LedgerGone: Local WhatsApp Khata & Receipt Auditor for Kirana Stores using Multimodal Gemma 3 and TigerData

Описание и применение

LedgerGone позволяет небольшим магазинам в Индии вести учёт поставок и платежей локально, обрабатывая фотографии, рукописные заметки и голосовые сообщения без внешних OCR‑пайплайнов, обеспечивая полную конфиденциальность и автономную работу.

Задача

Small Kirana and hardware shopkeepers in India receive dozens of supplier invoices, handwritten slips, UPI payment screenshots, and WhatsApp voice notes daily, making end‑of‑day tallying chaotic while they resist uploading sensitive data to third‑party cloud accounting SaaS.

Результат

LedgerGone enables Kirana stores to audit receipts and invoices locally with zero external OCR pipelines, achieving cost‑free processing, 100% data sovereignty, and offline resilience while providing real‑time cash‑flow analytics and semantic inventory search.

Рабочий процесс

• 1. Capture supplier GST bills, handwritten parchis, UPI screenshots, and WhatsApp voice notes on the shop counter.
• 2. Feed images and audio into the local Gemma 3 multimodal model via Ollama to extract structured JSON of vendor, items, taxes, payments, and balances.
• 3. Pass extracted data to FastEmbed ONNX to generate 384‑dimensional embeddings for each line item.
• 4. Store audit events in TigerData (TimescaleDB hypertables) and embeddings in pgvector HNSW index.
• 5. Use Streamlit dashboard to view daily sales, collections, and outstanding credit via sub‑millisecond queries.
• 6. Generate 1‑click WhatsApp reminders or downloadable PDF statements for customers or suppliers.
• 7. Review and approve outputs locally, ensuring data sovereignty and offline resilience.

Доказательства

LedgerGone is a 100% local‑first, privacy‑preserving Khata and receipt auditor for Kirana and small retail stores.

Источник

Multimodal Gemma 3: Ingests receipt photos, handwritten parchis, and WhatsApp voice notes directly (with zero external OCR pipelines) to extract structured line items, GST, payments, and pending balances.

Источник

TigerData (TimescaleDB + pgvector): Stores time‑series financial events in partitioned hypertables for cash‑flow analytics and indexes line items using pgvector embeddings for semantic natural language inventory search.

Источник

1‑Click WhatsApp Khata Statements: Automatically generates polite, audited WhatsApp payment reminders with Click‑to‑Chat deep links (wa.me) and downloadable PDF statements.

Источник

Pure CPU Optimization: Runs seamlessly on standard laptops without requiring a dedicated GPU.

Источник