Show HN: AI-Powered Receipt and Invoice Generator (LLM-Agnostic, Prompt-Based)

GitHub: https://github.com/WellApp-ai/Well/tree/main/ai-receipt-gene... Example output: https://imgur.com/a/YtFSodj

We’ve open-sourced a small tool to generate synthetic receipts and invoices using LLMs — no templates, no HTML, just prompts.

Why? We needed large, diverse, semi-realistic documents to evaluate our document extraction pipeline (OCR + AI). Most generators were either too rigid (static templates) or tied to a single provider. So we built our own:

Features:

LLM-agnostic (OpenAI, local models, etc)

Prompt-driven JSON output (not rendered docs)

Optional Faker-based fallback if the model isn't used

Configurable schemas, locales, and output count

Lightweight and hackable — runs with a single config

This makes it easy to:

Generate test data at scale

Create edge cases (missing fields, odd currencies, broken totals)

Evaluate or fine-tune document understanding models

Example use case: You can ask the model to generate a receipt from a Vietnamese coffee shop, with broken math and a typo in the merchant name — and get structured JSON output that mimics what real OCR might return.

Would love feedback from folks working on:

LLM eval tooling

Synthetic data generation

OCR / document AI

AI agents that touch financial data

Happy to support anyone who wants to extend it or wire up other backends (Claude, Mistral, LM Studio, etc).