Answers with sources, not guesses

Your documents,finally answeringback.

Upload a PDF, a spreadsheet, slides, or a scanned page, then simply ask. LumiDoc reads every word and replies with the exact passage it found.

20+ formatsEvery answer citedGDPR-ready
annual-report-2025.pdf
Source · page 12

“Enterprise revenue grew 23% QoQ, driven by the API launch.”

94%
PDFWordExcelPowerPointImagesScanned pagesMarkdownCSVPlain textCode filesGoogle DocsKeynote
PDFWordExcelPowerPointImagesScanned pagesMarkdownCSVPlain textCode filesGoogle DocsKeynote

How it works

01

Upload anything

Drop in a PDF, Word doc, spreadsheet, slide deck, image, or scanned page. LumiDoc reads them all.

02

Split into context

Each document is broken into overlapping passages: small enough to be precise, large enough to keep meaning.

03

Understand the meaning

Every passage becomes a vector: a mathematical fingerprint of what it means, not just the words it uses.

04

Answer, with the receipt

Your question retrieves the closest passages, and the answer arrives with the exact source attached.

annual-report.pdf
Uploaded
chunk 1
chunk 2
chunk 3
chunk 4
What drove revenue growth?
Revenue grew 23% QoQ, driven mainly by enterprise API subscriptions.
Source · page 12 · 94%

Capabilities

A real retrieval engine, not a chat wrapper.

Every answer, cited

No black box. Each response links to the exact passage, page, and a relevance score, so you can trust it, then verify in one click.

Source · page 12 · 94%

“Enterprise segment delivered $4.2M in net-new ARR…”

20+ file formats

Documents, spreadsheets, slides, images, and code, even scanned pages read with Vision OCR.

PDFDOCXXLSXPPTXPNGJPGTXTMDCSVCode

Hybrid search

Semantic vectors + keyword matching, fused, catching what either alone would miss.

Streaming answers

Replies arrive word by word. No spinners, no waiting for a wall of text.

Workspaces

Group related documents and conversations. Query a single file or an entire knowledge base at once.

Private by design

Row-level security, private storage, signed URLs. Your documents never train a model and stay yours.

See it in action

The answer points back at the page.

Q3-financial-report.pdf · page 12

“Enterprise segment expansion delivered $4.2M in net-new ARR, a 23% increase from Q2…”

What was the main driver of revenue growth this quarter?
Revenue grew 23% quarter-over-quarter, driven primarily by enterprise subscriptions, 68% of all net-new revenue. The report ties this to July's API platform launch.
Source · page 12 · Relevance 94%

“Enterprise segment expansion delivered $4.2M in net-new ARR…”

Privacy & trust

Your documents stay yours.

LumiDoc is built for European teams who can't compromise on data protection. Privacy isn't a setting here. It's the architecture.

GDPR-ready by default

Built for the EU. Right to access, export, and erasure are first-class. Your data, your control.

Row-level security

Every record is fenced to its owner at the database layer. No query can cross that line.

Never trains a model

Your documents are used to answer your questions, nothing else. They are never used as training data.