Bank alerts to a day-book, with a local Gemma

Description and use

A browser‑based day‑book app for a Nigerian POS operator uses Gemma 3‑1B to extract customer names from Moniepoint alerts, running offline on the user’s GPU.

Problem

Mama POS needs to record daily cash inflows from bank transfer alerts without manual note‑taking, but the alerts contain customer names in varying formats that code alone cannot reliably parse.

Outcome

The app correctly extracts 5/5 kinds and 4/5 names from real alerts, and 30/30 kinds and 23/30 names from synthetic alerts. It runs offline after a one‑time 859 MB download and requires a GPU for full functionality.

Workflow

• Open the day‑book app in the browser
• Paste the email alert into the input field
• Code extracts time, amount, and kind of money
• Gemma 3‑1B reads the customer name from the alert text
• App displays a row with all extracted data and a status (Landed/Not seen yet/2 payments)
• User reviews flagged rows (reversals, duplicate amounts)
• At end of day, app calculates totals, estimated charges, and generates a WhatsApp text for the business owner
• Data remains in browser storage
• App can run offline after initial model download

Evidence

She writes each payment in a notebook, and it "takes long".

Source

Gemma found most of them: on the run in the video it found 3 of 3; on an earlier test run it found 2 of 3.

Source

The first load downloads 859 MB once; after that it loads in a few seconds, with no internet (tested with the network off: ready in 6 s).

Source