feat(master): Ledger UI for Services and Agents (#18) (#29)
Docker Build / docker (push) Successful in 41s

## Summary

- Rebuilds the master's **Services** and **Agents** pages per the "Ledger" design: sidebar shell with live nav counts, machine filter and Live indicator; dark/light theme (follows the OS, toggle overrides); embedded Schibsted Grotesk and JetBrains Mono (fontsource 5.3.0, OFL).
- Services: Swarm grouped per cluster with its machines under it, offline banner ("Agent offline since HH:MM", browser-local), Simple Icons derived from the image ref (letter circle fallback), existing badges and mode switch restyled, client-side search, machine filter and Icon / Icon + text / Text toggle.
- Agents: Online / Offline / Never connected / Revoked, master version in the header with outdated agent versions marked, revoke via popover instead of `confirm()`, "Add agent" as a three-step panel that completes when the agent says hello (SSE `agents`).
- Store: migration `0005_agent_cluster.sql` persists each agent's `swarm_cluster_id` (needed to list a cluster's managers and its offline state after a master restart).
- SSE keepalive is now a named `ping` event so the Live indicator also shows the connection is alive. `revokeAgent` now publishes `agents`.

Closes #18

## Notes

- Workers never report a cluster id (Docker only sets `Swarm.Cluster` on managers), so their Swarm role shows "—". Fixing that needs a new hello field.
- Master binary grows by about 2.16 MB, almost all Simple Icons 16.35.0 path data (gzipped JSON, regenerated with `go generate ./internal/master/web/icons`). The agent binary does not import it.
- Jobs, Job and Service detail pages only get the new shell; their own artboards are out of scope, as are Update buttons, Update all, header stats and the drift badge.
- Unchanged: CSRF on every POST, `Cache-Control: no-store` on the token response, 401 + `HX-Redirect` without a session, no `template.HTML` on agent- or user-controlled values, no Node build step.

## Test plan

- [x] `go test ./...`, `go test -race ./internal/master/...`
- [x] `go test -tags integration ./internal/master/store/` against Postgres 17
- [x] `docker build`
- [x] Browser check on fake data: both themes, 1440 px and 390 px, add-agent flow through connect
- [ ] Services grouping and icons against real stacks and a real Swarm cluster
- [ ] "Agent offline since" after stopping an agent
- [ ] Theme toggle, search and machine filter across reloads and mode changes
- [ ] Add agent end to end with a real agent, Copy buttons
- [ ] Phone at 390 px

https://claude.ai/code/session_01GAZsp3P7Qto6JnhipxWUJQ
Reviewed-on: #29
This commit was merged in pull request #29.
This commit is contained in:
2026-10-13 09:20:18 +00:00
parent 919ad91710
commit ec28d1e027
29 changed files with 2140 additions and 199 deletions
+3
View File
@@ -40,6 +40,7 @@ DATABASE_URL=postgres://... go test -tags integration ./internal/master/store/
- `internal/master/poller` — every `POLL_INTERVAL`: one lookup per followed `repo:tag`, sets the available version, starts auto jobs
- `internal/master/jobs` — decides whether a service may be updated and starts a job on the right agent
- `internal/master/webhook` — `POST /update`: records the available version, starts jobs for auto services, waits for them
- `internal/master/web/icons` — embedded Simple Icons path data; slug derived from the image ref; regenerate with `go generate ./internal/master/web/icons`
- `internal/master/web`, `webauth` — HTMX UI behind OIDC login, SSE signals
- `internal/master/events` — in-process publish/subscribe for UI refresh signals
- `internal/master/model` — domain types shared by the above
@@ -162,6 +163,8 @@ DATABASE_URL=postgres://... go test -tags integration ./internal/master/store/
- **Merge lock order is agent, then service rows by id** (as in inventory writes), and the
duplicate row is deleted before the old row is updated, because `location_key` is unique
and not deferrable. A removal racing a merge can be lost; the next snapshot fixes it.
- **The SSE stream sends a named `ping` event every 25 s**; the sidebar Live indicator relies on it.
- **Agents persist `swarm_cluster_id`.** Only managers report it, so workers show no role.
## References