Service Admin UI guide

Service Admin is the browser console for a local Service Lasso runtime. Open its configured URL, then use the sidebar to move between operator surfaces. A page may be runtime-backed, metadata-only, preview, or unavailable; do not treat a visual status as durable unless the page states that it is runtime-backed.

First launch and navigation

1. Open the Service Admin URL reported by the Core demo/proof summary.
2. Start with Dashboard. If it cannot load, open Runtime; an unavailable banner means the UI has no usable runtime data.
3. Use the sidebar: Services, Dependencies, Routes, Logs, Runtime, Installed, Variables, Network, Operations, Secrets Broker, Settings, and Help Center.
4. Use page-toolbar and contextual Help links where present. A disabled control states its missing prerequisite or permission.

Core operational workflow

1. In Dashboard, read runtime health, warnings, service totals, and Broker posture.
2. In Services, use search, filtering and sorting, then open the exact service row.
3. On the detail page, review status, dependencies, endpoints, health, setup and recent lifecycle evidence before choosing Start, Stop, Restart, install/configuration or update actions that are exposed.
4. Confirm risky actions in their dialog. Cancellation leaves state unchanged; validation identifies missing inputs or authority.
5. Refresh the detail page and Dashboard, then inspect Logs or Runtime to confirm resulting state. A toast alone is not confirmation.

Page and control reference

Page / route	Purpose and key controls	State handling
Dashboard /	health summary, alerts, recovery links	loading/unavailable banner; verify by Runtime refresh
Services /services, /services/:id	search, filters, sorting, details and lifecycle actions	empty results, disabled actions, confirmation/cancel, error/recovery
Dependencies /dependencies	dependency and SecretRef relationships	metadata may be incomplete; use service detail for action evidence
Service Routes /service-routes, Network /network	configured endpoint/host/port metadata and open links	configuration is not reachability proof
Logs /logs	source selection and bounded log reading	empty/loading/read errors; never paste sensitive logs
Runtime /runtime, Installed /installed	runtime identity, readiness and installed metadata	separates UI from Core failures
Variables /variables	global/service variable and SecretRef posture	raw secret values are not displayed
Operations /inbox, /operations/telemetry, /operations/audit-logging	notices, telemetry and audit metadata	refresh after read/hide/restore mutations
Secrets Broker /secrets-broker/*	setup, inventory, providers, topology, policy and diagnostics	permission, confirmation, lockout and recovery are explicit
MCP /mcp, Security /security	safe MCP/security posture	configuration and mutations are runtime-authorized
Settings /settings/*	appearance, display, account and notification preferences	some data is local/metadata-only
Apps, Chats, Tasks, Users	available workspace surfaces	empty/unavailable states do not imply backend actions
Help Center /help-center	local operator guides and search	no-match state says No docs matched the current search
Sign-in, sign-up, OTP, forgot password and /401–/503	auth and error paths	availability depends on configured identity provider

Secrets Broker safety

Inventory, providers, topology, diagnostics, configuration, audit events, backup keys and operational controls are distinct routes. Reveal, mutation, decommission, rotation, policy and provider workflows require a trusted runtime actor and can require a one-time confirmation. Record only safe identifiers, outcomes and timestamps; never copy a value, recovery material, credential, token, private key or raw request/log payload.

RAM file usage

Secrets Broker → RAM files (/secrets-broker/webdav) opens directly on the inventory panel. It shows listener state, RAM capacity/use, active grants, filenames, service/workspace owners, sizes, completed downloads, bytes served and last access. The same panel appears in the Broker service's Secrets tab. It never offers content reveal, upload or download controls.

Refresh manually or wait for the 30-second poll. Service filters, search and sorting apply to the current page; pagination reaches the remaining inventory. Counters reset when grants are replaced/revoked or Broker restarts. HEAD and directory listings are excluded from completed downloads. Use the secure provisioning lesson to compare these server counters with Echo's safe consumer read status. Workspace-read permission and matching Core/Broker/Admin builds are required; live failures do not substitute demo inventory.

Refreshing the documentation tour

Use the capture manifest and refresh guide to run the reviewed, read-only Playwright tour against a live Service Admin URL. It creates an isolated 1512×982 review set and rejects setup, unavailable, and skeleton screens before it writes an image. A normal run regenerates the approved public assets in this guide's docs/static/img/service-admin-tour/ directory. Dashboard uses its separate public-capture policy: it remains review-only unless the named policy is selected, and its dynamic operational text is then masked before the frame can be promoted.

Current UI tour

The approved, read-only tour frames below are regenerated by the normal capture command. The Dashboard remains omitted until a fresh policy-controlled capture has passed independent visual review; the policy masks Dashboard values rather than treating password masking as sufficient protection.

Troubleshooting and glossary

• Blank, loading or unavailable page: open Runtime, refresh, and verify Core /api/health through the same-origin proxy.
• Action disabled: read the adjacent explanation; common causes are missing permission, unhealthy dependency, unsupported capability or confirmation.
• Healthy but unreachable: compare Network and Service Routes, then test the advertised interface separately.
• No results: clear search/filter controls before diagnosing a service.
• Runtime-backed means returned by Core; metadata-only is descriptive, not live proof; a SecretRef is a reference, not secret material; an operation id identifies a durable action outcome.