🤖 Customer Support Agentic RAG

An intelligent customer support system leveraging LangGraph and LangChain for Retrieval-Augmented Generation (RAG) with agent-like behavior to deliver accurate, context-aware responses.

---

🚀 Project Overview

This project implements a intelligent RAG-based customer support system that combines the power of LangGraph for workflow orchestration and LangChain for LLM interactions. The system provides intelligent, context-aware responses to customer queries through a multi-stage validation and retrieval pipeline.

Built with FastAPI, FAISS, LangGraph, and Ollama, this system efficiently processes customer support queries while maintaining high accuracy and safety standards through comprehensive validation checks.

---

✨ Key Features

✅ Intelligent Workflow Orchestration – LangGraph-powered pipeline for sophisticated query processing
✅ Advanced Document Retrieval – FAISS vector store for efficient semantic search
✅ Multi-Stage Validation – Comprehensive quality checks at each step
✅ Local LLM Support – Integration with Ollama for on-premise deployment
✅ Content Safety – LLM Guard implementation for safe responses
✅ Efficient Data Processing – Polars-based data preprocessing
✅ API-First Design – FastAPI backend for scalable deployment

---

🏗️ Tech Stack

Category	Tools Used
Programming	Python 3.9+
LLM Integration	LangChain, Ollama, OpenAI API (optional)
Vector Search	FAISS
Workflow Orchestration	LangGraph
Backend Framework	FastAPI
Data Processing	Polars
Safety & Validation	LLM Guard
Deployment	Docker, Docker Compose

---

🔧 System Architecture

The system follows a sophisticated agentic workflow with six main components:

1. Question Validation

   • Input safety checks
   • Token limit verification
   • Toxicity detection

2. Topic Classification

   • Customer support relevance verification
   • Query categorization

3. Document Retrieval

   • FAISS-powered semantic search
   • Context gathering

4. Document Grading

   • Relevance scoring
   • Context validation

5. Answer Generation

   • Context-aware response generation
   • Local or cloud LLM integration

6. Answer Validation

   • Output quality assessment
   • Safety verification

---

📁 Project Structure

├── data/
│   ├── indexes/          # FAISS index storage
│   └── customer_care_emails.csv
├── src/
│   ├── api/             # FastAPI application
│   ├── graph/           # LangGraph workflow components
│   │   ├── answer_check_node.py
│   │   ├── answer_node.py
│   │   ├── docs_grader_node.py
│   │   ├── graph.py
│   │   ├── question_check_node.py
│   │   ├── retriever_node.py
│   │   ├── state.py
│   │   ├── topic_check_node.py
│   │   └── utils.py
│   ├── static/          # Frontend assets
│   └── indexing/        # Data preprocessing and indexing
├── tests/               # Test cases
├── Dockerfile           # Docker file
└── docker-compose.yml   # Docker configuration

---

⚙️ Setup & Installation

1️⃣ Prerequisites

Before you begin, ensure you have:

• Python 3.9 or higher
• Docker (optional)
• Ollama installed (for local LLM support)
• OpenAI API key (optional, for cloud LLM)

2️⃣ Clone the Repository

git clone https://github.com/amine-akrout/customer-support-agentic-rag.git
cd customer-support-agentic-rag

3️⃣ Environment Setup

Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: .\venv\Scripts\activate

Install dependencies:

pip install -r requirements.txt

4️⃣ Configuration

Create a .env file with your settings:

OPENAI_API_KEY=your_api_key_here  # Optional
LANGCHAIN_API_KEY=your_api_key_here # Optional
LANGCHAIN_TRACING_V2= true # Optional
LANGCHAIN_PROJECT=your_project_id_here # Optional

---

🚀 Usage

Local Development

1. Preprocess and Index Data

python -m src.indexing.preprocess

2. Start the API Server

uvicorn src.main:app --reload

3. Access the API at http://localhost:8000

Docker Deployment

1. Build and Start Containers

docker-compose up --build

2. Access the API at http://localhost:8000

---

📡 API Endpoints

Method	Endpoint	Description
POST	/answer	Submit question and get response
GET	/health	Check API health status

Example request:

{
  "question": "How do I return a damaged product?"
}

---

🔄 Workflow Process

The system follows this process for each query:

1. Input Processing

   • Question validation
   • Safety checks
   • Topic classification

2. Context Retrieval

   • Document search
   • Relevance scoring
   • Context selection

3. Response Generation

   • Answer formulation
   • Quality validation
   • Safety verification

---

🤝 Contributing

We welcome contributions! Here's how you can help:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/improvement)
3. Make your changes
4. Commit (git commit -am 'Add new feature')
5. Push (git push origin feature/improvement)
6. Create a Pull Request

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

🙏 Acknowledgments

• LangChain - LLM framework
• LangGraph - Workflow orchestration
• FastAPI - API framework
• FAISS - Vector similarity search
• Ollama - Local LLM support

---

⭐ Star This Repo!

If you find this project useful, please consider giving it a star! 🌟

---

📩 Contact

For questions or feedback, please open an issue in the repository.

---