AI powered Customer Support
This PoC provides a chat interface that uses LlamaIndex for a RAG based chatbot fed with IT Support documents for context.
Target Users
- Customer Service Representatives
- Sales Representatives
- Client Service Representatives
Business Benefits
- Enhanced operational efficiency
- Reduced processing time
- Improved accuracy and consistency
- Better resource allocation
How It Works
The Customer Support Chat is the next generation of our RAG based chatbot implementation. It uses advanced intent detection and context management to provide a more accurate and natural conversation experience. Powered by LlamaIndex, it leverages a vector database to store and retrieve information from a knowledge base of IT Support documents.
Operational Impact
- Reduced time spent searching through documentation
- Consistent messaging about services across the organization
- Improved knowledge sharing and accessibility for all team members
Technical Highlights
Advanced intent detection and context management
LlamaIndex integration
Vector database for efficient information retrieval
AI-powered RAG Customer Support System
Executive Summary
A Retrieval-Augmented Generation (RAG) based customer support system that combines Mistral LLM with ChromaDB for intelligent document retrieval and response generation. The system processes customer queries, retrieves relevant documentation, and generates contextual responses while maintaining source traceability.
System Architecture
Core Components:
1. Query Processing Pipeline
- Generic Query Detection
- Context Classification
- Follow-up Question Handling
- Response Generation with Source Attribution
2. Document Management
- Real-time File Monitoring (Watchdog)
- Automatic Document Processing
- ChromaDB Vector Storage
- Document Embedding Generation
3. LLM Integration
- Mistral Model via Ollama
- Context-aware Response Generation
- Chat History Management
- Reliability Scoring
Key Features
1. Intelligent Query Processing
- Automatic query classification (generic/contextual/non-contextual)
- Follow-up question detection
- Context-aware response generation
- Source documentation tracking
2. Document Management
- Supported formats: PDF, DOC, DOCX, TXT
- Real-time document monitoring
- Automatic embedding generation
- Version tracking and updates
3. Quality Assurance
- Response reliability scoring (high/medium/low)
- Source attribution
- Context relevance verification
- Answer completeness checking
4. System Integration
- FastAPI REST endpoints
- File management API
- Real-time processing status updates
- Secure file handling
Future Enhancements
- Multi-model support
- Enhanced reliability metrics
- Automated knowledge base updates
- Advanced query preprocessing
- Improved context matching
Getting Started Guide: RAG Customer Support System
Prerequisites
Required Software
- Python (v3.9 or higher)
- Ollama (latest version)
- Node.js (v18 or higher) - for frontend development
- Git
System Requirements
- 8GB RAM minimum (16GB recommended)
- 10GB free disk space
- Internet connection for model downloads
Step 1: Environment Setup
Clone the Repository
git clone https://github.com/yourusername/ET-AI-CustomerServiceFaq.git cd ET-AI-CustomerServiceFaq
Create Python Environment
python -m venv venv source venv/bin/activate # On Windows: venv\Scripts\activate pip install -r requirements.txt
Install Ollama
# For Linux/MacOS curl https://ollama.ai/install.sh | sh # For Windows # Download from https://ollama.ai/download
Pull Mistral Model
ollama pull mistral
Step 2: Configuration Setup
Verify Directory Structure
ET-AI-CustomerServiceFaq/ ├── rag-custsupport/ │ ├── main_backend.py │ ├── llm_main.py │ └── services/ ├── docs/ └── data/ └── documents/ # Create this directory
Step 3: Start the Backend
- Navigate to RAG Customer Support Directory:
cd rag-custsupport
- Start FastAPI Server:
python main_backend.py
- Verify Backend Health:
- Open browser to: http://localhost:8001/docs
- Check API documentation is accessible
- Test basic endpoints
Common Issues & Troubleshooting
Ollama Issues
- Ensure Ollama service is running:
ollama serve - Check model availability:
ollama list - Verify no port conflicts on 11434
Document Processing
- Check file permissions in documents directory
- Verify supported file formats
- Monitor logs for processing errors
API Connection
- Default port: 8001
- Check firewall settings
- Verify CORS configuration if needed
Additional Resources
- FastAPI Documentation: http://localhost:8001/docs
- ChromaDB Docs: https://docs.trychroma.com/
- Ollama Guide: https://ollama.ai/docs
- GitHub Issues: Repository Issues
Knowledge Base Management
Upload Documents
Current Documents
No documents uploaded yet.