Region Based Routing

Simple routing based on user's geographic region for data residency and performance optimization.

Route requests to different models based on the user's geographic location. This helps with data residency requirements, regulatory compliance, and performance optimization by selecting geographically closer providers.

Use Case

• Data residency requirements

• Regional performance optimization

• Regulatory compliance (GDPR, data sovereignty)

• Provider availability by region

Configuration

{
  "model": "router/dynamic",
  "router": {
    "type": "conditional",
    "routes": [
      {
        "name": "european_users",
        "conditions": {
          "all": [
            {
              "metadata.region": {
                "$eq": "Europe"
              }
            }
          ]
        },
        "targets": {
          "$any": [
            "openai/gpt-5",
            "anthropic/claude-opus-4.1"
          ],
          "sort_by": "ttft",
          "sort_order": "min"
        }
      },
      {
        "name": "north_american_users",
        "conditions": {
          "all": [
            {
              "metadata.country": {
                "$in": ["US", "CA", "MX"]
              }
            }
          ]
        },
        "targets": {
          "$any": [
            "openai/gpt-4.1",
            "google/gemini-2.5-pro"
          ],
          "sort_by": "ttft",
          "sort_order": "min"
        }
      },
      {
        "name": "asian_users",
        "conditions": {
          "all": [
            {
              "metadata.region": {
                "$eq": "Asia"
              }
            }
          ]
        },
        "targets": {
          "$any": [
            "deepseek/deepseek-chat",
            "kimi-k2"
          ],
          "sort_by": "ttft",
          "sort_order": "min"
        }
      },
      {
        "name": "global_fallback",
        "conditions": {
          "all": []
        },
        "targets": "openai/gpt-4.1-mini"
      }
    ]
  }
}

How it works

1. European Users (using metadata.region = "Europe"): Get GPT-5 or Claude-Opus-4.1 models optimized for speed

2. North American Users (using specific country codes US, CA, MX): Get GPT-4.1 or Gemini-2.5-pro models optimized for speed

3. Asian Users (using metadata.region = "Asia"): Get DeepSeek-chat or Kimi-k2 models optimized for speed

4. Global Fallback: Any unmatched regions/countries get GPT-4.1-mini

Variables Used

• metadata.region: User's geographic region (automatically detected by LangDB based on IP address)

• metadata.country: User's country code (automatically detected by LangDB based on IP address)

Geographic Detection

LangDB automatically detects user location based on IP address geolocation and provides both:

• Regions: Broad geographic areas like "Europe", "Asia", "North America", etc.

• Countries: ISO 3166-1 alpha-2 country codes like "US", "GB", "DE", "CN", "JP", etc.

• No additional setup required - works out of the box

Customization

• Region-based: Use metadata.region for broad geographic routing ("Europe", "Asia", "Africa", etc.)

• Country-specific: Use metadata.country for precise country targeting ("US", "DE", "JP", etc.)

• Mixed approach: Combine both region and country routing in the same configuration

• Include region-specific model providers (e.g., local AI providers)

• Add failover logic for regional outages