ElkwaiExtract and search
Native PDF text where available; local English OCR for scans and PNG/JPEG images. Find matching text across your current batch.
LOCAL DOCUMENT OPERATIONS
Document Desk
Documents become records you can check.
Extract PDF text or English OCR, match labeled fields to your own template, and review the result against the original. Export your records with page references and source hashes, without sending your files to a document or AI service.
Free: one PDF or image, 5 MB and two pages. No account or upload. Offline edition: Python 3.9+ and a current desktop browser.
01 Keep the evidence with the record
Resolve uncertainty
Exact labels before a colon identify candidates. Conflicting values remain empty until you choose. Required fields and calendar dates are checked, not guessed.
Export usable records
Reviewed CSV, separate unreviewed records, extracted page text, reusable template and source receipts in one ZIP. Originals remain yours to retain.
02 Your rules, not a hidden interpretation
| Source text | What happens |
|---|---|
Reference: 00071 | Leading zeros preserved |
Total: 00120.50 | Decimal text preserved |
Date: 2024-02-30 | Invalid calendar date needs correction |
| Two different values under the same label | No automatic selection; both candidates retained |
OCR can be wrong. Engine confidence is not a factual accuracy score. Confirm fields against the original before moving them into reviewed exports.
03 Offline batches. One purchase.
Document Desk 1.0.0
A complete ZIP with PDF/OCR engines, English language data, application, local launcher, instructions, license and signed release records. No internet required after download.
- 10 files; 10 MB each, 30 MB total
- 25 pages per PDF; 50 pages per batch
- Reusable templates with up to 20 fields
- Batch search, explicit review and source-linked exports
- No AI subscription or per-document processing fee
$79one timeIntroductory pricing
One named user, current version. Python 3.9+ and a current desktop browser are required. Extract the full ZIP and start serve.py; it serves only this application on your own computer.
Checking purchase and private delivery availability.
Tax at checkout. Retrieve your ZIP within 30 days; continue using the downloaded version offline. Review the software license, purchase terms and privacy notice.
04 Choose it for the right work
Labeled records
Invoices, work orders and operational forms with consistent labels. Dates and amounts are checked for shape, not for accounting truth.
Scanned paperwork
Review English OCR text and find phrases in the current batch. Clear originals work best; low-quality scans still need human review.
Privacy-sensitive processing
Source files and results stay in the browser session. The offline launcher accepts no uploads. Retain your exports and originals in your own storage.
Explicit boundaries
PDF, PNG and JPEG only. Images up to 6 million pixels; 100,000 extracted characters per page, 2 million per batch and 20 MB generated output. No layout guessing, multilingual OCR, destination-system writes or future versions included.