ZRT Vercel AI Systems Suite
AI Gateway Failover Playground
DEMO PROVIDER MOCKS

Exercise provider routing, fallback, and budgets before production scale exists.

Unified `/api/chat` simulation with routing policies, outage scenarios, latency and cost budgets, circuit breakers, and an auditable request trace.

Plain English

This demo shows how an app can pick a backup AI provider when the first choice is too slow, too expensive, down, or rate-limited.

Selected Route RouteThe provider the gateway chose for this simulated request.

Provider ProviderA model service option the gateway can choose.Backup Balanced
Readiness ReadinessWhether the route meets the current cost and speed settings.READY
Latency LatencyHow long the request is expected to take, measured in milliseconds.540ms
Cost CostThe estimated model cost for this simulated request.$0.007

Primary model is unavailable; fallback pool must preserve reliability.

Routing Controls Routing controlsLevers that change how the simulated gateway chooses a provider.

Provider Pool Provider poolThe available model provider options the gateway can choose from.

Primary Large

iad1 | 820ms | $0.012 | reliability 99%

Fast Small

cle1 | 260ms | $0.003 | reliability 96%

Backup Balanced

sfo1 | 540ms | $0.007 | reliability 98.5%

Circuit Breakers Circuit breakerA safety switch that temporarily avoids a provider when it is down or rate-limited.

primary-large: OPEN

Synthetic outage event

fast-small: OPEN

Synthetic 429 rate-limit event

Budget Status Budget statusShows whether the selected route stays inside the speed and cost limits.

Latency PASS

Cost PASS

No real provider API keys are used.

Provider adapter behavior is deterministic.

Fallback decisions are auditable in the request trace.

Cross-App Build Repair Workflow Cross-app workflowThe shared reviewer path across all four demos, from pasted logs through final report export.

Request Trace Viewer TraceA step-by-step receipt explaining why the gateway chose this route.

1. request_received: PASS

Prompt accepted by unified /api/chat endpoint.

2. policy_eval: PASS

Routing policy: highest_reliability; scenario: primary_outage.

3. provider_health: REVIEW

primary-large unavailable; fallback pool activated.

4. latency_budget: PASS

540ms selected against 900ms budget.

5. cost_budget: PASS

$0.007 selected against $0.020 budget.

6. provider_selected: PASS

Backup Balanced selected.

7. cost_latency_recorded: PASS

$0.007 estimated; 540ms simulated latency.

Return to suite hub