[Success Story] Streamlit x GHCR x CI/CD

Tl;DR

Not a post about IQVIA nor FMCG, nor D&A tech stack for healthcare analytics.

But one to make the development operations more efficient for a PHD team in a hospital.

Intro

It all comes down to doing + sharing/helping.

Even better if you do so around a well defined offer, so people know what to expect.

SelfHosted Landing Repo

Consulting - Tier of Service

DIY webs via webook

A 101 Ive put together about building websites

In this case, it has been a success story based on github actions x containers best practices

SelfHosted Landing Repo

Tier of Service

What we have done:

1. Define the problem

2. Solution proposal: container image versioning and automatic creation

• Defining the workflow config: https://github.com/jlleongarcia/X-ray-imaging-analysis/commit/2a2eb68132834b39ce8ff3e9e77e8d7d87e801b5

• Making a manual test run towards v1.0: https://github.com/jlleongarcia/X-ray-imaging-analysis/actions/runs/20484822267/job/58865117916 which was before like so

3. Enabling the public ghcr artifacts to public to share within other hospitals: https://github.com/jlleongarcia/X-ray-imaging-analysis/pkgs/container/x-ray-imaging-analysis

Its also recommended to provide a docker-compose for users who want to build their own image:

version: '3.8'

services:
  streamlit-app:
    # Build context points to the current directory (where this file lives)
    build: 
      context: .
      dockerfile: Dockerfile
    container_name: X-ray-imaging-analysis
    restart: unless-stopped
    ports:
    - "8502:8502"
    # Define the healthcheck to ensure the container is truly ready for the tunnel
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8502"]
      interval: 30s
      timeout: 10s
      retries: 3

And one for the ones that just want to pull a particular already built image version with certain features:

docker pull ghcr.io/jlleongarcia/x-ray-imaging-analysis:v1.0

version: '3.8'

services:
  streamlit-app:
    image: ghcr.io/jlleongarcia/x-ray-imaging-analysis:v1.0 #this one is already build, you just pull it (and run it)
    container_name: x-ray-imaging-analysis
    restart: unless-stopped
    ports:
    - "8502:8502"
    # Define the healthcheck to ensure the container is truly ready for the tunnel
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8502"]
      interval: 30s
      timeout: 10s
      retries: 3

sudo docker stats x-ray-imaging-analysis

On my latest projects, I try to differenciate the docker-compose for:

• Development: which builts the image locally
• For Production: which just pull a particular image version with certain features
  • With NGINX/Cloudflare tunnel - which i tend to call portainer version
  • With Traefik - which i tend to call traefik version

Conclusions

We started out with the framework: Whats working, whats not and whats next?

The problem was clearly the way of shipping the code to production and the maintanability/ trazability of the versions.

Thats where bringing the CI/CD strategy to their software methodology with the containers automated built method was a game changer.

Among al CI/CD Tools and processes: we went with a focus on the CI part via github actions.

Fixing their bottleneck and major pain point with their operations and maintainance.

Regarding the what’s next: I recommended to have a look for automated SSL certificates for their internal domain.

Knowing how to control infrastructure (servers/homelab) and providing SSL/https is a must.

Healthcare Analytics is a broad field, and it can be applied to many different areas of healthcare:

mindmap
  root((Medical Physics))
    Radiotherapy
      Linac QA
      Patient Treatment
        3D CRT
        IMRT
        VMAT
    Nuclear Medicine
      Patient Treatment
        Radiopharmaceuticals
      Patient Diagnostic
        Radiotracers
        Machine QA
    Diagnostic Radiology
      Machine QA
      Diagnose
        Clinical Image Validation
    Radiation Protection
      Patient Treatment & Diagnose
        ALARA criteria
      Workers & Public Protection
        Time
        Distance
        Shielding

Powered by the start doing section of last year: Collect case studies / success stories: Flask web app https on a server for SMB / real estate DFY w custom chatbot / hospital D&A CI/CD radiophysics / from zero to astro web hero for a mechanical engineer…and add them to the consulting.jalcocer and/or as article on the main www site.

The success story has been recorded here

Be the next one to get a x10 workflow improvement:

Consulting Services

Consulting - Tier of Service

The power to DIY - via ebooks

Distilled knowledge via web/ooks that enable you to create

FAQ

Latest features shipped properly: https://github.com/JAlcocerT/X-ray-imaging-analysis + youtube video

Health Care Analytics - https://github.com/tuva-health

• https://github.com/tuva-health/tuva

Main repo including core data model, data marts, reference data, terminology, and the clinical concept library