MedicalGPT — AI Medical Assistant Telegram Bot

A Telegram bot that provides AI-powered medical consultation using OpenAI's GPT-4. Patients can register their medical history, receive symptom-based follow-up questions, and get filtered medication recommendations — all within Telegram. Built under the guidance of a certified doctor.

---

Features

• Conversational AI — Powered by GPT-4, acting as a personal medical assistant named "Maya"
• Patient registration — Collects age, gender, allergies, current medications, surgeries, medical conditions, and pregnancy status
• Symptom detection — Automatically detects when a message indicates a medical concern and routes to a structured diagnosis flow
• Prescription filtering — Medication recommendations are filtered against each patient's profile (allergies, conditions, medications, demographics)
• Message streaming — Real-time response delivery via Telegram with typing indicators
• Appointment booking — cal.com integration allows booking with a specialist directly from the bot
• Access control — Optionally restrict the bot to a list of allowed Telegram usernames/user IDs
• Conversation history — Dialog stored in MySQL with automatic context pruning when token limits are approached; sessions reset after a configurable timeout

---

Bot Commands

Command	Description
/start	Initialize the bot
/help	Show available commands
/register	Register as a patient (collects medical history)
/diagnose	Start a structured disease diagnosis flow
/choose	Select a concern that best fits your query
/call	Book an appointment with a recommended doctor
/new	Start a fresh conversation
/retry	Regenerate the last bot response
/cancel	Cancel the current conversation or task

---

Tech Stack

• Language: Python 3.8+
• Telegram: python-telegram-bot 20.x
• AI: OpenAI API (gpt-4) via openai 0.27.x
• Database: MySQL via SQLAlchemy ORM
• Async: aiohttp, aiolimiter
• Config: YAML + .env
• Deployment: Docker & Docker Compose
• External: AWS Lambda for cal.com webhook handling

---

Setup

Prerequisites

• Docker and Docker Compose
• An OpenAI API key
• A Telegram bot token from @BotFather
• A running MySQL instance

Steps

1. Clone the repository

git clone https://github.com/gargmegham/MedicalGPT.git
cd MedicalGPT

2. Create configuration files

mv config/config.example.yml config/config.yml
mv config/config.example.env config/config.env

3. Edit config/config.yml

telegram_token: ""              # From @BotFather
openai_api_key: ""              # From OpenAI
use_chatgpt_api: true
allowed_telegram_usernames: []  # Leave empty for public access, or list usernames/user IDs
new_dialog_timeout: 600         # Seconds of inactivity before a new dialog starts
developer_telegram_chatid: ""   # Your Telegram chat ID — receives error notifications
admin_telegram_username: ""     # Admin username

4. Edit config/config.env

MYSQL_USER=
MYSQL_PASSWORD=
MYSQL_HOST=
MYSQL_PORT=
MYSQL_DATABASE=chatgpt

5. Run with Docker Compose

docker-compose --env-file config/config.env up --build -d

---

Running Locally (Development)

pip install -r requirements.txt
python3 bot/bot.py

---

Project Structure

PubMedGPT/
├── bot/
│   ├── bot.py                  # Entry point, bot initialization
│   ├── config.py               # Config loader
│   ├── medicalgpt.py           # GPT interaction and token management
│   ├── mysql.py                # Database operations and prescription logic
│   ├── tables.py               # SQLAlchemy table definitions
│   ├── filters.py              # Custom Telegram message filters
│   ├── utils.py                # Utility helpers
│   └── handlers/
│       ├── commands.py         # Command handlers (/start, /help, etc.)
│       ├── disease.py          # Disease diagnosis conversation flow
│       ├── message.py          # Message streaming handler
│       └── registeration.py   # Patient registration conversation flow
├── config/
│   ├── config.example.yml      # YAML config template
│   ├── config.example.env      # ENV config template
│   └── chat_modes.yml          # System prompts for different chat modes
├── lambda_cal.com/
│   └── lambda_function.py      # AWS Lambda for cal.com booking webhooks
├── training_data/              # Disease-specific JSONL training datasets
├── Dockerfile
├── docker-compose.yml
└── requirements.txt

---

Database Schema (Key Tables)

Table	Purpose
user	Patient profiles — demographics, medical history
dialog	Conversation history stored as JSON
disease / disease_question	Structured diagnosis questions per disease
medicine	Medication database with eligibility filters
disposition	Prescriptions issued per patient/disease
booking	Appointment records from cal.com

---

cal.com Integration

The lambda_cal.com/ directory contains an AWS Lambda function that handles cal.com webhook events (BOOKING_CREATED, BOOKING_CANCELLED). It records bookings in the database and can notify the patient via Telegram.

Booking link template: https://cal.com/<your-username>/telegram-medicalgpt

---

Upcoming Features

• Invoicing and payment gateway integration
• PDF prescription generation with certified professional sign-off after case review

---

Disclaimer

This bot is intended as an informational aid and does not replace professional medical advice, diagnosis, or treatment. Always consult a qualified healthcare provider for medical decisions.