This is a port of the Mermaid CLI tool into an Azure Functions v2 (from here on out v2 is assumed) as a docker image.
Mermaid uses Puppeteer, a headless Chromium tool, to render diagrams and export them.
Azure Functions in App Service plans and cosnumption modesl have the following limitations:
- You have to deploy node_modules in your payload
- relatively costly (time-wise) deploy model
- since chromium is native, you have to build where you intend to run (win64, linux64, etc.)
- Windows plans fail to run Chromium, likely due to the isolation system
- Linux does not offer great support for adding extra dependencies needed such as libx11
Azure Functions is a run-time that consists of a .NET Core host and a set of worker run-times that can execute functions. With v2, the Function host run-time is freely available and NOT coupled to Azure at all. Microsoft also ships a base container image for this exact purpose.
Building is really just building the docker image. The image has the necessary versions of node installed, so no other dependencies should be required.
You can host the container anywhere you want, but Kubernetes is considered first class and Azure Container Instances is incredibly easy and performant on 1vCore and 1.5gb ram requirements.
Memory usage may build up over time, I haven't done too much testing at load, but simple use-cases (embedding images in docs by URL) work great.
Puppeteer has to be ran without a sandbox on docker, this means less security, so be cognizant of the trade-offs there and what level of content you send to a public instance.
No input validation is currently done, this is a pretty big security concern since script injection attacks would be trivial. If you plan on making an instance public, definitely conder those ramifications and hopefully add validation first!