Heart Attack Risk Predictor (HARPA)

Predicts the probability of heart disease from routine clinical + ECG-derived features using a trained ML model.

Live demo

https://harpredictor.streamlit.app/

Local run

pip install -r requirements.txt
python train_model.py          # optional – app auto-trains if model.joblib is missing
streamlit run app.py

Batch scoring

The app accepts CSV uploads in two formats:

• Encoded CSV: same columns as Datasets/Heart Attack/heart_processed.csv (excluding HeartDisease)
• Raw CSV: columns Age, RestingBP, Cholesterol, FastingBS, MaxHR, Oldpeak, Sex, ChestPainType, RestingECG, ExerciseAngina, ST_Slope

Coder Workspace Template

This repo also includes a Terraform template (main.tf) that integrates Coder's product suite.

What it does

• Uses coder_parameter to render an interactive workspace creation form in the Coder dashboard where users enter patient clinical values (age, blood pressure, chest pain type, etc.).
• On workspace provision the agent clones this repo, trains the model, runs a one-shot prediction from the form values, and launches the full Streamlit app inside the workspace.
• A coder_app resource exposes the Streamlit UI directly in the Coder dashboard.

How to use

1. Import main.tf as a template in your Coder instance.
2. Create a workspace — fill in the patient data form.
3. The workspace provisions a Docker container, trains the model, prints the prediction result, and starts Streamlit on port 8501.

Key files

File	Purpose
main.tf	Coder workspace template (Terraform)
predict_from_env.py	Reads patient values from env vars, runs prediction
app.py	Streamlit web app (single + batch scoring)
model_utils.py	Model training, saving, loading utilities
train_model.py	CLI script to train and export the best model

Notes

• model.joblib is gitignored; the app and workspace both auto-train on first run.
• Dataset: Datasets/Heart Attack/heart_processed.csv