Database Explorer
You give the focus. The AI answers to the point.
Every knowledge base drawn as a graph, a map and an overview — so you choose what the AI should read before it answers.
Why focus wins
A question asked of a whole database gets an answer assembled from whichever passages happen to match the words. Ask it about the right node instead — one composition, one client, the two reports that matter — and the AI reads those documents in full before it writes. You bring the judgement only a person has; the AI does the reading.
In the recording a research group asks for one figure. Asked of the whole database, the answer is a vague interim value; asked of the selected study, it is the exact measured result, cited to its page.
Human-enhanced
Your selection decides what is read. The answer is as good as your focus — and much better than a guess over everything.
Every claim traceable
Each sentence links to the passage it rests on; open it in the document with the words highlighted.
Word files from the chat
“Write a 10-page report on this client” becomes a real .docx — outline, sections and sources — ready to download.
Deep research on a node
Run a full deep research limited to the documents of the node you picked.
Choose your situation
A quality manager checks whether a complaint is really closed.
Every customer, supplier and incident — one click away.
Complaints, 8D reports, audits, contracts, quotations, minutes: the explorer reads them all and recognises your customers and suppliers. Click the customer, ask whether the problem is really closed — the answer cites the report and the review that say otherwise.
Three ways to look at your documents
See the shape of what you know.
Graph
Documents linked by meaning, gathered around their themes, tied to the people, organisations, places and keywords they share. Hover to light up a neighbourhood; click to open it.
Map
Every passage as a point, placed by meaning, on islands per topic. Hover to read; draw a lasso around a region to ask about exactly those passages.
Overview
Themes, kinds of documents, timeline, people, which document covers which topic — and the files worth a look: near-duplicates, isolated or empty ones.
Answers you can check
Every answer shows where it comes from.
Under each answer: what was searched, how much of it the model read, which documents the answer is made of — in proportion — and a sentence-by-sentence check that flags anything no passage supports.
Sources light up
The passages and documents behind the answer pulse on the graph and the map, numbered like the citations.
Made of which documents
A bar shows how much of the answer each document contributed, and what part is not backed at all.
Sentence by sentence
Each sentence is matched against the passages: supported, partly, or not found in the sources.
Read all of it
With a focus set, one switch hands the model every passage you selected, not a sample.
See your own documents this way.
We set up a demo with a sample of your files — on your hardware or ours — and show you what the explorer finds.
- Your documents stay on your server
- No cloud upload
- Access follows your database rights
Questions
Does it work with our existing databases?
Yes. Every Butler database gets an explorer automatically — nothing to configure. The first build takes a minute or two; after that it updates itself when files change.
Which files can it read?
Everything Butler indexes: PDF (including scanned, with OCR), Word, PowerPoint, Excel, text and web pages archived with Web Vacuum.
How large can a database be?
Hundreds of documents and thousands of passages: the map and the graph stay fluid in the browser, the heavy work runs on the server, and each file is read by the AI only once.
Who can see a database’s explorer?
Exactly the people who can open the database: its owner, the group it is shared with, and administrators.