Banking Intelligence Agent

This project implements a Banking Intelligence Agent that leverages LangGraph to process user queries, retrieve relevant information from a mock database, and interact with policies and tools to provide actionable insights.

Workflow Diagram

Below is the workflow diagram used in this project:

Features

• Query Processing: Use LangChain to understand and process user queries.
• Policy Interaction: Retrieves and interacts with banking policies.
• Mock Database Integration: Simulates database interactions for query responses.
• Modular Tools: Includes tools for accounts, transactions, fraud alerts, and more.
• Error Handling: Robust error-handling mechanisms for seamless operation.

Project Structure

.
├── bank.db                   # SQLite database file
├── banking-policies.md       # Banking policies documentation
├── main.py                   # Entry point for the application
├── requirements.txt          # Python dependencies
└── src                       # Source code directory
    ├── config                # Configuration module
    │   ├── __init__.py
    │   └── settings.py       # Configuration settings
    ├── db                    # Database interaction module
    │   ├── __init__.py
    │   └── functions.py      # Functions for database queries
    ├── retrievers            # Data retrieval module
    │   ├── __init__.py
    │   └── vector_store_retriever.py  # Retriever for vector store
    ├── state                 # State management and graph module
    │   ├── __init__.py
    │   ├── assistant.py      # Core assistant logic
    │   ├── error_handling.py # Error handling utilities
    │   └── state_graph.py    # State graph management
    ├── tools                 # Tools for various banking functionalities
    │   ├── __init__.py
    │   ├── accounts.py       # Account-related tools
    │   ├── fraud_alerts.py   # Fraud detection tools
    │   ├── loans.py          # Loan-related tools
    │   ├── payments.py       # Payment-related tools
    │   ├── policy.py         # Banking policy tools
    │   ├── safe.py           # Safety-related tools
    │   ├── sensitive.py      # Sensitive information tools
    │   ├── transactions.py   # Transaction tools
    │   └── user_info.py      # User information tools
    └── utils                 # Utility functions
        ├── __init__.py
        ├── event_printer.py  # Event logging utility
        ├── policy.py         # Policy-related utilities
        └── prompt_templates.py  # Prompt templates for LangChain

Installation

1. Clone the repository:

   git clone https://github.com/nikavgeros/banking-intelligence-agent.git
   cd banking-intelligence-agent

2. Create a virtual environment and activate it:

   python3 -m venv venv
   source venv/bin/activate   # On Windows use `venv\Scripts\activate`

3. Install the required dependencies:

   pip install -r requirements.txt

4. Set up the environment file:

• Create a .env file in the root directory:

  touch .env

• Add your API keys to the .env file:

  OPENAI_API_KEY=your_openai_api_key_here
  TAVILY_API_KEY=your_tavily_api_key_here

5. Run the application:

   python main.py

Usage

After following the installation steps, run the application using the following command:

python main.py

The Banking Intelligence Agent will start, process sample user queries, and provide responses. Below is a sample output:

Example Output

1. Initializing the Database

Initializing the banking database...
Database 'bank.db' and tables created successfully.
Mock data inserted successfully.

2. Query Examples

1. Discovering Accounts and Balances

   Input:
   Hi there, what accounts do I have with your system?

   Output:

   You have the following accounts with our system:
   
   1. **Checking Account**
      - Account Balance: $1,500.00 USD
      - Created At: 2024-11-24 15:16:17
   
   2. **Savings Account**
      - Account Balance: $5,000.00 USD
      - Created At: 2024-11-24 15:16:17
   

2. Sending Money

   Input:
   Please send $50 to user id 2

   Output:
   Tool prompt for action approval:

   Do you approve of the above actions? Type 'y' to continue; otherwise, explain your requested changes.
   
   y
   

3. Recurring Payments

   Input:
   What recurring payments are currently active for me?

   Output:

   You currently have the following active recurring payment:
   
   - **Monthly Rent**
     - Amount: $1,200.00
     - Due Date: 2024-12-01
   

4. Fraud Alerts

   Input:
   I noticed there are some alerts on my account. Can you check if they're resolved?

   Output:

   You have an unresolved fraud alert on your account:
   
   - **Alert ID**: 1
   - **Description**: Suspicious login detected
   - **Alert Date**: 2024-11-24 15:16:17
   - **Resolved**: No
   
   Would you like me to assist you in resolving this alert?
   

   Follow-up:
   Please resolve all active alerts on my account.

   Tool Prompt:

   Do you approve of the above actions? Type 'y' to continue; otherwise, explain your requested changes.
   
   y
   

   Final Check: Have the alerts on my account been resolved?

   Output:

   Yes, the alerts on your account have been resolved. The previously reported alert for a suspicious login detected has been marked as resolved.
   

5. Loans

   Input:
   Can you fetch all the loan details and tell me their current status?

   Output:

   You currently have the following loans and their statuses:
   
   1. **Personal Loan**
      - Loan Amount: $10,000.00
      - Status: Approved
      - Applied At: 2024-11-24 15:16:17
   
   2. **Mortgage Loan**
      - Loan Amount: $250,000.00
      - Status: Pending
      - Applied At: 2024-11-24 15:16:17
   

6. Banking Policies

   Input:
   What is the minimum credit score required to be eligible for a loan?

   Output:

   The minimum credit score required to be eligible for a loan is 650.
   

   Input:
   Is there a fee for closing a checking account?

   Output:

   There is no specific fee mentioned for closing a checking account as long as the balance is $0 and there are no outstanding overdraft fees at the time of closure.
   

7. Web Retrieval

   Input:
   What are the current interest rates for savings accounts offered by major banks?

   Output:

   As of November 22, 2024, the best interest rate for high-yield savings accounts requiring a minimum deposit of $10,000 is 5.35%. Savings account interest rates are often influenced by the federal funds rate, meaning banks typically adjust their rates in line with changes to the Fed's rates. Online banks and credit unions usually offer the best yields compared to traditional banks due to their lower overhead costs and efforts to attract new customers.
   
   For more details, you can visit [Forbes Advisor](https://www.forbes.com/advisor/banking/savings/savings-account-rates-today-11-22-24/) for a comprehensive guide on the best high-yield savings accounts and current rates.
   

Further Improvements

• Vector Database Integration: Replace the in-memory implementation of the VectorStoreRetriever with a vector database such as Pinecone or Milvus. This will enable efficient similarity search, enhanced scalability, and better management of embeddings and metadata. It will also reduce costs by minimizing calls to OpenAI for repeated embedding generations and allow policies to be read and stored once for efficient retrieval.

• Add More Assistants: Introduce additional specialized assistants for areas such as investment planning, regulatory compliance, or customer support. These assistants can work independently or collaboratively to provide holistic insights.

Interactive Approvals

When tool actions are invoked (e.g., send $50 or resolve alerts), the system prompts for user approval. Approve by typing y or provide alternative instructions.

Contributing

Contributions are welcome! Please open an issue or submit a pull request.

License

This project is licensed under the MIT License.