Agentic AI chatbot - Equation genie

 Abstract

The field of AI and natural language processing has driven significant growth in intelligent agents for specific problem-solving. This paper details the development of an AI agent for mathematical problem-solving, integrating advanced AI tools and custom software.

The system combines a large language model (LLM), custom tools, and an agent-controller architecture to interact with users in real-time, addressing queries from basic arithmetic to complex functions. This paper presents the system's design, implementation, and testing, evaluating its performance and outlining future improvements.

Existing Systems

Wolfram Alpha: A computational knowledge engine that answers queries by computing answers from curated data. It provides solutions to mathematical, scientific, and general knowledge queries.

2. Microsoft Math Solver: A free app that can solve a variety of math problems, including algebra, calculus, and trigonometry. It uses machine learning algorithms to recognize problems through images and provides step-by-step solutions.

3. Symbolab: An advanced math solver that specializes in algebra, calculus, and other advanced math topics. It offers step-by-step solutions to various problems

Proposed System

The proposed system consists of several key components:

Agent Controller: Acts as the core component, controlling the interaction between the user and the mathematical functions. The agent controller is responsible for managing the input from the user and ensuring the correct tool is invoked based on the user's query.

Mathematical Tools: A set of predefined mathematical tools such as addition,

multiplication, trigonometric functions (sin, cos), logarithmic operations, and

exponentiation are used to perform the calculations.

 Language Model: A large language model (LLM) such as LLaMA is used to understand and generate natural language responses.

User Interface: Gradio provides an interactive interface for the user to input

mathematical queries and receive responses in real time.

Web Interface: The Flask web framework allows the system to be accessed via a web browser, providing a platform-independent interface for users.

Software Requirements

1. Python 3.x: The primary programming language used for the development of the

system.

2. Gradio: A Python library for creating user interfaces that allows users to interact with the agent through a web interface.

3. Flask: A micro web framework used to expose the agent’s capabilities via a REST API.

4. LLaMA Model: A large language model used for natural language understanding and

response generation. Groq Model is used, since it is open source and free. API key is taken

5. Mathematical Libraries: Libraries such as math, numpy, and custom-built tools for

performing mathematical operations.

6. Logging Libraries: Python's logging module for logging events and actions taken by

the system.

The process for building this chat bot, the reference is taken from the following link :

Inspiration Link

Hardware Requirements

1. CPU: A multi-core processor capable of handling the computations required for real-time problem-solving.

   2. RAM: Minimum of 8 GB of RAM to handle the memory-intensive tasks of the language model and mathematical operations.

   3. GPU: A GPU may be required if the LLaMA model is to be run locally, although cloud-based deployment of the model can alleviate this requirement.

   4. Network: Stable internet connection for accessing external services and models.

The project info is hosted as a paper in the following link

Paper Link

Project Advantages

Comprehensive Learning

Practical Implementation

Hands On Example

Production & Deployment

Simple prompting

Chat History

Individual Tools

Tool calls

The moto of the project is: “Harnessing AI agents to transform interactions into intelligent, task-driven processes.”