Files
git.stella-ops.org/docs/architecture/console.md
root 68da90a11a
Some checks failed
Docs CI / lint-and-preview (push) Has been cancelled
Restructure solution layout by module
2025-10-28 15:10:40 +02:00

14 KiB
Raw Blame History

StellaOps Console Architecture (Sprint23)

Ownership: Console Guild • Docs Guild
Delivery scope: StellaOps.Web Angular workspace, Console Web Gateway routes (/console/*), Downloads manifest surfacing, SSE fan-out for Scheduler & telemetry.
Related docs: Console overview, Navigation, Runs workspace, Downloads, Console security posture, Console observability, Deployment guide

This dossier describes the end-to-end architecture of the StellaOps Console as delivered in Sprint23. It covers the Angular workspace layout, API/gateway integration points, live-update channels, performance budgets, offline workflows, and observability hooks needed to keep the console deterministic and air-gap friendly.


1·Mission & Boundaries

  • Present an operator-grade UI that surfaces Concelier, Excititor, Policy Engine, Scheduler, Attestor, and SBOM Service data without mutating aggregation or policy state.
  • Enforce Authority-issued scopes and tenant claims on every call through the Console Web Gateway.
  • Deliver deterministic builds (<1MB initial bundle) that can be mirrored in Offline Kits, with runtime configuration loaded from /config.json.
  • Stream live status (ingestion deltas, scheduler progress, telemetry) via SSE with graceful degradation to polling when offline or throttled.
  • Maintain CLI parity by embedding stella commands alongside interactive actions.

Non-goals: authoring ingestion logic, mutating Policy overlays, exposing internal Mongo collections, or performing cryptographic signing in-browser.


2·Workspace & Packages

The console is implemented in src/Web/StellaOps.Web, an Angular17 workspace built on standalone components and Signals.

Path Purpose Highlights
src/app/core/auth DPoP + PKCE authentication, Authority session store, HTTP interceptors. WebCrypto keygen (crypto.subtle), session metadata persisted in sessionStorage, DPoP nonce replay guard.
src/app/core/api Typed API clients for Console gateway (/console/*) and downstream services. DTOs for Scanner, Notify, Concelier exporters; fetch-based clients with abort signals.
src/app/core/config Runtime configuration loader (/config.json), feature flag gating. Supports air-gap overrides and injects API base URLs, Authority issuer/client.
src/app/features/* Route-level shells (auth bootstrap, scans detail, notifications inbox, Trivy DB settings). Each feature is a standalone module with lazy loading and Angular Signals state.
src/app/testing Fixtures and harnesses used in unit tests and storybook-like previews. Deterministic data used for Playwright and Jest scenarios.

Workspace characteristics:

  • Toolchain: Node20.11+, npm10.2+, Angular CLI 17.3. npm run ci:install primes dependencies without network audits; scripts/verify-chromium.js ensures headless Chromium availability for Karma.
  • Build budgets: angular.json enforces 500KB warning / 1MB error for initial bundle and 2KB warning / 4KB error per component stylesheet. Output hashing (outputHashing: all) keeps assets cache-safe.
  • Testing: Karma + Jasmine for unit tests, Playwright for e2e with dev server autotuning. CI (DEVOPS-CONSOLE-23-001) runs Lighthouse against the production bundle.
  • Runtime config: /config.json merged at bootstrap; gateways can rewrite it on the fly to avoid rebuilding for environment changes.

3·Runtime Topology & Data Flow

The console SPA relies on the Console Web Gateway to proxy tenant-scoped API calls to downstream services. Tenant isolation and Aggregation-Only guardrails are enforced at every hop.

graph TD
    subgraph Browser["Browser (Angular SPA)"]
        UI[Console Shell<br/>(Signals, Feature Modules)]
        SSE[EventSource / SSE Clients]
    end
    subgraph Gateway["Console Web Gateway"]
        Router[Minimal API / ASP.NET Core Router]
        StatusCache[Status Cache & Manifest signer]
    end
    Authority[Authority<br/>(DPoP + PKCE)]
    Concelier[Concelier.WebService]
    Excititor[Excititor.WebService]
    Scheduler[Scheduler.WebService]
    Policy[Policy Engine API]
    SBOM[SBOM Service]
    Attestor[Attestor API]
    Downloads[Downloads Manifest Store]

