CaseMark stops Mike's matter chat from quietly losing history
A post-merge cleanup turns a handful of silent failure modes in the chat experience into honest, visible ones.
CaseMark spent this round hardening the parts of Mike that lawyers touch day to day. The headline fix: matter chat messages could fail to save without anyone noticing, quietly dropping pieces of a case's conversation history. Now a failed save stops and reports itself instead of vanishing. And when the AI hits a mid-answer error - a quota or budget limit, say - the interface surfaces it as a clear failure rather than freezing mid-stream.
The same pass adds guardrails around the AI features: a cap on how many requests a single user can fire in a given window, which protects against runaway cost and abuse, plus a filter that keeps internal system error messages from leaking back to whoever is using the API.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?