Structured Data Extractor

An AI-powered microservice designed to transform unstructured text (emails, invoices, chat logs) into strictly validated, structured data schemas.

Features

• Schema Enforcement: Uses Pydantic V2 to ensure parsed data matches your business rules.
• Smart Parsing: Powered by GPT-4o-mini and instructor for high-accuracy extraction.
• Input Validation: Custom Pydantic validators to catch logical errors (e.g., negative quantities).
• Production Infrastructure: Built-in Docker, Makefile, and CI/CD support.

---

Prerequisites

• Python: 3.10+
• UV: Fast Python package manager
• OpenAI API Key: Required for the parsing engine

---

Usage

1. Configuration

Create a .env file:

OPENAI_API_KEY=sk-...

2. Run API

make dev

Access the interactive documentation at http://localhost:8000/docs.

3. API Scenarios

Scenario: Extracting Order from Email

Request: POST /parse-order

{
  "text": "Hi, I'm John. I want to buy 3 mechanical keyboards and I need express shipping. Please leave it at the front door."
}

Output:

{
  "customer_name": "John",
  "items": [
    {
      "product_name": "mechanical keyboards",
      "quantity": 3
    }
  ],
  "shipping": "express",
  "notes": "Please leave it at the front door."
}

---

Roadmap

• Order extraction logic with Pydantic validation.
• FastAPI modularization.
• Multi-document support (Batch parsing).
• Table extraction from raw text/markdown.

---

Development

• Linting: make lint
• Testing: make test
• Docker: make up