Medical AI Agent Development by Rabiul IslamMedical AI Agent Development by Rabiul Islam

Medical AI Agent Development

Rabiul Islam

Completed work

AI Developer

LangChain

Python

SQLite

Healthcare

🏥 Multi-Tool AI Agent for Medical Datasets and Web

A LangChain-based Medical AI Agent that answers dataset-specific medical questions using three SQLite databases and provides general medical information using web search.

The agent automatically selects the appropriate tool based on the user's question.

🚀 Features

❤️ Heart Disease dataset queries

🎗️ Cancer dataset queries

🩺 Diabetes dataset queries

🌐 General medical information through web search

🤖 Groq LLM with tool calling

🔎 Automatic tool selection

🗄️ SQLite databases

🔐 Read-only SQL validation

🔑 Secure API key management using environment variables

💬 Natural-language answers

🧠 Project Overview

This project is a multi-tool AI agent designed to work with medical datasets.

The agent can answer two major types of questions:

1. Dataset-specific questions

For example:

These questions are answered using the appropriate SQLite database.

2. General medical questions

For example:

These questions are answered using the web search tool.

🏗️ System Architecture

🛠️ Tools

The project contains four main tools.

❤️ HeartDiseaseDBTool

Queries the Heart Disease SQLite database.

Used for:

Record counts

Averages

Percentages

Filtering

Dataset statistics

Numerical questions

Table:

🎗️ CancerDBTool

Queries the Cancer SQLite database.

Used for:

Record counts

Averages

Percentages

Filtering

Dataset statistics

Numerical questions

Table:

🩺 DiabetesDBTool

Queries the Diabetes SQLite database.

Used for:

Record counts

Averages

Percentages

Filtering

Dataset statistics

Numerical questions

Table:

🌐 MedicalWebSearchTool

Searches the web for general medical information.

Used for:

Definitions

Symptoms

Causes

Risk factors

Prevention

Diagnosis

General treatment information

📊 Datasets

1. Heart Disease Dataset

Kaggle dataset:

Records:

SQLite database:

Table:

Columns:

2. Cancer Prediction Dataset

Kaggle dataset:

Records:

SQLite database:

Table:

Columns:

3. Diabetes Dataset

Kaggle dataset:

Records:

SQLite database:

Table:

Columns:

📁 Project Structure

The raw datasets are kept locally inside the data/ directory and are excluded from GitHub using .gitignore.

💻 Technologies Used

Python

LangChain

LangChain Core

LangChain Groq

Groq

SQLite

Pandas

Tavily

python-dotenv

⚙️ Installation

Step 1: Clone the Repository

Go to the project directory:

Step 2: Create Virtual Environment

For Windows:

Activate the virtual environment:

Step 3: Install Dependencies

🔑 API Key Configuration

Create a .env file in the project root directory.

Use .env.example as a reference.

Never upload your .env file to GitHub.

The .gitignore file is configured to prevent .env from being committed.

🗄️ Database Setup

To create the SQLite databases, run:

This creates:

The database setup script also verifies the number of records.

Expected records:

▶️ Running the Agent

Start the application with:

You should see:

Now enter a question.

Example:

The agent will select the Diabetes database tool and return the result.

💡 Example Questions

❤️ Heart Disease

🎗️ Cancer

🩺 Diabetes

🌐 General Medical Questions

🔀 Automatic Tool Routing

The AI agent automatically determines which tool should be used.

For example:

For a general medical question:

🔐 SQL Safety

The database tools are designed to perform read-only SQL queries.

Allowed query types:

The following SQL operations are blocked:

This helps prevent accidental modification of the medical databases.

🧪 Verified Test Results

The agent was tested with several questions.

Diabetes Record Count

Heart Disease

Cancer

Average Glucose

Web Search

🔄 Agent Workflow

🔒 Security

API keys are stored in environment variables.

The following files/directories are excluded from GitHub:

This prevents API keys, virtual environments and raw datasets from being accidentally uploaded.

⚠️ Medical Disclaimer

This project is created for educational and demonstration purposes only.

The AI agent does not provide professional medical diagnosis or treatment.

Information returned by the web search tool may not always be appropriate for an individual's specific medical situation.

Always consult a qualified healthcare professional for medical advice.

👨‍💻 Author

Rabiul Islam

AI/ML Developer specializing in:

Computer Vision

Generative AI

Machine Learning

Intelligent Automation

GitHub:

🔗 Repository

Like this project

Completed work

Posted Sep 18, 2026

Developed an AI agent to answer medical dataset and web queries.

Likes

Views