Skip to content
← Telemetry UI for Laravel dev-main View on GitHub

Roadmap

Roadmap

Phase 1 — foundation (done)

  • Connector layer: MetricsSource/TracesSource/LogsSource, drivers for Prometheus, Mimir (tenanted), Tempo and Loki, lazy ConnectionManager with extend().
  • Schema autodetection: pages with detectMetric patterns; built-in Statamic page lights up when statamic-telemetry metrics exist.
  • Card/page registry, Livewire base Card, gate + routes, ECharts bundle, period selector. (Replaced by panels + the SPA in v2.)
  • Tooling parity with cboxdk/laravel-telemetry: Pest 4, PHPStan level 8, Pint, arch tests.

Phase 2 + 3 — full information architecture (done)

  • Dashboard: requests (status-class stacked bars), duration (avg/p95), exceptions, jobs.
  • Service/environment switcher in the sidebar, scoping PromQL, TraceQL and LogQL on every card.
  • Requests (route table → trace drill-down), Jobs (outcomes, queue lag, per-job table), Commands, Scheduled Tasks, Exceptions (chart + by-class table + error-trace link), Queries (slowest query spans via TraceQL spanSets), Cache (ops + hit ratio), Outgoing (per-host table), Mail & Notifications, Users (sampled from enduser.* traces), Logs (Loki viewer with search + trace-id links), System (memory/CPU/filesystem/network).
  • Traces: TraceQL search (raw query or quick filters, deep-linked from all drill-downs) + server-rendered waterfall with span attribute expansion.
  • Dev environment: ../laravel-telemetry-demo (LGTM stack in one container with query APIs exposed).

Later polish

  • Detection heuristics beyond a metric-name family (shape, cardinality).
  • Exceptions grouping with first/last-seen (needs app-side state).

Done: short-TTL query cache (cache.ttl, cached GET responses so a busy dashboard doesn't hammer the backends), transient-blip retries, route throttling, auto-refresh polling.

Phase 4 — integrations (done)

  • Issue trackers as a fourth signal (IssuesSource): GitHub (issues + PRs, verified live), Sentry (issue groups) and Linear (GraphQL), config-gated Issues page. Add your own via ConnectionManager::extend().
  • Actions: create a ticket from an exception — a CreatesIssues capability (GitHub, Linear) with a compose form in the drawer prefilled from the exception analysis; the drawer lands on the created ticket.

Phase 5 — correlation, drill-down & agents (done)

The things a generic, app-only or read-only dashboard can't do:

  • Signal correlation — opening a trace surfaces the host/runtime signals recorded around it (CPU, load, memory, network, RSS), each flagged against its typical baseline. Config-driven and fail-open per signal; headless Analysis\SignalContext is the reusable core. See correlation.
  • Purpose-built drill-down — routes/jobs/exceptions/hosts open dedicated detail pages scoped to one entity (throughput → latency → status codes → its traces), not a filtered trace search. Built on hidden pages + a scopeMatchers() card hook + ScopesTo* traits, so overview cards are reused scoped. See custom detail pages.
  • Dimensional filtering — every trace attribute is a click-to-filter link; a Hosts page lists every reporting server.
  • Annotations, written — telemetry-ui:annotate emits deploy/incident/… markers through the telemetry pipeline; telemetry-ui:scan-versions proactively annotates un-announced deploys. See annotations.
  • MCP server — metrics/traces/logs + the correlation tools over the Model Context Protocol (stdio, or HTTP with OAuth 2.1 + DCR via laravel/passport), so an agent can drive incident RCA. See MCP.

Production hardening (done)

  • Backend failures log full detail server-side but show a generic message in the UI (no endpoint/body leak); MCP web fails loud on a half-configured OAuth setup; MCP tools are bounded (row/series/window caps + throttle); correlation baselines are cached; tagValues lookups are time+limit bounded.
  • Feature-tested drill-down/detail pages, dimensional filtering, and the info-leak boundary.

v2 — JSON API + React SPA (done, 2.0.0)

Plan: v2 architecture · implementation plan · ADR 0003.

  • Livewire removed. Cards became framework-free panels returning typed payloads (Panels\Panel, Panels\Ui); the registry, config key and built-in classes renamed accordingly.

  • Versioned JSON API under {path}/api/v2 with typed errors and an SSE live tail; the gate and scope lock run on every endpoint.

  • React + Vite + TypeScript SPA (TanStack Query/Router/Virtual, ECharts as a lazy chunk), committed to public/build so hosts need no Node.

  • Dimensions as a first-class primitive (TelemetryUi::dimension()), Explore over requests/traces/logs/errors with facets and group-by, and generated entity pages that tell a story instead of dumping attributes.

  • Stacked, deep-linkable drawer (?drawer=), ⌘K palette, brush-to-zoom on the global range, live tail over SSE with a polling fallback.

  • Developer integrations: names instead of ids (resolve()), derived dimensions read out of another attribute, route families as pages, and declared metric panels for sidecars in other languages.

  • Embeddable React components for host apps, installed from vendor/ as an npm package — the replacement for the 1.x Livewire widgets.

1.x keeps a maintenance branch (1.x) for fixes. Still open: token auth for an externally hosted SPA.

Next

  • AI triage/chat on a trace, exception or incident (streamed over SSE), building on the MCP tools and SignalContext.
  • Post-to-Slack action alongside ticket creation.
  • Threshold / alert hints.
  • Detection heuristics beyond a metric-name family (shape, cardinality).
← Previous Laravel Telemetry UI Next → Installation