Utility Billing Multi-Agent Conversational Chatbot

This project implements a multi-agent conversational AI system to support customer billing inquiries in the utility services domain (such as gas, water, or electricity companies). The goal is to provide accurate, context-aware, and human-like responses by decomposing the chatbot into multiple specialized agents controlled by a central Manager Agent.

---

Objective

To automate customer support for billing queries by enabling the chatbot to:

• Understand user intent
• Retrieve and extract bill details from PDF bills
• Explain bill components in simple language
• Escalate unresolved or complaint-related queries
• Collect customer feedback to improve service

---

System Architecture

The system is composed of six specialized agents, each responsible for a specific part of the conversation:

1. Manager Agent

   • Orchestrates tasks and decides which agent should take action next based on context.

2. Intent Recognition Agent

   • Identifies what the customer is asking (e.g., bill amount inquiry, due date query, complaint).

3. Bill Extraction Agent

   • Parses and extracts structured data from PDF bills (meter usage, bill amount, taxes, due date, etc.).

4. Billing Explanation Agent

   • Converts bill data into simple explanations that customers can easily understand.

5. Escalation Agent

   • Handles scenarios where the user is dissatisfied or when issues require human support.

6. Feedback Agent

   • Collects customer experience feedback at the end of the interaction.

---

Conversation Flow

• User Message → Intent Recognition Agent
• Manager Agent (decision point)
• Bill Extraction Agent (if needed)
• Billing Explanation Agent
• Escalation Agent (if required)
• Feedback Agent (conversation closure)

---

Features

• Natural language understanding of billing questions
• Automated PDF bill reading and structured data extraction
• Friendly, human-readable billing breakdown
• Complaint escalation logic
• End-of-chat feedback collection

---

Tech Stack

• Python
• LLM / Chat Model: OpenAI / HuggingFace
• PDF Parsing: PyMuPDF
• Agent Orchestration: Manager agent routing logic
• UI (optional): Streamlit

---

Installation

git clone https://github.com/<your-username>/utility-billing-multi-agent-chatbot.git
cd utility-billing-multi-agent-chatbot
pip install -r requirements.txt
streamlit run app.py