Find your next open-source project

Search by name, purpose, or keyword — or paste a GitHub repository URL

Nahomi055/medical-risk-screening-bot

Productivity

A self-hosted Telegram bot that walks users through informal diabetes and heart disease risk screening using XGBoost models.

153
0
TypeScript
MIT
2 days old

What it does

This repository contains a Telegram bot that guides a user through two informal health risk screenings: diabetes and heart disease. Each screening is backed by its own trained XGBoost classifier, and a shared conversation engine asks the model's expected questions one at a time, echoes the answers back for review, then returns a probability estimate plus a plain-language result. It is explicitly a screening aid, not a diagnostic tool, and every result carries a disclaimer pointing to a healthcare professional README.

The design is modular: each screening lives in its own module exposing an ID, display name, feature list, and prediction function, so a new check can be registered without changing the conversation flow. The loader prefers XGBoost's native JSON model format and falls back to pickle files; if a model needs a feature the bot has no question for, it refuses to start rather than producing a bogus prediction.

Quick start

The setup instructions currently describe a PowerShell 5.1+ flow on Windows 10/11 x64, with no administrator rights required and an internet connection to fetch the script setup. Fill in the .env file, at minimum TELEGRAM_BOT_TOKEN, and point the model paths at your trained artifacts; defaults work if the standard layout is kept. Then start the bot and use /start to open the main menu, /screen to choose diabetes, heart disease, or both, or /diabetes and /heart to jump directly into one questionnaire. After a result, /summary replays the last answers, /again reruns the same screening, /cancel aborts an in-progress run, and /reset clears the conversation state.

Usage constraints

The README documents a Windows-only PowerShell installation path and requires a Telegram bot token before the bot will start. Prediction quality depends heavily on the training data, and the heart model's target-label mapping is flagged in the code as a best-effort assumption with # ASSUMED comments. Mismatched scikit-learn and XGBoost versions can cause model-loading problems, so mirroring the training environment is recommended. Local activity logs include user IDs and usernames and rotate at 5 MB with five backups, so the log directory should be treated as sensitive and kept out of version control. The project is MIT licensed license.

HotRepo's take

On the 2026-11-17 new-project list for TypeScript, the repo gained 48 stars over a window of roughly two and a half hours, reaching 153 stars with no forks; there is no recorded all-language daily rank. That is a fast early pickup, but the available series is too short to say anything about sustained interest, and the README's installation and model details are the more useful signal for anyone evaluating whether to run it.

Written by AI from the project's README, updated . The project's own docs are the source of truth.