🤖 Customer Service Agent (Multi-Agent System)

A modular multi-agent customer support system that classifies incoming user queries and routes them to the most relevant agent, with an LLM-based fallback dispatcher for handling ambiguous or unseen requests.

System Overview

The system follows a router → agent → response pipeline:

1. User submits a query
2. A classifier/router identifies the intent
3. The query is forwarded to a specialized agent
4. If no clear intent is found, an LLM fallback handles the request

This design allows scalable, maintainable, and intelligent customer support automation.

Agent Architecture

Each agent is responsible for a specific category of queries:

agent_contact.py

• Handles contact-related queries
• Examples: support email, phone number, office hours

agent_db.py

• Handles database or data-backed queries
• Fetches customer-related information from stored data

agent_misc.py

• Handles uncategorized or general queries
• Acts as a soft fallback before LLM escalation

Agents are loosely coupled, making it easy to add or replace functionality.

🔀 Router & Dispatcher

• Classifies user intent based on keywords / logic
• Routes queries to the appropriate agent
• Uses LLM fallback when intent confidence is low

🛠 Tech Stack

• Python
• Multi-Agent Architecture
• LLM (fallback reasoning)
• Streamlit (UI)

How to Run It?

pip install -r requirements.txt
python app.py