periods, model profile) is deliberately not editable here: it lives in the deployment's .env, and the Security & privacy tab shows it read-only so the compliance manifest can never drift from reality.
Branding
Set your company name under Settings → Branding. The assistant is shown to users as “<Your company> Assistant” across the sign-in page, chat, admin and emails.
Upload your logo there too — it becomes the browser icon, the menu logo and the chat-bot picture where the platform allows it. PNG, JPEG, WebP or GIF under 440 KB; a square mark works best.
System language (and the languages the assistant speaks)
Two different things speak two different kinds of language here, and only one of them is a setting.
The system language — a setting
Menus, buttons, the sign-in and activation pages, invitation and reset emails, the website widget's frame and the chat-bots' system messages (help texts, linking instructions) follow the system language. It is chosen in first-run setup and changed any time under Settings → Language. English, Greek, German, French and Spanish are available.
Each person can override it for themselves: the Language entry in the user menu (bottom of the sidebar, in the chat and the console) sets a personal display language that beats the instance default on their account only.
The assistant's languages — not a setting
The assistant understands and answers in whichever language each person writes, regardless of the system language: a Greek question gets a Greek answer, an English question an English one — in the web chat, on every messaging channel, and over documents written in any language. There is nothing to configure.
The help centre's articles are written in English; the help AI beneath them answers questions about the product in your own language.
Email (SMTP)
Configure an SMTP server so invitations and password-reset links are emailed automatically. It's optional — the links are always available to copy from the Users page, so you can run Naxis Assistant without email. Use Send test to check the settings.
| Host / Port | Your provider's SMTP endpoint — typically port 587 with STARTTLS. |
|---|---|
| Username / Password | The sending account's credentials; providers with two-factor authentication issue app passwords for this. |
| From address | What recipients see — e.g. assistant@acme.com. |
What email unlocks
- Invitation and password-reset links are delivered automatically instead of being copied by hand.
- “Forgot password?” on the sign-in page works on its own — a single-use link, valid 24 hours, sent to the account's email.
- Administrator sign-ins gain a second factor: after the correct password, a 6-digit single-use code (valid 10 minutes) is emailed and must be entered. If the mail server is down, the sign-in falls back to password-only and the failure is recorded in the audit log — administrators are never locked out by an outage.
The Email (auto-reply) chat channel also sends its replies through this server — configure SMTP before setting that channel up.
Background activity (jobs) — what runs on its own
Syncing, document analysis and retention don't happen while you watch — they run as background jobs: queued, executed by the server, retried if they fail. You never manage jobs. There is nothing to start, stop, cancel or retry — the queue tends itself, and Settings → Background activity simply shows you what it has been doing.
The three kinds of job
| Sync (sweep) | On a schedule, and whenever you press Sync: each source is compared with the knowledge base and only what changed is re-processed. |
|---|---|
| Document analysis | New and changed documents get their card — date, kind, entities, summary, authority — right after each sync. |
| Data retention | Every night, conversations and audit events older than the deployment's retention periods are deleted. The Security & privacy tab shows the last run. |
Reading the activity feed
| running… | Being executed right now. Big first syncs can run for a while — that's normal. |
|---|---|
| waiting | Queued; the server picks it up within seconds (or at its scheduled time). |
| completed | Finished successfully. |
| failed | Gave up after its retries — the reason is shown on the row, and the next scheduled run takes over. A source that keeps failing also appears under Needs attention on the Overview, with the button that fixes it. |
What if a job fails or gets stuck?
- A job that fails retries itself with increasing pauses, up to three attempts.
- A job can't get stuck: running jobs report a heartbeat, and one orphaned by a crash or restart is detected and re-queued automatically.
- A job that keeps failing stays marked failed — and the next scheduled sync supersedes it, so the knowledge base catches up on its own. Fix the cause it names (usually a source's credentials) and the next run succeeds.
So: what do you do with jobs? Nothing — that's the design. If a sync can't wait for the schedule, press Sync sources (Overview) or a source's own Sync now; that just queues the same job early.
Audit log
An append-only, tamper-evident record of every question and admin action, under Settings → Security & privacy. Each event is hash-chained to the one before, so nothing can be edited or removed unnoticed. “Verify hash chain” recomputes the whole chain to confirm nothing has been altered — run it whenever you like; it takes seconds.
Filter by event type or actor to answer questions like “who changed this document's access” or “what did the assistant retrieve for that answer”. Each query event records which passages were retrieved and which were cited — the full evidence trail behind every answer.
Security & privacy
Settings → Security & privacy gathers everything that makes the system trustworthy: the tamper-evident audit log, automatic retention, a generated processing manifest (GDPR Art. 30), erasing a person's or a document's data (Article 17), and exporting everything held about a person (Article 15). Green means working — none of it needs day-to-day attention.
| Erasure (Art. 17) | Removes a person's conversations, or one document, everywhere — index included. A hard delete, itself recorded in the audit log. |
|---|---|
| Subject export (Art. 15) | Everything stored about a person — account, conversations, audit trail — as a JSON file they can be handed. |
| Compliance manifest | The processing record (purposes, retention, sub-processors, security measures) generated live from the running configuration, so it can never drift from reality. |
| Deployment configuration | Read-only view of the .env settings this deployment runs with. |
This help centre (and its AI)
Everything in this window is about how Naxis Assistant itself works — it never searches your company documents (that's what the chat is for). Three ways in:
- Browse: the outline on the left covers every feature in detail.
- Search: type in the box above — answers rank instantly as you type, entirely on your device.
- Ask the AI: the bubble in the corner opens a conversation grounded in these very articles. Ask in your own words, ask follow-ups; every answer cites the articles it drew from, and clicking a citation opens the article.
The small ? marks throughout the product open a short explanation right where you are, with “Learn more” leading back here.
The AI answers only from these articles. If something about the product isn't covered, it says so — and if your question is about your own documents, ask in the chat instead.