If native text extraction returns nothing usable, the page is sent through a vision-capable model that transcribes the content.

Can I see exactly which page the answer came from?

Yes. Every answer cites file + page; the chat surfaces clickable references.

Where are my documents stored?

In your project's encrypted storage. Delete the folder source to remove the files.

Can I combine document search with SQL?

Yes. The same chat can answer 'how many incidents in the DB last month' AND 'what does the safety policy say about reporting them'.

How big can a folder be?

Configurable per workspace. Default supports thousands of pages comfortably.

Document RAG

Document Q&A — search PDFs, contracts, specs with real citations.

Upload a folder of documents. iDBQuery extracts the text (vision-OCR for scans) and the AI answers your questions with file + page citations.

Start free See it for construction

Documents / Search

“termination clause vendors”

live

termination clause vendors42 docs

Top citations

Vendor MSA — Section 4.2 Termination

92%

MSA_Atlas_v3.pdf · p.14

Either party may terminate for material breach with 30 days written notice...

Engineering RFP response — pricing addendum

81%

RFP_Atlas_response.docx · §6

Discount tier of 12% applies above 500 seats, billed annually, net-30.

Quarterly board memo — Q3 risk register

67%

Q3_board_memo.pdf · p.7

Top mitigations include vendor consolidation and renewing the legacy contract...

Mock preview

The why

Documents shouldn't be a black hole

PDFs eat answers

The clause you need is on page 87 of a 200-page contract. Ctrl-F doesn't help.

Scans aren't searchable

Half the document set is scanned PDFs that defeat normal text search.

Citations matter

An AI answer without a source is a guess. You need to click through to the page.

How it works

From folder to answer in four moves

Add a Folder source

Drop a folder of PDFs / Word / Excel files into a project. Sync runs in the background.

Extraction + OCR

Native text is extracted; scanned PDFs go through vision-OCR fallback so even image-only docs are searchable.

Ask in chat

The AI calls search_documents to retrieve relevant passages, then synthesizes an answer.

Get citations

Every answer includes file + page references — click to jump to the original.

Capabilities

Everything the document layer does

PDF, DOCX, XLSX

Mainstream document formats supported out of the box.

Vision-OCR fallback

When native text extraction fails (scanned PDFs, image-heavy slides), a vision model reads the page.

File + page citations

Every passage in an answer carries a file name and page number — click to open the source.

search_documents tool

The AI's RAG entry point: it phrases an internal query, searches the index, and grounds its answer.

Mix with structured data

Document RAG sits alongside SQL tools — one chat can answer 'what does the spec say about X' and 'what's the SQL count of X'.

Folder sync

Drop new files in, they're indexed. Re-sync on demand or on schedule.

Who's it for

Legal / Procurement

Searches 200-page contracts for the obscure clause in seconds.

Construction PM

Asks the spec library 'what's the fire-rating requirement for this assembly?'

Operations

Pulls the right SOP without remembering which folder it lives in.

Try this

Sample prompts

>What does the master service agreement say about termination?
>Find all clauses mentioning 'indemnification' across our standard contracts
>Summarize the safety section of the latest site-induction document
>Which spec sections cover concrete strength?

Frequently asked questions

PDF (text + scanned), DOCX, XLSX, plus plain text/markdown. More formats added on request.

Ready to chat with your data?

Free tier covers most teams. Connect a database in 30 seconds.

Start free Contact us

No credit card required · 1M tokens / month free

Loading…