{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "$id": "https://raw.githubusercontent.com/api-evangelist/ai-gateway/refs/heads/main/json-schema/ai-gateway-route-schema.json",
  "title": "AIGatewayRoute",
  "description": "A single model route defined on an AI gateway. A route binds an inbound matcher (model alias, header, identity, path) to one or more upstream LLM provider backends, with optional caching, fallback, fanout, rate limiting, budget, and guardrail policy attached.",
  "type": "object",
  "properties": {
    "name": {
      "type": "string",
      "description": "Unique route name within the gateway.",
      "example": "primary-llm"
    },
    "modelAlias": {
      "type": "string",
      "description": "Client-facing model alias the gateway exposes; resolved to one or more upstream models.",
      "example": "gpt-4o-primary"
    },
    "matchers": {
      "type": "array",
      "description": "Inbound matchers — header, model name, path, identity, or content-based conditions that select this route.",
      "items": {
        "type": "object",
        "properties": {
          "type": {
            "type": "string",
            "enum": ["header", "model", "path", "identity", "time", "content"]
          },
          "key": { "type": "string" },
          "value": { "type": "string" }
        },
        "required": ["type"]
      }
    },
    "backends": {
      "type": "array",
      "description": "One or more upstream provider backends serving this route. Multiple backends enable fallback or fanout depending on strategy.",
      "items": {
        "type": "object",
        "properties": {
          "provider": {
            "type": "string",
            "description": "The LLM provider identifier (slug) — see provider schema.",
            "example": "openai"
          },
          "model": {
            "type": "string",
            "description": "Provider-side model identifier.",
            "example": "gpt-4o-2024-08-06"
          },
          "weight": {
            "type": "integer",
            "description": "Load-balancing weight; higher means more traffic.",
            "example": 100
          },
          "priority": {
            "type": "integer",
            "description": "Fallback priority; lower wins.",
            "example": 1
          },
          "credentialRef": {
            "type": "string",
            "description": "Reference to a stored upstream API key or secret.",
            "example": "openai-prod"
          }
        },
        "required": ["provider", "model"]
      },
      "minItems": 1
    },
    "strategy": {
      "type": "string",
      "description": "How the gateway selects among multiple backends.",
      "enum": ["fallback", "load-balance", "fanout", "shadow", "cost-optimized", "latency-optimized"],
      "example": "fallback"
    },
    "cache": {
      "type": "object",
      "description": "Response caching configuration for this route.",
      "properties": {
        "mode": {
          "type": "string",
          "enum": ["off", "exact", "semantic"],
          "example": "semantic"
        },
        "ttlSeconds": { "type": "integer", "example": 3600 },
        "similarityThreshold": {
          "type": "number",
          "description": "Cosine similarity threshold for semantic-cache hits.",
          "example": 0.95
        }
      }
    },
    "rateLimit": {
      "type": "object",
      "description": "Per-route rate limits enforced at the gateway.",
      "properties": {
        "requestsPerMinute": { "type": "integer", "example": 600 },
        "tokensPerMinute": { "type": "integer", "example": 200000 },
        "concurrency": { "type": "integer", "example": 32 }
      }
    },
    "budget": {
      "type": "object",
      "description": "Spend caps applied to this route.",
      "properties": {
        "currency": { "type": "string", "example": "USD" },
        "limit": { "type": "number", "example": 500.0 },
        "period": {
          "type": "string",
          "enum": ["daily", "weekly", "monthly"],
          "example": "monthly"
        },
        "action": {
          "type": "string",
          "enum": ["alert", "throttle", "block"],
          "example": "block"
        }
      }
    },
    "guardrails": {
      "type": "array",
      "description": "Guardrail policies applied to inbound prompts and outbound completions on this route.",
      "items": {
        "type": "object",
        "properties": {
          "name": { "type": "string", "example": "pii-redaction" },
          "stage": {
            "type": "string",
            "enum": ["request", "response", "both"],
            "example": "both"
          },
          "policyRef": {
            "type": "string",
            "description": "Reference to a policy defined in the policy schema."
          }
        },
        "required": ["name", "stage"]
      }
    },
    "retries": {
      "type": "object",
      "properties": {
        "attempts": { "type": "integer", "example": 3 },
        "perTryTimeout": { "type": "string", "example": "30s" }
      }
    },
    "observability": {
      "type": "object",
      "description": "Where this route emits telemetry.",
      "properties": {
        "openTelemetry": { "type": "boolean", "example": true },
        "langfuse": { "type": "boolean", "example": false },
        "logRequestBody": { "type": "boolean", "example": false }
      }
    },
    "tenant": {
      "type": "string",
      "description": "Logical tenant or team owning this route.",
      "example": "platform"
    }
  },
  "required": ["name", "backends"]
}