    UI -->|/config.json| Gateway
    UI -->|/console/* (Bearer+DPoP)| Router
    SSE -->|/console/status/stream| Router
    Router --> Authority
    Router --> Concelier
    Router --> Excititor
    Router --> Scheduler
    Router --> Policy
    Router --> SBOM
    Router --> Attestor
    Router --> Downloads
    StatusCache -.-> Gateway
    Gateway -.-> UI

Key interactions:

  • Auth bootstrap: UI retrieves Authority metadata and exchanges an authorization code + PKCE verifier for a DPoP-bound token (aud=console, tenant=<id>). Tokens expire in 120s; refresh tokens rotate, triggering new DPoP proofs.
  • Tenant switch: Picker issues Authority /fresh-auth when required, then refreshes UI caches (ui.tenant.switch log). Gateway injects X-Stella-Tenant headers downstream.
  • Aggregation-only reads: Gateway proxies /console/advisories, /console/vex, /console/findings, etc., without mutating Concelier or Policy data. Provenance badges and merge hashes come directly from upstream responses.
  • Downloads parity: /console/downloads merges DevOps signed manifest and Offline Kit metadata; UI renders digest, signature, and CLI parity command.
  • Offline resilience: Gateway exposes /console/status heartbeat. If unavailable, UI enters offline mode, disables SSE, and surfaces CLI fallbacks.

4·Live Updates & SSE Design

Live surfaces use HTTP/1.1 SSE with heartbeat frames to keep operators informed without polling storms.

Endpoint Payload Source Behaviour
/console/status/stream statusChanged, ingestionDelta, attestorQueue, offlineBanner events Concelier WebService, Excititor WebService, Attestor metrics 5s heartbeat; gateway disables proxy buffering (X-Accel-Buffering: no) and sets Cache-Control: no-store.
/console/runs/{id}/stream stateChanged, segmentProgress, deltaSummary, log Scheduler WebService SSE fan-out Event payloads carry traceId, runId, tenant; UI reconnects with exponential backoff and resumes using Last-Event-ID.
/console/telemetry/stream metricSample, alert, collectorStatus Observability aggregator Gated by ui.telemetry scope; disabled when CONSOLE_TELEMETRY_SSE_ENABLED=false.

Sequence overview:

sequenceDiagram
    autonumber
    participant UI as Console SPA
    participant GW as Console Gateway
    participant SCHED as Scheduler WebService

    UI->>GW: GET /console/runs/42/stream (Authorization + DPoP)
    GW->>SCHED: GET /runs/42/stream (X-Stella-Tenant)
    SCHED-->>GW: event: stateChanged data: {...}
    GW-->>UI: event: stateChanged data: {..., traceId}
    Note over UI,GW: Gateway injects retry-after + heartbeat every 15s
    UI-->>GW: (disconnect)
    UI->>GW: GET /console/runs/42/stream (Last-Event-ID: <seq>)
    GW->>SCHED: GET /runs/42/stream?since=<seq>

Offline behaviour:

  • If SSE fails three times within 60s, UI falls back to polling (/console/status, /console/runs/{id}) every 30s and shows an amber banner.
  • When console.offlineMode=true, SSE endpoints return 204 immediately; UI suppresses auto-reconnect to preserve resources.

5·Performance & Budgets

Surface Target Enforcement
First meaningful paint (dashboard) 2.5s on 4vCPU offline runner Lighthouse CI gate (DEVOPS-CONSOLE-23-001), ui_route_render_seconds P95 alert.
Route hydration (feature shells) 1.5s after token acquisition Angular Signals + lazy loading; route-level budgets tracked via custom telemetry.
Initial bundle size Warn ≥500KB, fail ≥1MB angular.json budgets; CI fails build on overflow.
Component stylesheet Warn ≥2KB, fail ≥4KB angular.json budgets; ensures Tailwind utilities stay tree-shaken.
SSE heartbeat Every 15s max Gateway emits comment heartbeats; UI resets timers on each frame.

Optimisation levers:

  • Standalone components with ChangeDetectionStrategy.OnPush and Angular Signals avoid zone.js churn.
  • fetch + AbortController guard double fetches.
  • Assets served with immutable caching (cache-control: public, max-age=31536000, immutable) thanks to hashed filenames.
  • Compression (gzip/brotli) enabled at gateway; offline bundles include precompressed assets.
  • Command palette, tenants, and filters rely on IndexedDB caches to avoid refetching static metadata.

