Run a receipt-photo extractor with a personal expense dashboard
At a glance
Hema Nambiradje built ReceiptMind to extract purchase details from receipt images and save them in a Neon-backed dashboard. This workflow is listed under Expense tracking, Document extraction, Local app.
- Agents / tools
- ReceiptMind, Gemma 4 26B, OpenRouter, Node.js, Neon Postgres
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Prepare Node.js 18 or later, an OpenRouter account, and a Neon Postgres database. Clone https://github.com/Hema-Nambi/ReceiptMind.git, enter the ReceiptMind directory, and run npm install. 2) Configure the project's environment with your own OPENROUTER_API_KEY and DATABASE_URL. Run node server.js and open localhost:3000; the backend creates its receipts table on startup. 3) Upload a JPG or PNG receipt. Express receives the image, converts it to base64, and sends it with an extraction instruction to the vision-capable Gemma 4 26B model through OpenRouter. The requested fields are merchant, amount, date, category, and tax. 4) The backend parses the returned JSON, saves the record in Neon, and updates the dashboard totals. Compare the extracted amount and tax against the photographed receipt before relying on the expense record.
How it went
The author reports building the app and testing receipts with varied taxes, discounts, and unit prices.
Worth knowing
The author encountered misread discounts and responses wrapped in Markdown fences; parsing required removing those fences.