Phase 2 references for the host-agent Dune adapter, moved out of volatile /tmp
into docs/reference-repos/ (per Commander). Three upstream projects, .git +
node_modules + compiled binaries stripped (16MB source). Nested AI-instruction
files (.claude/, CLAUDE.md) removed so they don't pollute Corrosion sessions.
- icehunter/ dune-admin (Go+React) — 4 control planes; SETUP_DOCKER.md is the
closest analog to our agent's Dune docker control plane (compose
lifecycle, docker logs, RabbitMQ-via-exec, dune Postgres schema)
- adainrivers/ Rust/Tauri desktop — SSH+k8s BattleGroup control, maintenance
daemon, in-game admin console (Rust idiom reference)
- the4rchangel/ Node web UI replacing battlegroup.bat — matches the Commander's
Hyper-V self-host path + game-config schema
See docs/reference-repos/README.md for the full index + how we use each.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
5.4 KiB
5.4 KiB
Changelog
1.0.7 — 2026-06-02
SSH — fix false "SSH exited with code 1" on battlegroup commands
- Root cause — Interactive SSH (
-tt) was spawned with stdin closed. OpenSSH exits code 1 with no output in that case, so status/start/stop looked broken even when the VM was fine. lib/ssh.js— Pipe stdin when a pseudo-TTY is requested; treat PTY sessions with stdout as success when stderr is only "Connection closed".
1.0.6 — 2026-06-02
SSH key path — fix battlegroup failure after reboot (WSL)
- Root cause — When the manager runs from WSL,
LOCALAPPDATAis unset. SSH was given a relative path (AppData/Local/DuneAwakeningServer/sshKey) instead of the real key atC:\Users\<you>\AppData\Local\DuneAwakeningServer\sshKey. lib/paths.js— Resolves WindowsLOCALAPPDATAvia PowerShell when env vars are missing; mirrors the key into~/.dune-awakening-server-manager/sshKeywith0600permissions for WSL OpenSSH.- Status / battlegroup —
/api/statusreportsssh.keyPresent; battlegroup routes fail fast with a clear message; dashboard shows an SSH key warning banner.
1.0.5 — 2026-06-02
Start VM — fix silent failure on low host RAM
- Root cause — Dashboard Start VM called Hyper-V with the VM's configured startup RAM (often 30 GB). When the Windows host couldn't allocate that much, Hyper-V returned
OutOfMemory/0x8007000E. The error only appeared in the collapsed console, so it looked like nothing happened. lib/vm.js— SharedstartVm()helper with automatic memory step-down (40→32→30→24→20→18→16→14→12 GB) when the host is low on RAM, plus clearer error messages.POST /api/vm/start— UsesstartVm(); accepts optional{ memoryGB }for manual retry; returns HTTP 507 withstartFailed: trueon OOM.- Dashboard UI — Shows an alert on failure, auto-expands the console, and displays a Retry Start panel with a memory selector when start fails.
- Status — VM card shows configured startup memory when the VM is off.
- Setup import — Also uses auto step-down on first start.
1.0.4 — 2026-06-02
Tech tree — Unlock All Recipes fix
- Root cause — The old endpoint only flipped
UnlockedStateon recipes already present in the character save (~128 entries). The in-game tech tree has 356 nodes (DA_GRP_*groups +DA_REC_*recipes) that were never added to the save, so they stayed locked even after “Unlock All.” public/data/tech-recipe-catalog.json— Full tech node list extracted from game pak files viatools/Cue4ParsePatents(regenerate withdotnet runin that folder).POST /api/characters/:id/tech/unlock-all— Merges every catalog node intom_TechKnowledgeDataasPurchased, preserves existingRCP_*/BLD_*save entries, and setsm_TechKnowledgePointsto 99999.- UI — Tech Tree badge shows
purchased / in save / in game; unlock result reports how many nodes were added. Reminder to stop battlegroup and relog after changes.
1.0.3 — 2026-06-03
Incomplete bootstrap repair
needsBootstrapin/api/status— Detects the “No resources found in funcom-seabass-… namespace” state (VM imported but bootstrap never finished).POST /api/setup/repair— Deletes empty seabass namespace(s) when no battlegroup CR exists, then re-runs bootstrap automatically.- Dashboard repair panel — Yellow banner with token/world/region form when incomplete setup is detected.
1.0.2 — 2026-06-03
Setup — delete and start fresh
- Setup tab → Delete & Start Fresh — When pre-flight detects an existing
dune-awakeningVM, a reset panel appears on step 1. POST /api/setup/reset— Stops the battlegroup (if reachable), removes the Hyper-V VM, deletesDuneAwakeningServerfolders on all drives, removes SSH keys, and clears cached manager state.- Requires typing DELETE to confirm. Re-runs pre-flight automatically after reset so you can walk through the wizard again.
1.0.1 — 2026-06-03
Server finder / WAN visibility fixes
- Reliable visibility IP writes —
settings.confis now written via base64 over SSH instead of fragileprintfescaping. Line 4 is what the gateway reads at startup asGameRmqAddresswhen registering with Funcom. - Stop then start required — UI and console now clearly state that visibility changes require a full battlegroup stop → start (not just restart-in-place). The gateway only publishes the join address on startup.
- WAN port-forward guide — When Public (WAN) or custom public IP is selected in Game Config → Server Visibility, a port-forward checklist appears:
- 31982 TCP — queue/matchmaking (commonly missed; required for server finder)
- Director NodePort TCP — from your Dashboard (e.g. 31402)
- 7777–7810 UDP — game traffic
All forwards must target the VM IP, not the Windows host.
- Setup wizard — Same port-forward notice when choosing a public/custom player IP during initial setup.
- Experimental tab reminder — Self-hosted worlds appear under Servers → Experimental in the game client, not Official or Private.
- API —
GET /api/server-visibilitynow returnsdirectorPort,isWan, and structuredportForwardinfo.POSTreturns restart guidance.
Other
- Improved Server Display Name field hint in Game Config (empty sietch names can hide servers in the browser).