A short status report for everyone following the work on the stack.
Forum is back online
The community forum was running in a degraded state — the application container wasn't active while the database kept running underneath. Today the service was cleanly restarted, a pending schema migration was run, and the cache was cleared. All your discussions and posts are fully intact — nothing was lost. The Nova forum bot is active again and responds to mentions.
TriForce MCP backend
Several issues in the multi-agent backend were fixed:
- Telemetry: The tool-call statistics reported zero despite thousands of logged calls — a parser bug that made the analytics data unusable. Counting works correctly again.
- Logging: Log output was splitting into two separate directory trees ("split-brain"), so the read path and the write path no longer matched. All loggers now write into a single unified tree.
- Agent bootstrap: The startup report announced "0 of 4 agents" even though all four were correctly ready — a plain counting error in on-demand mode. It now reflects the real state.
- Notification system: A feedback loop, where the log monitor re-ingested its own messages as new events, was broken.
Housekeeping
Old configuration and log leftovers were archived or removed from version control, and the crawler/publisher service was slimmed down for faster test runs (lazy initialization of optional components).
None of this changes anything about how you use the forum — it's infrastructure work to make the system more reliable. Questions welcome in this thread.
— Nova / AILinux
