Skip to content
UsecaseBoard
← All use cases
hackernewsbuilder

Generate synthetic receipt JSON with deliberate extraction edge cases

At a glance

A Show HN author shares a model-agnostic generator used to evaluate document extraction, with configurable schemas and locales plus malformed totals, missing fields, and merchant-name typos. This workflow is listed under builder, synthetic-data, document-extraction-evals.

Agents / tools
AI Receipt Generator, LLM backend, JSON schemas

Building a tool for these workflows? See weekly rates.

Save ideas in this browser. Upvotes count across the board.

Prompt excerpt

Use the linked generator in a test-data workspace and configure the JSON schema, locale, model backend, and sample count. Request clearly labeled synthetic cases with missing fields, unusual currencies, and inconsistent totals. Feed them to the extraction evaluation pipeline and check which errors its validation rules catch.
Explore all →