🧠 Multi-Tool Medical AI Agent using LangChain & OpenAI API

Query Medical Databases + Perform Web Search Automatically via AI Routing

Built with FastAPI, LangChain, SQLite, GitHub OpenAI Models, and Tavily/SerpAPI.

---

📌 Overview

This project is a Multi-Tool Medical AI Agent that can:

• Query Heart Disease, Cancer, and Diabetes datasets using SQL
• Perform medical web search (symptoms, treatment, diagnosis, etc.)
• Automatically detect the correct tool based on the question
• Provide answers in natural language
• Expose a clean FastAPI backend for external integration

It is ideal for:

• Medical research assistants
• Data-driven healthcare analysis
• Intelligent chatbot backends
• AI-powered medical dashboards

---

🚀 Features

🔍 Intelligent Tool Routing

You do not need to provide a tool name. The agent automatically selects the correct tool based on your query.

Type of Question	Tool Used
"How many patients...?"	SQL Database Tool
"Average cholesterol?"	SQL Database Tool
"Symptoms of breast cancer?"	Web Search Tool
"Treatment for diabetes?"	Web Search Tool

🗄 SQLite Database Querying (LangChain SQL Agent)

Each dataset is stored as SQLite:

• heart_disease.db
• cancer.db
• diabetes.db

The AI writes SQL automatically and returns human-readable answers.

🌐 Medical Web Search (Tavily / SerpAPI)

Used when the query requires external knowledge.

Example:

“What are the symptoms of breast cancer?”

⚙️ Modular Architecture

File	Description
convert_csv_to_sqlite.py	Converts CSV → SQLite DB
tools.py	Creates SQL database tools
web_search_tool.py	Medical web search (Tavily/SerpAPI)
agent_runner.py	Automatic tool routing logic
app.py	FastAPI backend

---

📁 Project Structure

📁 Multi-Tool-Medical-AI-Agent
│── app.py # FastAPI backend
│── agent_runner.py # AI routing logic
│── tools.py # SQL DB tools using LangChain
│── web_search_tool.py # Tavily / SerpAPI medical search
│── convert_csv_to_sqlite.py # CSV → SQLite converter
│── heart_disease.db
│── cancer.db
│── diabetes.db
│── requirements.txt
└── README.md

---

🔧 Installation

1️⃣ Clone the repository

git clone https://github.com/ZobayerAkib/Multi-Tool-Medical-AI-Agent-using-Langchan-and-OpenAI-Api.git

cd Multi-Tool-Medical-AI-Agent-using-Langchan-and-OpenAI-Api

2️⃣ Create virtual environment

python -m venv venv
venv\Scripts\activate      # Windows
# OR
source venv/bin/activate   # Mac/Linux

3️⃣ Install dependencies

pip install -r requirements.txt

🔑 Environment Variables

• 1. GitHub OpenAI Model Token

  export GITHUB_TOKEN="your_github_token"

• 2. Tavily Search Key (for web search)

  export TAVILY_API_KEY="your_tavily_api_key"

• Optional: SerpAPI alternative

  export SERPAPI_API_KEY="your_serpapi_key"

▶️ Running the Server

Start FastAPI:

uvicorn app:app --reload

Open API docs: 👉 http://127.0.0.1:8000/docs

📡 API Usage

Endpoint

POST /ask
Request
{
  "question": "How many heart disease patients are above age 50?",
  "tool": "auto"
}

Response (example)
{
  "tool_used": "HeartDiseaseDBTool",
  "response": "There are nth(number based on your dataset) heart disease patients above age 50."
}  

🔍 Tools Explanation

🗄 Database Tools (tools.py)

• LangChain SQL Agent
• SQLite
• GitHub OpenAI Models

Tools include:

• HeartDiseaseDBTool
• CancerDBTool
• DiabetesDBTool

🌐 Web Search Tool (web_search_tool.py)

• Supports Tavily API or SerpAPI

🧠 Agent Router (agent_runner.py)

• Automatically chooses the tool based on query keywords:

db_keywords = [
    "count", "total", "number of", "average",
    "patients", "age", "cholesterol", "bp",
    "compare", "below", "above"

🔄 Convert CSV → SQLite Database

Place your CSV files in the repo and run:

python convert_csv_to_sqlite.py

Generates:
heart_disease.db
cancer.db
diabetes.db

🛠 Tech Stack

Technology	Purpose
FastAPI	Backend API
LangChain	SQL Agent + Tool Routing
OpenAI GitHub Models	NLP and reasoning
SQLite	Local medical datasets
Tavily / SerpAPI	Real-time medical search
Python	Main programming language

🤝 Contributing

Pull requests are welcome!
If you have suggestions or questions, feel free to open an issue.

⭐ Support

If you found this project useful, please ⭐ star the repository to support further development.