Platform

Enterprise Knowledge AI Platform

Sonela connects the places your business keeps what it knows, finds what answers each question, and answers from that alone. It never changes anything, and it says so when nothing it found can answer.

In the trade that is retrieval-augmented generation (RAG), and here it retrieves three kinds of thing: rows read from your live records, passages found in your written material, and replies from your own systems. The assistant is one application built on it. Below are the four layers underneath, and the applications that run on them.

Connect

Every source your business already keeps

You attach each one from your dashboard, and Sonela reads only what you approved or gave it.

  • Your live records

    PostgreSQL, SQL Server, MySQL and Oracle, read where they live: connect directly from your dashboard, or keep the database inside your own network and run the Gateway, the high-security option and the one way Oracle connects. How a database connects

  • Your written procedures

    Handbooks, SOPs and policies as text, a spreadsheet, a Word document or a PDF, beside the context you write about your business. Start from your SOPs

  • Your website

    Name your public site and Sonela reads it, then reads it again every week. What it reads

  • Google Drive and SharePoint

    A Drive folder shared with a service account you own, or SharePoint and OneDrive for work or school through your own Microsoft app, read every hour. On Growth and Scale. Synced files

  • Your social pages

    The latest posts of a public Facebook page or Instagram profile, each answer giving the link of the post it came from. On Growth and Scale. Social pages

  • Your shop’s products

    The product feed you already make for Google Shopping or a Meta catalog, read again every day. On every plan. For shops

  • Your own APIs and forms

    Register an endpoint, or a search form your site already has, and say what it is for; when a question calls for it, the reply is read. Lookups

  • The page someone has open

    The data the page already fetched for the person asking, and a picture of the screen while they keep the camera pressed, travel with their question on the signed-in assistant. Each can be switched off. Where it stops

Ground

Answers from what was found, or a plain no

For each question it retrieves the rows, passages, posts and pages that bear on it, and the model answers from those.

  • Only from what it retrieved

    Answers come from your records and your written material, never from what a model remembers about the world. Where answers come from

  • Says so when it cannot answer

    When nothing it found supports an answer, it says so instead of guessing. That is the platform working, not failing. Grounded, or it says so

  • Every passage cited

    A number after a sentence opens the passage it came from, and the rows that are the answer sit under it. It shows its working

  • Every figure held to its source

    Each figure in an answer is checked against what the assistant was shown, by rules and not another AI; one it could not find is counted on Gaps. The figure check

  • The questions you trust, asked again

    Write the questions you know the answers to; Sonela asks them again after a change to the glossary, the schema, the manuals or your own AI keys, and every week, before and after side by side. Golden questions

  • Your own vocabulary

    The context you write for your workspace is the vocabulary it answers in, kept apart from the procedures it answers from. Two inputs, kept apart

  • A changed column, found first

    Sonela compares each database every day with the schema you approved; a changed column waits for your review, and answers keep to the approved one. The daily schema check

Govern

Rules that hold whatever the model does

Enforced in code, outside the AI, and written up for whoever has to sign it off.

  • Read-only, enforced three times

    A grammar that cannot express a write, the database’s own guard, and a rollback rather than a commit. The security brief

  • Each customer sees only their own

    Every query is rewritten to the asking customer’s rows before it reaches your database, and the model never sees who that customer is. Isolation by rewrite

  • Personal data, masked in flight

    Before the AI model — yours or ours — reads a question, your records or your systems’ replies, the personal details Sonela recognises become placeholders. Where it stops

  • Nothing written down

    No question, answer or query is kept, and usage is counted rather than read; the exceptions are an answer someone reports and chooses to send you, and a conversation someone hands to your people. What Sonela can see

  • What ran, on every database answer

    Every answer that read your database opens to show the query the AI wrote beside the one your database actually ran. What ran

  • A record of refusals and changes

    Your admin sees what was refused and every change anyone made to the setup. Your controls

Deliver

Where people reach it, and on whose key

The same rules on every channel; the table on How it works marks each difference.

  • In your software

    One line of code puts the assistant inside your application, where your staff already work. The channels

  • On a page we host

    No change to your application: invite people by email, and they sign in with a short code. The hosted page

  • On their phone

    Apps that wrap the hosted page are built, and not in the stores yet. The phone

  • On your public website

    A key marked public answers anyone on your site with no sign-in, deliberately narrower, and only on your own website. Public assistant guide

  • One key for every customer you serve

    For a software vendor: your server signs who is asking, your pages carry one tag, and each of your customers sees only their own. ISV guide

  • Your AI key, or ours

    Bring your own and your rows reach only the provider you chose, under your own agreement; run on Sonela’s key and there is nothing to set up, which those plans price in. Both, side by side

Applications

What runs on the platform

Five applications on one set of sources and one set of rules.

  • The assistant

    People ask in their own words and get the answer, its sources and its working: in your software, on a hosted page, on their phone or on your public website. Where it runs

  • The API and integrations

    Your own software asks over the API, on secret keys you issue and revoke; Slack, Microsoft Teams or a tool of yours reach it through a small service you run. Integrations guide

  • Documents from your templates

    Asked for a document, it fills your Word template from the conversation, shows every value, and on a plain yes emails the file. On Growth and Scale. Templates

  • Product recommendations

    Visitors ask what to buy and get the products as cards, with picture, price and link. It recommends; your shop sells. For shops

  • A knowledge base on any subject

    Switch a key to knowledge base and it answers on any subject your documents cover, from them alone. Knowledge base

The record

What it keeps, and what it will not

Your written material, synced files and social posts, and the products of a feed you add are kept in your workspace, because they are what it answers from. No question, no answer, no row of yours is written down here — except what a visitor chooses to send you. The security brief lists each, and What ran shows the query behind every database answer.

Start with one source, and add the rest

Open a workspace, connect one source and ask it something. Write to us first if you would rather talk it through.

14 days, 320 question credits, no card. No AI key to supply — trials run on Sonela's key.