Skip to main content

Enterprise Lab

Spring Prism ships a repo-only Enterprise Lab that demonstrates the full v1.1.0 story in a clone-and-play sandbox:

  • Spring AI, LangChain4j, and MCP payload shapes
  • two application nodes
  • Redis shared vault restore
  • Grafana with the default Prism dashboard preloaded
  • Big 7 regional rulepack selection
  • NLP heuristic and hybrid modes
  • FAIL_SAFE and FAIL_CLOSED
  • cross-node restore and Redis outage simulation

This lab is for:

  • manual release validation
  • architecture demos
  • contributor debugging
  • realistic smoke tests before cutting a release

It is repo-only and is not part of the published Maven Central surface.

Why it matters​

This lab is more than a sample page. It is a release-grade validation and marketing asset that lets teams see the operational value of Spring Prism without wiring their own infrastructure first.

It demonstrates:

  • distributed tokenization and restoration with two application nodes
  • Redis-backed shared vault behavior under healthy and degraded conditions
  • FAIL_SAFE versus FAIL_CLOSED posture in a realistic UI
  • Big 7 regional rulepack selection in one workspace
  • optional NLP rollout paths for person-name protection
  • operational visibility through the embedded Prism dashboard and Grafana

Architecture at a glance​

The one-command sandbox starts a small enterprise-shaped topology:

  • demo-node-a
  • demo-node-b
  • redis
  • demo-proxy
  • grafana

Traffic enters through demo-proxy, while both app nodes share the same Redis-backed vault and the same Prism app secret. That makes it possible to sanitize on one node and restore on the other.

One-command startup​

On Windows CMD, start the full sandbox from the repository root with:

run-demo.cmd

On Unix-like shells:

./run-demo.sh

The command starts:

  • demo-node-a
  • demo-node-b
  • redis
  • demo-proxy
  • grafana

Docker is the only required local prerequisite. No Maven, Node.js, Redis, or Grafana setup is needed on the host machine.

Default URLs:

  • http://localhost:8080/lab/
  • http://localhost:8080/prism/index.html
  • http://localhost:3000

Grafana is preprovisioned with the Spring Prism Overview dashboard and reads live data from the same /actuator/prism snapshot exposed by the demo stack.

Use these stop commands when you are done:

stop-demo.cmd
./stop-demo.sh

What the UI demonstrates​

The React-based Enterprise Lab is intentionally a single-page command center rather than a fake multi-page product shell.

The UI provides:

  • cluster posture controls for:
    • failure mode
    • NLP mode
    • integration path
    • route strategy
  • Big 7 rulepack selection in the sidebar
  • cluster status cards for nodes and Redis shared vault state
  • direct operations shortcuts for:
    • embedded dashboard
    • Grafana
    • raw Actuator metrics JSON
  • preset payloads next to the raw input editor so testers can switch scenarios quickly
  • a dual workspace for:
    • raw input
    • sanitized outbound content
    • mock model response
    • restored output
  • a trace flow that shows tokenize, vault, masked payload, and restore phases
  • a timeline with node-aware trace events
  • runtime metrics including:
    • protected fields
    • active rules
    • shared vault state
    • request and response blocks
  • an outage toggle that arms the next run for degraded-vault testing

Manual validation checklist​

Use the lab to verify:

  • a mixed prompt is tokenized and restored correctly
  • regional packs such as RO, US, DE, GB, FR, NL, and PL can be toggled explicitly
  • FAIL_CLOSED blocks requests when Redis outage simulation is enabled
  • cross-node restore uses the shared Redis vault
  • Grafana reflects blocked requests and active rules after several runs
  • Grafana reflects protected fields, shared-vault readiness, and history charts after several runs

If you want the highest-signal walkthrough, run this exact sequence:

  1. Start the lab with run-demo.cmd or ./run-demo.sh.
  2. Open http://localhost:8080/lab/.
  3. Run the Mixed Enterprise Payload preset with all default rulepacks enabled.
  4. Switch NLP Mode to HYBRID and run the NLP Person Names preset.
  5. Switch route mode to CROSS_NODE and verify sanitize-on-A / restore-on-B behavior.
  6. Enable Simulate Redis outage, switch to FAIL_CLOSED, and run a protected payload again.
  7. Open Grafana and confirm the dashboard reflects protected fields, active rules, and blocked requests.

Grafana troubleshooting​

If the Spring Prism Overview dashboard shows No data even though http://localhost:8080/actuator/prism returns a populated JSON document, restart Grafana and hard-refresh the browser:

docker compose -f prism-examples\demo-app\compose.yaml restart grafana
docker compose -f prism-examples/demo-app/compose.yaml restart grafana

Then reload:

  • http://localhost:3000/d/spring-prism-overview/spring-prism-overview
  • http://localhost:8080/prism/index.html

Marketing-ready talking points​

When showing Spring Prism to evaluators or internal stakeholders, the lab lets you demonstrate:

  • privacy controls before the LLM boundary, not after
  • distributed restore with a shared Redis vault
  • visible enforcement posture through FAIL_CLOSED
  • international coverage through modular regional rulepacks
  • optional person-name protection without polluting prism-core
  • operational readiness through dashboards and metrics instead of console logs

Notes​

  • The lab does not require a real LLM API key.
  • The frontend is served from /lab/.
  • The embedded Prism dashboard remains available at /prism/index.html.
  • The focused framework examples remain available under prism-examples/ for narrower debugging scenarios.