💳 Financial Intelligence Platform

An AutoGen-Powered Multi-Agent Bill Management System

Transform bill images into actionable financial intelligence using deterministic analytics, collaborative AI reasoning, and explainable financial recommendations.

---

Python • AutoGen AgentChat • Streamlit • Gemini • Groq • Plotly • Pydantic • EasyOCR

---

🎥 Demo

Demo Video:

---

📸 Application Preview

Landing Page

Financial Dashboard

AI Financial Consultation

---

🌟 Overview

Managing expenses involves far more than simply extracting totals from receipts.

Most expense tracking applications stop after Optical Character Recognition (OCR), leaving users to interpret their own spending behaviour.

This project extends traditional bill processing into a Financial Intelligence Platform that combines deterministic financial analytics with AutoGen-powered collaborative AI agents.

The platform processes uploaded bills, extracts structured financial information, analyzes spending patterns, detects recurring expenses, forecasts future expenditure, and generates verified financial recommendations through an Advisor–Reviewer collaboration workflow.

---

✨ Key Features

• 📄 OCR-based bill processing
• 🧾 Automatic expense categorization
• 📊 Spending analytics dashboard
• 🔁 Recurring expense detection
• 📈 Monthly spending forecasting
• 🤖 AutoGen Multi-Agent collaboration
• 🧠 Advisor–Reviewer financial consultation
• 🪞 Explainable AI with reflection traces
• 📋 Structured execution planning
• 📊 Interactive Streamlit dashboard

---

🏗️ System Architecture

The platform follows a modular architecture consisting of deterministic financial analytics followed by collaborative AI reasoning.

Core Components

• Streamlit User Interface
• Financial Coordinator
• Planning Service
• Execution Service
• Bill Processing Agent
• Expense Analytics Agent
• Recurring Expense Agent
• Spending Forecast Agent
• Financial Team
• Financial Serializer
• Response Synthesizer

Each component has a clearly defined responsibility, making the system modular, maintainable, and easily extensible.

---

🤝 AutoGen Multi-Agent Collaboration

Unlike traditional single-LLM applications, this project employs multiple specialized AI agents that collaborate before producing the final financial recommendation.

Financial Advisor

Responsible for:

• Financial assessment
• Risk identification
• Recommendation generation

Financial Reviewer

Responsible for:

• Independent review
• Detecting missing observations
• Improving recommendation quality
• Providing constructive critique

The reviewer evaluates the advisor's assessment before the advisor produces a verified financial consultation.

This collaborative workflow significantly improves explainability and recommendation quality.

---

🔄 End-to-End Workflow

Upload Bill
      │
      ▼
OCR & Information Extraction
      │
      ▼
Expense Categorization
      │
      ▼
Spending Analytics
      │
      ▼
Recurring Expense Detection
      │
      ▼
Monthly Spending Forecast
      │
      ▼
AutoGen Financial Team
      │
      ▼
Verified Financial Consultation
      │
      ▼
Interactive Dashboard

---

🧠 Engineering Highlights

This project combines deterministic software engineering with modern Agentic AI design.

Highlights

• Planner–Executor architecture
• Modular coordinator design
• Deterministic financial analytics
• AutoGen AgentChat 0.7.5
• Advisor–Reviewer collaboration
• Structured serialization pipeline
• Reflection-based explainability
• Pydantic model validation
• Modular service architecture
• Interactive financial dashboard

---

📊 Financial Intelligence Pipeline

                     User Upload
                          │
                          ▼
                 Bill Processing Agent
                          │
                          ▼
               Expense Analytics Agent
                          │
                          ▼
             Recurring Expense Detection
                          │
                          ▼
              Spending Forecast Agent
                          │
                          ▼
          Advisor ↔ Reviewer (AutoGen)
                          │
                          ▼
             Financial Serializer
                          │
                          ▼
            Response Synthesizer
                          │
                          ▼
               Interactive Dashboard

---

📂 Project Structure

Bill-Management-Agent/
│
├── agents/
│   ├── autogen/
│   ├── coordinator/
│   ├── financial/
│   ├── registry/
│   └── planning/
│
├── models/
│
├── services/
│
├── streamlit/
│
├── tests/
│
├── sample_bills/
│
├── images/
│
└── requirements.txt

---

🛠️ Technology Stack

Layer	Technology
Language	Python 3.12
UI	Streamlit
Multi-Agent Framework	AutoGen AgentChat 0.7.5
LLM Providers	Google Gemini, Groq
OCR	EasyOCR
Data Processing	Pandas
Validation	Pydantic
Visualization	Plotly

---

🚀 Installation

Clone the repository

git clone https://github.com/<username>/Bill-Management-Agent.git

cd Bill-Management-Agent

Create a virtual environment

python -m venv .venv

Activate the environment

source .venv/bin/activate

or

.venv\Scripts\activate

Install dependencies

pip install -r requirements.txt

Configure environment variables

GOOGLE_API_KEY=

GROQ_API_KEY=

Run the application

streamlit run app.py

---

📈 Results

The system successfully provides:

• Automated bill processing
• Expense categorization
• Spending analytics
• Recurring expense analysis
• Monthly expenditure forecasting
• Multi-agent financial consultation
• Explainable AI reasoning
• Interactive financial dashboard

---

🔮 Roadmap

Future enhancements include:

• Persistent financial memory
• Budget planning
• Goal tracking
• Investment portfolio analysis
• Cloud deployment
• Multi-user authentication
• Mobile application
• Long-term financial insights

---

🤝 Contributing

Contributions, feature suggestions, and discussions are welcome.

Feel free to fork the repository and submit a pull request.

---

📄 License

This project is released under the MIT License.

---

👨‍💻 Author

R. Ruthuraraj

Assistant Professor | AI Trainer | Generative AI & Agentic AI Enthusiast

• GitHub: (GitHub link)

---

⭐ If you found this project useful, please consider giving it a star!