LLM Email Auto Reply System

A Python project I built to automatically handle customer emails using AI. It reads an email, figures out what the customer is asking about, finds the relevant product, and writes back a proper reply — all without any human involvement.

---

What it does

• Reads customer emails from a CSV file
• Classifies each email as a product inquiry, complaint, or other
• Extracts the Product ID if mentioned in the email
• Searches for the product using exact match or semantic search (FAISS)
• Generates a professional reply for each email type using Google Gemini

---

Sample Output

Product Inquiry

Complaint

---

Dataset

The project uses two CSV files.

File	Description	Size
products.csv	Clothing product catalog with name, category, description, price and stock	~25 products
emails.csv	Simulated customer emails covering inquiries and complaints	~30 emails

To run this project you can create your own CSV files with the same column structure or use any similar dataset.

products.csv columns: Product_ID, Product_Name, Category, Description, Price, Stock

emails.csv columns: Email_Text

---

Tools and libraries used

Library	Purpose
Google Gemini 2.0 Flash	Email classification and reply generation
FAISS	Finding the closest product when no ID is mentioned
Sentence Transformers	Converting text into vectors
Pandas	Reading CSV data
python-dotenv	Keeping the API key secure

---

How to run it

1. Clone the repo

git clone https://github.com/Aman132584/llm-email-auto-reply.git
cd llm-email-auto-reply

2. Create and activate a virtual environment

python -m venv venv
venv\Scripts\activate

3. Install dependencies

pip install -r requirements.txt

4. Add your API key

Create a .env file in the root folder and add:

Get a free key at aistudio.google.com

5. Add your data files

Place your CSV files inside a data/ folder:

• data/products.csv
• data/emails.csv

6. Run

python main.py

---

Project Structure

📁 llm-email-auto-reply
│
├── 📄 main.py               # Main pipeline — holds the code
├── 📄 config.py             # Reads environment variables and file paths
├── 📄 requirements.txt      # All required libraries
├── 📄 .gitignore            # Files excluded
│
├── 📁 Outputs            # Sample output images
│   ├── output_1.PNG
│   ├── ouput_2.PNG
│   ├── output_3.PNG
│   └── ouput_4.PNG
│
└── 📁 data                  
    ├── products.csv         # Product catalog 
    └── emails.csv           # Customer emails 

---

Note

An intelligent email automation system that classifies incoming customer emails and generates context-aware replies using LLMs. Designed to reduce manual effort in customer support workflows by automatically handling product inquiries and complaints.