AI Email Generator using Generative AI (Gemini + LangChain + Flask)

This Flask-based web application generates professional, personalized emails using Google's Gemini AI (gemini-2.0-flash-001) through LangChain. It leverages a Retrieval-Augmented Generation (RAG) approach, retrieving relevant email templates from a FAISS vector store for context-aware email generation.

🔗 Live Demo (Note: The application may take up to 50 seconds to load due to hosting constraints.)

Features

• Generates emails based on user input (prompt, recipient, tone, and length).
• Uses RAG with FAISS for context-aware email generation.
• Supports formal and informal tones with customizable email lengths.
• Provides an editable and copyable email subject and body.
• Simple, responsive UI built with Bootstrap.

Tech Stack

• Frontend: HTML, CSS (Bootstrap)
• Backend: Python, Flask
• AI/LLM: Google Gemini (gemini-2.0-flash-001) via LangChain
• Embeddings: models/embedding-001 via GoogleGenerativeAIEmbeddings
• Vector Store: FAISS
• Environment Management: python-dotenv

Installation & Setup

1. Clone the Repository

git clone https://github.com/your-username/ai-mail-generator.git
cd ai-mail-generator

2. Create and Activate a Virtual Environment

Windows

python -m venv venv
venv\Scripts\activate

---

macOS

python3 -m venv venv
source venv/bin/activate

---

Linux (Ubuntu)

python3 -m venv venv
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

4. Set Up Environment Variables

Create a .env file in the project root with the following content:

GOOGLE_API_KEY=your_google_api_key
PORT=10000

Note: Obtain a Google API key with access to Gemini and embedding models from Google AI Studio.

5. Prepare Email Templates

• Create an email_templates.txt file in the project root with sample email templates (optional, as the app includes default templates).
• Example format for email_templates.txt:

  Formal Business Inquiry:
  Subject: Business Inquiry from [Your Name]
  Dear [Recipient Name],
  ...
  
  Informal Job Application:
  Subject: Application for [Position Name] - Let's Connect!
  Hey [Hiring Manager's Name],
  ...
  

File Structure

├── app.py                    # Main Flask application
├── templates/
│   └── index.html            # Web interface template
├── email_templates.txt       # Predefined email templates (optional)
├── .env                      # Environment variables
├── requirements.txt          # Python dependencies
└── README.md                 # Project documentation

Running the App

python app.py

Access the app at: 🔗 http://localhost:10000

Example Usage

1. Open the web interface at http://localhost:10000.
2. Fill in the form:
   • Prompt: Describe the email's purpose (e.g., "Inquire about a product").
   • Recipient: Name of the recipient.
   • Tone: Select "Formal", "Informal", "Harsh" or even "Urgent" .
   • Length: Choose "Short" (50-100 words), "Medium" (100-200 words), or "Long" (200-300 words).
3. Submit the form to generate the email.
4. Edit or copy the generated subject and body directly from the interface.

How It Works

• The app loads email templates from email_templates.txt (or uses defaults if not found).
• Templates are split into chunks using CharacterTextSplitter and indexed in a FAISS vector store with GoogleGenerativeAIEmbeddings.
• Upon form submission:
  • The app retrieves relevant templates using the user's prompt.
  • The Gemini LLM generates a personalized email via a PromptTemplate.
• The output is displayed in a clean Bootstrap interface with options to edit or copy.

Dependencies

• Flask
• python-dotenv
• langchain
• langchain-google-genai
• langchain-community
• faiss-cpu

Install with:

pip install -r requirements.txt

Troubleshooting

• Error: "email_templates.txt not found": Create the file in the project root or rely on default templates.
• Error: API key issues: Verify your Google API key in .env and ensure it has the necessary permissions.
• Error: Port conflict: Update the PORT variable in .env or app.py.

Resources

• LangChain Documentation
• Google Generative AI (Gemini)
• Bootstrap
• FAISS

License

This project is licensed under the MIT License for educational and non-commercial use.