terpstra-209 fixes a deployment blind spot between Mike's front and back ends
A small container-build change aims to stop production deployments from losing the shared definitions that keep the interface and legal-AI service aligned.
The issue is less glamorous than a new feature, but consequential: the web interface relies on shared information from Mike's back-end service. Building it in isolation meant that connection disappeared during production builds, potentially turning important checks into guesswork and causing the build to fail.
- A fuller build package: the proposed setup includes both sides of Mike when creating the web image, so their shared information remains available.
- A leaner delivery bundle: an exclusion policy keeps bulky dependencies, source-control history and an unrelated add-in out of the build, avoiding unnecessary transfer and build time.
terpstra-209 reports that the rebuilt local stack started cleanly, although the pull request was closed without being merged upstream.
So what Teams self-hosting Mike should care because dependable deployment is a prerequisite for trusting a tool around privileged documents.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?