Resolve conflicts between main's 4.5 changes and the v5 Node work:
- GetLogs: keep main's browser-side log viewer; read cluster logs from
Sentinel, with since_unix_seconds for streaming.
- Terminal: keep main's session-failure events and attach mode; keep
the Node terminal target.
- Dashboard: keep main's layout; add the cluster map as a third servers view.
- Remove the settings sidebar accordion from Node, cluster, and cluster
application sidebars, as main did for servers.
- Replace removed x-forms.select with x-forms.listbox.
- Node factory creates its SSH key in the node team (main scopes key lookup).
- Add v5 models to the server transfer coverage and app key re-encrypt review.
Allow cluster workloads to configure domains and HTTP ports, and enable
individual nodes as ingress proxies. Reconcile Caddy routes and firewall
rules as network revisions so domain changes apply without redeploying.
Refresh ingress routing when workload placements change and merge node
metadata under a row lock to preserve concurrent updates.
Add ingress controls, domain links, and regression and end-to-end coverage.
Add a Lima backend to scripts/dev. Auto mode picks it on macOS when
limactl (Lima >= 2.0) is installed, and COOLIFY_DEV_SERVER_BACKEND=lima
forces it. Lima VMs are named coolify-dev-<slot>-<profile>. Coolify
connects to them through Lima's forwarded SSH port on
host.docker.internal, and only guest ports 80/443 are forwarded to
the host.
dev:qemu:seed now takes --ip and --port options, so the seeded server
can use a forwarded SSH address instead of the profile IP.
Add tests for the Lima start flow and for seeding with an overridden
address, and document the macOS setup in AGENTS.md.
Add UpdateNodeWorkloadConfiguration to create a new workload revision
when the command, ports or environment change. Revisions now store the
environment separately, and a shared NodeWorkload::createRevision helper
builds them. The cluster application page can edit these settings.
Deployments take a pull policy ('missing' or 'newer'), which is passed
through to Sentinel. User redeploys use 'newer' so mutable tags pull the
latest image. Before a move starts, the source node must support
workload.lifecycle.v1 and the target must support workload.deploy.v1.
The coordinated workload.move.v1 command no longer needs its own
capability. Adds a migration for Sentinel capabilities on nodes.
Adds a 'naked' dev QEMU profile runtime. These VMs boot with the
coolify-flux hosts entry and are not seeded, so they can be added
through the Node onboarding UI. Removes the standalone scripts/dev-qemu
script in favour of scripts/dev.
Conflict resolution:
- Development QEMU: keep next's instance mode and prepared template images;
re-add v5 Podman worker Node profiles, Node seeding, Podman templates, and
the node-dev compose overlay through ./scripts/dev.
- Terminal: use next's token-based TerminalSessionService; add Node targets
with the v5 readiness and team checks at issue and redeem time.
- Sentinel: keep v5 Flux imports and next's SentinelSynchronized event; use
next's viewSentinel policy.
- Clusters page: follow next's no-placeholder rule for empty descriptions.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Merge scripts/dev-instances into scripts/dev. Each branch gets its own
Compose project, container, volumes, port block, libvirt network, and
KVM VMs, so stop and start reuse the same data and worktrees do not
collide.
- Name instances after the branch; keep slots and APP_KEYs in the main
checkout's .dev-instances/ so they survive worktree removal
- Reuse existing QEMU VMs instead of recreating them; add dev:qemu --fresh
- Give each instance an isolated libvirt network (10.221.<slot>.0/24)
and VM names coolify-dev-<branch>--<profile>
- Publish per-instance browser ports and publish them with tailscale
serve when APP_URL is a tailnet host
- Add urls, exec, logs, container, destroy, and teardown commands; run
teardown from jean.json before Jean deletes a worktree
- Fix Reverb/terminal browser ports and the testing-host alias in the
instance compose file
- Keep fresh worktree instances working: pre-create laravel.log, log the
host dev:qemu call to stderr, and stop Vite from watching vendor/
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Andras Bacsai