{
  "name": "primary-llm",
  "modelAlias": "gpt-4o-primary",
  "matchers": [
    {
      "type": "header",
      "key": "x-model-alias",
      "value": "gpt-4o-primary"
    }
  ],
  "backends": [
    {
      "provider": "openai",
      "model": "gpt-4o-2024-08-06",
      "weight": 100,
      "priority": 1,
      "credentialRef": "openai-prod"
    },
    {
      "provider": "azure-openai",
      "model": "gpt-4o",
      "weight": 100,
      "priority": 2,
      "credentialRef": "azure-prod-eastus"
    },
    {
      "provider": "anthropic",
      "model": "claude-sonnet-4-5",
      "weight": 100,
      "priority": 3,
      "credentialRef": "anthropic-prod"
    }
  ],
  "strategy": "fallback",
  "cache": {
    "mode": "semantic",
    "ttlSeconds": 3600,
    "similarityThreshold": 0.95
  },
  "rateLimit": {
    "requestsPerMinute": 600,
    "tokensPerMinute": 200000,
    "concurrency": 32
  },
  "budget": {
    "currency": "USD",
    "limit": 500.0,
    "period": "monthly",
    "action": "block"
  },
  "guardrails": [
    { "name": "pii-redaction", "stage": "both", "policyRef": "pii-redaction-default" },
    { "name": "prompt-injection-guard", "stage": "request", "policyRef": "prompt-injection-block" }
  ],
  "retries": {
    "attempts": 3,
    "perTryTimeout": "30s"
  },
  "observability": {
    "openTelemetry": true,
    "langfuse": true,
    "logRequestBody": false
  },
  "tenant": "platform-eng"
}