Build a receipt-to-Sheets app with an editable review step
At a glance
A developer used ChatGPT for a build plan and Copilot for a receipt parser that lets users correct data before submission. This workflow is listed under Receipts, Prototyping, Human Review.
- Agents / tools
- ChatGPT, GitHub Copilot, FastAPI, Vision-language model, Google Sheets
SourceBuilding a tool for these workflows? See weekly rates.
What they did
1) Have Python, VS Code with Copilot, ChatGPT, vision-model access and a writable Google Sheet. Enable Google Sheets API in a Cloud project, create a service account and give it access to the sheet. Configure the spreadsheet ID and credential-file path locally. 2) Ask ChatGPT to produce a README plan for a FastAPI app with image upload, extraction, parsing, an editable review table and Sheets submission. Use that README as Copilot's implementation brief. 3) Run the prototype on a receipt image. The author replaced EasyOCR with a vision-language model when ingredient recognition failed. 4) Review and correct the extracted ingredients and prices in the table before submitting. Check that output fields land in the intended spreadsheet columns, then add meal and shelf-life notes manually.
How it went
The author reported an MVP in about an hour. The vision model recognized all ingredients but only about 70% of prices on a crumpled test receipt.
Worth knowing
The first submission had mismatched columns. This is a build workflow, not a packaged app or a complete source-code listing.