A virtual assistant based on RAG (Retrieval Augmented Generation) that processes received emails, generates contextual responses, and automatically replies to VerdeVive customers.

VerdeVive Assistant is an automated customer service solution that uses AI technology to process customer emails and generate personalized responses based on company documentation. The system integrates with the Postmark service to receive and send emails, and uses an advanced language model (Llama3) through the Groq API to generate contextual and relevant responses.

• Backend Webhook: Node.js, Express
• RAG API: Python, Flask
• Language Processing: LangChain, FAISS, HuggingFace Embeddings
• AI Model: Llama3 via Groq API
• Email Processing: Postmark
• Process Management: PM2

email-ai-assistant/
├── backend/                  # Webhook server and RAG API
│   ├── error_emails/         # Stores emails with processing errors
│   ├── rag/                  # Retrieval Augmented Generation API
│   │   ├── indexador.py      # Document indexing script
│   │   ├── main.py           # Flask API for RAG processing
│   │   └── requirements.txt  # Python dependencies
│   ├── package.json          # Node.js dependencies
│   └── server.js             # Webhook server for Postmark
├── docs_empresa/             # Company knowledge base
│   └── content.md            # Company documentation in Markdown format
├── ecosystem.config.js       # PM2 configuration
├── scripts/                  # Utility scripts
│   ├── deploy.sh             # Deployment script
│   └── monitor.sh            # Monitoring script
└── README.md                 # Project documentation

• Node.js (v14 or higher)
• Python 3.8 or higher
• Postmark account with configured server
• Groq API key
• PM2 (for process management in production)

git clone https://github.com/vec21/email-ai-assistant.git
cd email-ai-assistant

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

# Webhook server settings
PORT=3000
[email protected]
POSTMARK_SERVER_TOKEN=your-postmark-token

# RAG API settings
RAG_API_URL=http://localhost:5000
RAG_API_PORT=5000
GROQ_API_KEY=your-groq-api-key
LLM_MODEL=llama3-8b-8192
FLASK_DEBUG=false

cd backend
npm install

cd rag
pip install -r requirements.txt

Before starting the service, you need to index the knowledge base documents:

cd backend/rag
python indexador.py

This command will create a faiss_index directory with the document vectors.

To start the services in development mode:

# Terminal 1 - RAG API
cd backend/rag
python main.py

# Terminal 2 - Webhook server
cd backend
node server.js

For production environment, use PM2:

# In the project root
pm2 start ecosystem.config.js

Verify that the services are working correctly:

# Check webhook status
curl http://localhost:3000/health

# Check RAG API status
curl http://localhost:5000/health

1. Set up a server in Postmark (https://postmarkapp.com)
2. Configure a webhook to send received emails to http://your-server.com/hook
3. Make sure the Postmark server token is configured in the .env file

1. A customer sends an email to the address configured in Postmark
2. Postmark sends a webhook to the Node.js server
3. The server extracts the email content and sends it to the RAG API
4. The RAG API queries the knowledge base and generates a contextual response
5. The server formats the response and sends it back to the customer via Postmark

To monitor the services in production:

pm2 status           # View service status
pm2 logs             # View real-time logs
pm2 monit            # Advanced monitoring

To test the system, send an email to the address configured in Postmark and check if you receive an automatic response.

• RAG API initialization error: Check if the faiss_index directory was created correctly
• Email sending error: Check if the Postmark token is configured correctly
• Inadequate responses: Review and update the content in docs_empresa/content.md
• Connection error: Check if the configured ports (3000 for webhook, 5000 for RAG API) are available

• The system uses SSL for secure communication
• Credentials are stored in environment variables, not in code
• It is recommended to configure a firewall to limit access to service ports

To customize the assistant's responses:

1. Edit the docs_empresa/content.md file with your company information
2. Run the indexer again to update the knowledge base
3. Restart the services to apply the changes

Contributions are welcome! Feel free to open issues or send pull requests.

This project is property of VerdeVive.

Developed by Veríssimo Cassange (CEO of VerdeVive) GitHub: https://github.com/vec21