The Durga Fleet · Bot 01

Lakshmi

Invoice data extraction — automated. Stop re-typing invoices into your accounting system. Lakshmi reads them for you.

Live · operational since 19 March 2026
Lakshmi
The problem she solves
4.2hrs

Lost every week to manual data entry

The average SME owner or bookkeeper spends over four hours a week manually typing invoice data into accounting software. At £25/hr that's £5,460 a year — for a task a computer should be doing.

23%

Of invoices contain at least one error

Manual re-keying introduces mistakes. Wrong invoice numbers, transposed amounts, missed line items. These errors cause delayed payments, reconciliation headaches, and VAT return corrections.

Every supplier formats differently

Some suppliers send PDFs, some send Word docs, some email a table. Each format requires a different approach. Lakshmi handles all of them — structured or unstructured.

Accounting integrations are expensive

Enterprise OCR and invoice capture tools start at £200/month. Lakshmi delivers the same extraction capability as part of the Durga fleet — no separate subscription required.

How Lakshmi works
1
Send invoice
Forward to Durga via Telegram — PDF, image, or plain text
2
Durga routes
A2A message dispatched to Lakshmi on port 18802
3
Lakshmi extracts
Structured JSON — supplier, date, amount, line items, VAT
4
Data returned
Ready to paste into Xero, QuickBooks, or any accounting system
See it in action
Sample invoice extraction
Powered by Lakshmi
INVOICE #INV-2026-0142
From:Acme Supplies Ltd, Manchester M1 4AB
To:CTECH Ltd, London EC4A 4EN
Date:15 March 2026
Due:14 April 2026 (30 days)
Web hosting services x3£450.00
Domain registration x2£24.00
SSL certificate£89.00
Subtotal£563.00
VAT (20%)£112.60
Total due£675.60
Lakshmi extracts ↓
{ "invoice_number": "INV-2026-0142", "supplier": "Acme Supplies Ltd", "supplier_address": "Manchester M1 4AB", "client": "CTECH Ltd", "date": "2026-03-15", "due_date": "2026-04-14", "payment_terms": "30 days", "line_items": [ {"description": "Web hosting services", "qty": 3, "amount": 450.00}, {"description": "Domain registration", "qty": 2, "amount": 24.00}, {"description": "SSL certificate", "qty": 1, "amount": 89.00} ], "subtotal": 563.00, "vat_rate": 0.20, "vat_amount": 112.60, "total": 675.60, "currency": "GBP" }
Durga ↔ Lakshmi — build conversation
D
Durga · 19 Mar 2026 · 19:59
Lakshmi, your extract_invoice.py script is deployed. Your mission: when you receive an invoice via A2A, extract all structured data and return it as JSON. Supplier name, address, invoice number, date, due date, line items, VAT, total. Confirm you understand.
L
Lakshmi · 19 Mar 2026 · 19:59
Confirmed. I receive invoice text via A2A, run extract_invoice.py, and return structured JSON covering supplier details, invoice metadata, line items with quantities and amounts, VAT breakdown, and total. If fields are missing from the invoice I flag them as null rather than guessing.
D
Durga · 20 Mar 2026 · 09:10
PDF extraction is now available — pdfplumber 0.11.9 installed. You can now extract text from PDF invoices, not just plain text. Update your SOUL.md to note this capability.
L
Lakshmi · 20 Mar 2026 · 09:10
SOUL.md updated. PDF extraction via pdfplumber is now noted as an available capability. I can process both plain text invoices and PDF files. Waiting for next invoice.