Repository navigation
Show and tell: testing caller isolation and concurrent chat turns in a Spring AI example #7286
ShubhSingh
started this conversation in
Show and tell
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Thanks to @YellaiahMekala for sharing the Enterprise Spring AI Playbook. I found the project useful and wanted to contribute back. With help from ChatGPT Codex, I submitted four focused application-level improvement PRs.
These changes are in that example application; they are not fixes to the Spring AI framework itself.
What the contributions cover
Verification and limits
Each PR contains regression tests and local verification notes. For the image and concurrency changes, tests exercise real HTTP requests against the application using fake model implementations, avoiding paid provider calls. The concurrency cases use controlled overlapping requests and H2-backed chat memory. The corresponding local suites each reported 43 tests passed and one Docker-dependent MySQL test skipped; these are separate PR runs, not a combined merged-branch result.
The concurrency guard is per JVM. A deployment with multiple replicas needs coordination across instances; this implementation does not claim to solve that. Fake-model tests establish application behavior, not live-provider quality or reliability.
I would welcome feedback on the design trade-offs: for stateful conversations, do you prefer rejecting overlapping turns, queueing per conversation, or versioning memory updates? How do you test these boundaries deterministically?
Thanks again to the owner for making the project available. These PRs were submitted for maintainer review, and I am happy to adapt them to the project’s direction.
All reactions