🤖 RAG-powered AI Customer Support Agent with Memory

An intelligent AI customer support agent built with Python, Anthropic Claude API, Retrieval-Augmented Generation (RAG), and Conversational Memory.

The agent answers customer questions by retrieving information from a custom knowledge base (product catalog, FAQs, and store policies) while remembering previous conversations to provide more natural and context-aware responses.

---

🚀 Project Overview

This project simulates a real-world AI customer support assistant for an online business.

Instead of relying only on an LLM, the agent:

• Retrieves relevant business information from a custom knowledge base.
• Maintains conversation history during the session.
• Generates accurate, context-aware responses using Claude.
• Reduces hallucinations by grounding responses in business data.

---

✨ Features

• 🤖 AI-powered customer support assistant
• 🧠 Conversational memory for multi-turn interactions
• 📚 Retrieval-Augmented Generation (RAG)
• 🛍️ Product catalog search
• ❓ FAQ retrieval
• 📜 Store policy lookup
• 💬 Context-aware responses using Claude API
• ⚡ Fast and interactive command-line interface
• 🔒 Secure API key management with .env

---

🏗️ System Architecture

                User Question
                      │
                      ▼
        Load Conversation History
                      │
                      ▼
      Search Knowledge Base (RAG)
      ├── Product Catalog
      ├── FAQ
      └── Store Policy
                      │
                      ▼
      Build Context + Conversation
                      │
                      ▼
          Anthropic Claude API
                      │
                      ▼
            AI Generated Response
                      │
                      ▼
      Update Conversation Memory

---

📁 Project Structure

ai-agent-with-memory/
│
├── README.md
├── main.py
├── requirements.txt
├── .env.example
├── .gitignore
│
├── knowledge_base/
│   ├── product_catalog.txt  BD Bangla
│   ├── faq.txt              BD Bangla
│   └── store_policy.txt     BD Bangla
│
└── assets/
    └── demo.png

---

⚙️ Installation & Setup

1. Clone the Repository

git clone https://github.com/saifyea/ai-agent-with-memory.git

Navigate into the project folder:

cd ai-agent-with-memory

---

2. Create Virtual Environment

python -m venv venv

Activate the virtual environment:

Windows:

venv\Scripts\activate

Mac/Linux:

source venv/bin/activate

---

3. Install Dependencies

pip install -r requirements.txt

---

4. Configure Environment Variables

Create a .env file:

ANTHROPIC_API_KEY=your_api_key_here

Replace your_api_key_here with your Anthropic API key.

---

▶️ Run the Application

python main.py

🌐 Language Support

This AI Agent demonstrates RAG capabilities using Bengali business knowledge sources.

The knowledge base contains real-world business information written in Bengali:

• 🇧🇩 Product catalog
• 🇧🇩 Frequently Asked Questions (FAQ)
• 🇧🇩 Store policies

The agent can retrieve Bengali information and generate context-aware responses using an LLM.

---

💬 Example Conversation

The AI Agent can answer customer queries using information retrieved from the Bengali knowledge base while maintaining conversation context.

🛍️ Product Query

User:

Flash Cards এর দাম কত?

AI Agent:

Flash Cards এর দাম: ২৫০ টাকা

বৈশিষ্ট্য:
- বয়স: ৩-৬ বছর
- ৩০টি রঙিন কার্ড প্রতিটি সেটে
- বর্ণমালা, সংখ্যা, ফল, প্রাণী শেখায়
- টেকসই laminated কার্ড

---

🧠 Memory-based Follow-up Question

User:

৩টা কিনলে কত লাগবে?

AI Agent:

৩টা Flash Cards এর মোট দাম: ৭৫০ টাকা
(২৫০ টাকা × ৩ = ৭৫০ টাকা)

The agent understands the previous conversation context and answers based on the previous product discussion.

---

📦 Delivery Information

User:

সিলেটে delivery কতদিনে?

AI Agent:

সিলেটে ডেলিভারি: ৩-৫ দিনের মধ্যে।

আপনার Flash Cards অর্ডার করলে ৩-৫ দিনের মধ্যে পৌঁছে যাবে।

---

🔄 Policy Retrieval

User:

Return policy কী?

AI Agent:

Saif's Kids Store এর Return Policy:

- ৭ দিনের মধ্যে রিটার্ন করা যাবে
- পণ্য অবশ্যই অব্যবহৃত থাকতে হবে
- Original packaging এ থাকতে হবে

---

This demonstrates:

✅ RAG-based knowledge retrieval
✅ Conversational memory
✅ Bengali language support
✅ Business-specific AI assistance

---

🎯 Learning Outcomes

Through this project, I gained practical experience in building AI-powered applications using modern Generative AI techniques.

Key learnings:

• Building AI Agents with LLM APIs
• Implementing Retrieval-Augmented Generation (RAG)
• Managing conversational memory
• Working with custom knowledge bases
• Designing business-focused AI solutions
• Prompt engineering for accurate responses
• Integrating AI into real-world workflows

---

🔮 Future Improvements

Planned improvements for this project:

• 🗄️ Add vector database integration (FAISS / ChromaDB)
• 🧠 Implement long-term persistent memory
• 🔧 Add tool calling capabilities
• 🌐 Build a web interface using Streamlit
• 📊 Add analytics for customer interactions
• 🎙️ Add voice-based interaction
• ☁️ Deploy as a cloud-based AI customer support service

---

👨‍💻 Author

Saifuddin Yeahea

AI Integration Specialist | Python | RAG | AI Agents

I combine 17+ years of enterprise IT experience with modern AI technologies to build practical automation and AI solutions.

🌐 LinkedIn: https://www.linkedin.com/in/saifuddin-yeahea-b557a1109

🚀 Building AI solutions through my #100DaysOfAI journey

📸 Demo