The Scenario
Large wholesale distributors receive thousands of purchase orders via natural language emails every day. Manually reading these emails, extracting the numbers, calculating discounts, and generating invoices is a massive bottleneck.

Your mission is to build an Agentic AI pipeline that completely automates this process.

The Challenge
Large Language Models (LLMs) are fantastic at reading text, but smaller models are notoriously bad at doing precise arithmetic. If you ask an 8B parameter model to calculate an 18% discount on 47 monitors priced at $215.50, it will likely hallucinate the math.

To solve this, you must build an AI Agent using the ReAct (Reason and Act) framework. You will provide the LLM with three custom-built Python tools. The model must extract the entities from the email and strictly use your tools to calculate the final invoice.

What You Must Build:
You are required to write the underlying Python logic for three specific tools and bind them to your agent:

calculate_subtotal(quantity, unit_price)

apply_discount(subtotal, discount_percent)

calculate_final_total(discounted_amount, shipping_fee)

Your agent will read the email_text from the test dataset, use the tools to perform the math, and output the exact Total_Bill.