6·Offline & Configuration Workflows

  • Config manifest: /config.json includes Authority issuer/client ID, gateway base URL, feature flags, telemetry endpoints, and offline hints. Operators can swap config by copying src/config/config.sample.json and editing before build, or by rewriting the response at gateway runtime.
  • Deterministic install: Documented in src/Web/StellaOps.Web/docs/DeterministicInstall.mdnpm run ci:install plus Chromium provisioning ensures offline runners reproduce builds.
  • Offline Kit parity: UI validates downloads manifest signatures (cosign) and surfaces snapshot timestamps per tenant. When offline, buttons switch to CLI snippets (stella runs export, stella downloads sync).
  • Feature flags: CONSOLE_FEATURE_FLAGS toggles modules (runs, downloads, telemetry); offline bundles include flag manifest so UI can render only supported panes.
  • Snapshot awareness: Global banner shows snapshot timestamp and disables actions needing Authority fresh-auth when running in sealed mode.

7·Security & Tenancy

  • DPoP + PKCE: Every request carries Authorization + DPoP header and gateway enforces nonce replay protection. Private keys live in IndexedDB and never leave the browser.
  • Scope enforcement: Gateway checks scope claims before proxying (ui.read, runs.manage, downloads.read, etc.) and propagates denials as Problem+JSON with ERR_* codes.
  • Tenant propagation: X-Stella-Tenant header derived from token; downstream services reject mismatches. Tenant switches log ui.tenant.switch and require fresh-auth for privileged actions.
  • CSP & headers: Default CSP forbids third-party scripts, only allows same-origin connect-src. HSTS, Referrer-Policy no-referrer, and Permissions-Policy configured via gateway (deploy/console.md).
  • Evidence handling: Downloads never cache secrets; UI renders SHA-256 + signature references and steers users to CLI for sensitive exports.
  • See Console security posture for full scope table and threat model alignment.

8·Observability & Telemetry

  • Metrics: Prometheus scrape at /metrics (enabled when CONSOLE_METRICS_ENABLED=true). Key histograms/counters documented in Console observability (ui_route_render_seconds, ui_tenant_switch_total, ui_download_manifest_refresh_seconds).
  • Logs: Structured JSON with traceId, tenant, action. Categories include ui.action, ui.tenant.switch, ui.security.anomaly. Sampled per feature flag to balance volume.
  • Traces: Browser OTLP exporter ships spans to configured collector; gateway adds server-side spans so traces cross client/server boundary.
  • Alerts: Burn-rate rules for route latency, telemetry batch failures, download manifest refresh, and SSE stalls integrate with Notifier.
  • Correlation: SSE events carry traceId so operators can jump from UI to backend logs using shared correlation IDs.

9·Integration Points & Dependencies

Service Console dependency Notes
Authority OIDC, DPoP tokens, tenant catalog, fresh-auth Requires client console-ui with scopes listed in security guide.
Concelier WebService /console/advisories, feed health, export triggers Gateway must enforce Aggregation-Only guardrails and surface merge hashes.
Excititor WebService /console/vex, consensus overlays SSE ticker shows provider deltas.
Policy Engine Findings views, policy previews, simulation diffs Console never writes overlays; uses effective_finding_* data via API.
Scheduler WebService Runs dashboard, SSE streams, queue metrics Heartbeat drives status ticker; cancellation actions require runs.manage.
SBOM Service SBOM explorer tree, component lookup Responses cached per tenant; offline bundles preload snapshots.
Attestor Attestation verification, evidence links Console displays verification status and CLI parity commands.
DevOps downloads pipeline Signed manifest for /console/downloads Manifest signatures validated with cosign key shipped in Offline Kit.

10·Compliance Checklist

  • Frontend package map (core/auth/api/config + feature shells) documented with ownership and tooling details.
  • Data flow diagram captures SPA ↔ Gateway ↔ downstream services with tenant & scope enforcement notes.
  • SSE design documented (endpoints, payloads, heartbeat, retry/backoff, offline fallback).
  • Performance budgets (<1MB initial bundle, route hydration ≤1.5s, SSE heartbeat) stated alongside enforcement mechanisms.
  • Offline workflows (/config.json, deterministic install, Offline Kit parity) described with operator guidance.
  • Security section references DPoP, scopes, CSP, evidence handling, and tenancy propagation.
  • Observability metrics/logs/traces coverage listed with alert hooks.
  • Integration dependencies table links Console responsibilities to upstream services.
  • Document cross-references validated (UI guides, security, observability, deployment).
  • Last updated timestamp refreshed after review.

Last updated: 2025-10-27 (Sprint23).