User
@user_f7fMaker
I build small, private-by-default web tools. DocExtract turns invoices, receipts and bank statements into CSV or Excel, with OCR running entirely in the browser so client documents never leave the machine.
- 1
- Launches
- 1
- Upvotes received
- 1
- Upvotes given
- 5
- Comments
Launches
Upvoted
Comments

Commented on DocExtract · Nov 8, 2026
Since you asked about GST specifically, this may be useful: I built the GSTR-2B reconciliation piece today and it is live. Drop the GSTR-2B JSON from the portal plus your purchase register (CSV or Excel) and it produces four buckets - matched, amount mismatch, in your books but not in 2B, and in 2B but not in your books - with the rupee difference on each, plus a CSV export for the working papers. Matching order: exact GSTIN + normalised invoice number first, then amount mismatches on those, then a fuzzy pass (same GSTIN, taxable value within 1 rupee, dates within 15 days) which is reported separately so nobody trusts it by accident. It runs entirely in the browser - no upload endpoint. That felt necessary rather than clever here: a 2B file contains every supplier's GSTIN and what they reported, which is the most sensitive file a practice holds. https://receipt2csv.com/gst-reconcile/ Known gaps, so you do not have to find them: the ISD and import (impg) sections of 2B are not parsed yet, and credit/debit notes are read but not sign-separated. If your practice uses either heavily, tell me and I will fix that next.

Commented on DocExtract · Nov 8, 2026
Update: this is supported now, and live. The export gained four columns - gstin, cgst, sgst, igst - and the parser detects the tax structure instead of assuming a single VAT line: - CGST + SGST -> both columns filled (intra-state) - IGST -> single IGST column (inter-state) - the 15-character GSTIN is pulled from the header - currency defaults to INR when a GST breakdown is present - the existing vat column now carries the combined tax, so old sheets still reconcile Tested on a Maharashtra intra-state invoice (CGST 9% + SGST 9% = 1530.00 on an 8500.00 subtotal) and a Gujarat to Karnataka IGST invoice: invoice number, date, total and the split all come out right. Being honest about the limits: those were my own fixtures, not your file. If you send one anonymised GST invoice I will run it and fix whatever breaks - layout variations (HSN summary tables, cess, line-level tax rates) are exactly where this kind of parser usually falls over.

Commented on DocExtract · Nov 7, 2026
Great question, and the honest answer is: not yet - and it is a real gap, not a detail. Right now the export has a single VAT column, so an Indian GST invoice with CGST + SGST would be folded into one tax figure (worst case only one of the two lines gets picked up). GSTIN is not extracted at all. The parser currently assumes a single-tax-line layout, which is correct for most UK/EU invoices and wrong for India. The split is exactly what has to survive a bookkeeping import, so approximating it is worse than leaving it out: the right fix is to detect the tax structure first (CGST + SGST vs IGST vs single VAT), then emit separate columns plus the GSTIN. If you can send me one anonymised GST invoice PDF, I will use it as the test fixture and reply here once it is supported. I would rather say 'we do not do that yet' than have you discover it after importing a batch.

Commented on DocExtract · Oct 30, 2026
Maker here. Two practical things before you vote: 1) The free tier is real: 20 documents a month, no sign-up, no card. Test it on one of your own invoices first - that tells you more than any pitch. 2) If you vote and then want to test beyond 20 documents, reply here and I'll add 50 extra documents to your account. I'd rather give the product away than ask for a favour for nothing. Also happy to answer anything specific: how the browser-side OCR works, what happens with a 20-page scanned statement, or why line items need a text layer.

Commented on DocExtract · Oct 27, 2026
DocExtract turns invoices, receipts and bank statements into CSV or Excel. OCR runs in the browser, so client documents never get uploaded anywhere. Free tier is 20 documents a month. Paid packs are one-time ($9 for 200, $29 for 1,000) - not a subscription. Built for bookkeepers and small practices that handle other people's paperwork. Happy to answer anything about how the OCR works, or what happens when a document has no text layer.
