Multi-Provider Failover: Automatic Fallback When a Provider Goes Down

Provider Outages Are Inevitable

OpenAI, Anthropic, Google — every provider experiences downtime. If your app depends on a single provider, an outage means your users see errors. Multi-provider failover eliminates this single point of failure.

How Failover Works

Define a priority chain of equivalent models. If the primary model times out or returns an error, Vincony instantly retries with the next model in the chain — typically in under 200ms.

import Vincony from "vincony";

const client = new Vincony({ apiKey: "YOUR_API_KEY" });

const response = await client.chat.completions.create({
  model: "gpt-4o",
  messages: [{ role: "user", content: "Summarize this report..." }],
  fallback_models: ["claude-sonnet-4", "gemini-2-flash"],
  failover: {
    timeout_ms: 5000,
    retry_on: ["timeout", "rate_limit", "server_error"],
    max_retries: 2
  }
});

console.log(response.model_used);  // Shows which model actually served the request
console.log(response.failover_triggered);  // true if primary failed

Health Monitoring Dashboard

The real-time provider health dashboard shows latency, error rates, and uptime for every provider. Set custom thresholds to trigger failover before users notice degradation.

Cost-Aware Failover

Configure fallback preferences that balance reliability with cost. Prefer cheaper models as fallbacks, or prioritize quality — Vincony respects your constraints while keeping requests flowing.

// Cost-aware failover configuration
const config = await client.failover.configure({
  primary: "gpt-4o",
  fallbacks: [
    { model: "claude-sonnet-4", max_cost_multiplier: 1.2 },
    { model: "gemini-2-flash", max_cost_multiplier: 0.8 }
  ],
  strategy: "cost_optimized"  // or "quality_first"
});

Pricing

Failover routing is included on Pro and Enterprise plans. Free plans get basic failover with up to 2 fallback models.