Team Ai
Apppublic

IDKHowToCodeFr/tinyml-backend

sourceHugging Faceupdated 2d agoView on Hugging Face
1likes
App README

<div align="center"> <h1>HeartFlow OS</h1> <p><strong>Cardiovascular Telemetry & MLOps Engine</strong></p>

<p> <img src="https://img.shields.io/badge/Next.js-16+-black.svg?style=for-the-badge&logo=next.js" alt="Next.js" /> <img src="https://img.shields.io/badge/React-19-61DAFB.svg?style=for-the-badge&logo=react&logoColor=black" alt="React" /> <img src="https://img.shields.io/badge/FastAPI-0.111.0+-009688.svg?style=for-the-badge&logo=fastapi" alt="FastAPI" /> <img src="https://img.shields.io/badge/Scikit--Learn-1.5.1+-F7931E.svg?style=for-the-badge&logo=scikit-learn&logoColor=white" alt="Scikit-Learn" /> <img src="https://img.shields.io/badge/License-MIT-green.svg?style=for-the-badge" alt="License" /> </p> </div>

HeartFlow OS takes real-time patient vitals via WebSockets, runs them through an scikit-learn ensemble, and outputs probability distributions for cardiovascular conditions. It also transpiles these models into standalone C headers you can flash directly to constrained edge devices like an ESP32.

Components

DomainTechDetails
FrontendNext.js 16, React 19Renders 60Hz telemetry streams using WebSockets and SVG
BackendFastAPI, Python 3.10Handles WebSocket routing and MLOps API endpoints
MLScikit-LearnSoft-voting ensemble (RF, SVM, KNN, LogReg, MLP)
DatabaseSQLite, aiosqliteNon-blocking telemetry and prediction logging
CompilerCustom AST TranspilerGenerates INT8 quantized, malloc-free C code

Features

  • —Live Telemetry: Streams data at 60Hz via WebSockets to a React 19 dashboard.
  • —Edge Compilation: Converts trained scikit-learn models into C headers. The generated code requires no dependencies and does not use dynamic memory allocation.
  • —Quantization: Scales 64-bit float weights down to 8-bit integers, reducing flash memory usage by ~75%.
  • —Soft-Voting: Combines predictions from 5 different models to output a probability distribution.
  • —Background Retraining: Upload new CSV data to /retrain and the backend will impute missing values, retrain the models in a background thread, and update the live ensemble without dropping active socket connections.

Architecture

See ARCHITECTURE.md for data flow diagrams and the file map.

Quickstart

1. Backend

Requires Python 3.10+.

bash
cd backend
python -m venv .venv
source .venv/bin/activate  # Windows: .\.venv\Scripts\activate
pip install -r requirements.txt

# Train initial models and seed the DB
python models.py

# Start the server
uvicorn main:app --reload --port 8000

2. Frontend

Requires Node.js 18+.

bash
cd frontend
npm install
npm run dev

Open http://localhost:3000 to view the dashboard, upload new data, or export C headers.

License

MIT License. See LICENSE.