WatchTower 1.1.0
WatchTower 1.1.0 — smarter crashes, sharper mods, ops you can trust
This build is about making WatchTower actually understand your server — not just list files and guess. Crash reading, mod triage, and report/live pipelines are more accurate, more complete, and less likely to lie to you. The dashboard polish is there so that intelligence is easy to act on.
Artifacts
watchtower-neoforge-1.1.0+mc1.21.jarwatchtower-cli-1.1.0.jar
Install: replace the mod in mods/ → restart → hard-refresh the dashboard (Ctrl+Shift+R).
Smarter crash intelligence
WatchTower doesn’t stop at “something crashed.” It reads the report the way an experienced operator would:
- Classifies failure kinds — watchdog/hang, FML mod-load, mixin/config/duplicate/worldgen/compat patterns, OOM, env locks, and more — instead of dumping raw stacks
- Names the real suspect — primary / stall mod ids from evidence (stack, FML banners, Mod loading issue blocks), not the loudest unrelated warning
- Scan before the next full report — light classify on crash file heads so the inbox isn’t a wall of “Unknown”
- Titles you can act on — plain-English / exception fallbacks so groups never read blank or useless
- Evidence-first fix plans — “Do this now” steps match the failure (e.g. stop a stuck Create assembly before blaming Flywheel); fewer generic “update the mod” dead ends
- Declarative crash rule packs — optional YAML rules under
config/watchtower/rules/for your pack’s known patterns, without unsafe scripting
Result: fewer mystery crashes, faster triage, advice that matches what actually broke.
Sharper mod analysis
Mod troubleshooting is no longer a flat table of noise:
- Log errors rebuilt — expandable cards with full sample lines, category breakdown, and concrete Do this next from recommendations / scan issues
- No more fake mods — vanilla loggers (
ItemStack, etc.) aren’t truncated into junk ids likeitemstack - Conflicts stay honest — update/compat recommendations only; duplicate “Scan issues” clutter removed (Log errors is the triage surface)
- Create / ecosystem protection — client-only scoring won’t casually recommend removing Create/Flywheel/Registrate stack pieces
- Mod forensics — corrupt jars, config health, class/package find, UCVE owning-jar attribution when enabled
- Optional Modrinth identity — SHA lookup for slugs, links, and outdated hints (never downloads jars for you)
Result: you see which mods are actually hurting the server, with lines of evidence and a next step — not a pile of false positives.
More reliable ops (less lying UI)
- Live charts restored — RX/TX and disk IO from
/api/samplesagain; Network heroes; ambient temps that don’t clip - Startup false-Failed fixed — last
Done!boot window, not the incremental cutoff on a healthy running JVM - Overview uptime live — JVM RuntimeMXBean (not Linux-only
/proc), with report fallbacks - Reports that stick — Collecting stage marks immediately; step/elapsed progress; Hide while running; hydrate on refresh so facts don’t vanish
- Insights / Sources / Issues wiring — data on mount; schedule in hours; Don’t show again hides immediately and persists
- Support bundles include ops-cache — scan aggregates for debugging when something still looks wrong
Result: what you see in Live, Startup, Overview, and Reports matches what the server is actually doing.
Dashboard that showcases the intelligence
| Area | What you’ll notice |
|---|---|
| Sidebar | WatchTower brand mark, glass rail, report controls, theme on rail |
| Overview | A/C/F grade legend, colourful By dimension bars, uptime, backup mute / Not tracking |
| Session | Hero KPIs, Top playtime, online/offline directory |
| Crashes | Classified groups, better titles, bug icon |
| Mods | Real Log errors triage; Conflicts without noise |
| Live | Network section + working IO history |
| Backups | Status/inventory on Backups; setup in Settings → Backups |
Platform
NeoForge 1.21.x · GPL-3.0-or-later
Wiki: https://github.com/djinnbanter/WatchTower/wiki
Full detail: repo CHANGELOG.md