KenR

Wooohooo 🙌🎉🚀🚀🚀

👏👏👏🎉🎉🎉

EasyAuth for Container Apps is already possible but is a bit awkward to configure using the Provisioning APIs

Woohoo 🙌 this looks amazing 👏 multi repo support will be a game changer, dev tunnels will be nice to have. But being able to define and enforce subnet boundaries, along with environment support for deployment will really change how useful #aspire is outside of local development

I'm now only using #aspire for local development (which is awesome) but it would have been nice if #azd could handle the multiple environments problem

However now that the MiniPAS has now been widely adopted and we want to deploy to various testing environments, staging environments and production, it has quickly outgrown where #aspire and #azd are useful and now I've had to build my own build and deploy pipelines and bicep infra

Building the application was the easy part, but as mentioned in the linked thread, lots of other things took a lot more time...

Initially I deployed the app to #azure using #aspire and AZD. This was great to get going quickly. It took care of spinning up all the infrastructure, container apps etc

the fact that a PAS is a glorified CRUD application is where #llm #coding #agents can really excel

#ai also accelerators how quickly Greenfield development happens. But there are also a few non obvious answers...

#fhir is extremely well thought-out and structured, this was a dream come true for #cursor and #copilot that combined with...

A couple of weeks ago at a company off-site the CEO asked why it was possible to use #ai and #vibe code a PAS in the matter of days but we aren't seeing the same results in the rest of the product... there's the obvious answer about how Greenfield app development is always faster than brownfield...

Don't forget, #boringtech works nicely with #ai

I'm having a lot of fun figuring out how to best to leverage the tools we have available to us in 2025
#ai #copilot #aspire

Not sure where I'm going with this thread now 😅 so will wrap up by saying #ai coding agents have made building certain kind of products incredibly easy. However lots of things are still hard #bicep and #infraascode, deployment pipelines etc

The MirthMock is for local development and integrates the MiniPAS with our product. It's not quite finished but working well enough to demonstrate it's usefulness. Again it's partially written by #ai

Coming back to the challenges our team faces... The MiniPAS made testing easier when our code was deployed and running in a test environment but locally things still sucked... So last week on the train to our company off-site I started to build a MirthMock.

Since launching it internally at a company All Hands it's used by our dev teams, commercial team, transformation team etc to test and demo to clients. It's been a huge success. Maybe one day it will be something we sell.

It took a few sessions with #cursor to realise this was going to work. And proceeded to build a pared down PAS. It handles patients, appointments, clinics and slots. Where #ai was utterly useless, converting #fhir to #hl7. I had to write all that code myself, with some help from our integration devs

I've been at this job for a long time. Over the years various devs have talked about building a PAS. But the effort was too much to fathom. This is where #ai has really been a game changer.
I started the project out of curiosity to see what was possible.

One of the advantages of choosing boring tech (and minimal JavaScript)... #cursor and then #github #copilot was able implement everything I asked it to with very few mistakes that needed to be fixed.

When it's deployed and running in our test environments it is able to integrate with our platform the same way that our clients do (#hl7 messages via Mirth). And then mirth is able to integrate back into it using #FHIR.
This mimics how we actually work with our clients.

To overcome this problem a few months ago I #vibe coded a #fhir based patient administration system (PAS). It uses really boring tech, #dotnet razor pages, very minimal JavaScript and a #fhir server backend.

One of the biggest challenges is integrating with all the various systems. That aside, the other challenge is testing our products. We obviously can't go and buy #epic or #cerner 🤑💰

Working in #healthtech has many challenges, the team I'm currently in is in the scheduling space. Reducing the problem significantly, we're offering slots to patients, and then automating the booking back into the PAS/EPR. #reducingthebacklog

I've opened an issue on GitHub...let me know if you need more info

I do get this error after a little while...but it seems like everything works regardless. I'll open a GitHub issue with more details

Ahh I think the problem was that my two AppHosts were on different versions, 9.2 vs 9.3...i think it's working

Both AppHosts are adding rabbitmq using the same parameters

So I have two solutions open, each with an Aspire AppHost, each with a .AddRabbitMQ. I've specified container name, data volume and persistent lifetime...but I can't run them both simultaneously. The container becomes unhealthy

This is part of an effort to get multiple AppHosts working and running alongside each other. We use RabbitMQ as our messaging broker, so I only want one instance when working locally.

Yes.