Model Routing & Mapping

Configure intelligent model routing and custom mappings for API requests

Overview

Model routing allows you to map incoming model requests to different backend models. This enables seamless integration with existing tools while leveraging the best available models.

Custom Mappings

Adding Mappings

Create custom model mappings in the API Proxy settings:

1. Navigate to Model Router

   Open API Proxy → Model Router & Mapping

2. Add Mapping

   Select the source model pattern and target model from the dropdowns.

3. Apply Changes

   Click Add Mapping to save the configuration.

// From src/pages/ApiProxy.tsx:462
const handleMappingUpdate = async (type: 'custom', key: string, value: string) => {
  const newConfig = { ...appConfig.proxy };
  newConfig.custom_mapping = { 
    ...(newConfig.custom_mapping || {}), 
    [key]: value 
  };
  
  await invoke('update_model_mapping', { config: newConfig });
  showToast(t('common.saved'), 'success');
};

Wildcard Patterns

Use wildcards for flexible routing:

// Example mappings
{
  "gpt-4*": "gemini-3.1-pro-high",      // All GPT-4 variants
  "gpt-4o*": "gemini-3-flash",          // GPT-4o series
  "claude-3-5-sonnet-*": "claude-sonnet-4-6",  // Claude Sonnet
  "o1-*": "gemini-3.1-pro-high"         // O1 models
}

Wildcard matching is case-insensitive and uses suffix/prefix matching.

Preset Configurations

Antigravity includes several built-in presets for common use cases:

Default Preset

Balanced performance and compatibility:

// From src/pages/ApiProxy.tsx:513
{
  id: 'default',
  mappings: {
    "gpt-4*": "gemini-3.1-pro-high",
    "gpt-4o*": "gemini-3-flash",
    "gpt-3.5*": "gemini-2.5-flash",
    "o1-*": "gemini-3.1-pro-high",
    "claude-3-5-sonnet-*": "claude-sonnet-4-6",
    "claude-3-opus-*": "claude-opus-4-6-thinking",
    "claude-haiku-*": "gemini-2.5-flash"
  }
}

Performance Preset

Maximize response quality:

// From src/pages/ApiProxy.tsx:526
{
  id: 'performance',
  mappings: {
    "gpt-4*": "claude-opus-4-6-thinking",
    "gpt-4o*": "claude-sonnet-4-6",
    "gpt-3.5*": "gemini-3-flash",
    "o1-*": "claude-opus-4-6-thinking",
    "claude-3-5-sonnet-*": "claude-sonnet-4-6",
    "claude-haiku-*": "claude-sonnet-4-6"
  }
}

Cost-Effective Preset

Optimize for quota consumption:

// From src/pages/ApiProxy.tsx:544
{
  id: 'cost-effective',
  mappings: {
    "gpt-4*": "gemini-3-flash",
    "gpt-4o*": "gemini-2.5-flash",
    "gpt-3.5*": "gemini-2.5-flash",
    "o1-*": "gemini-3-flash",
    "claude-3-5-sonnet-*": "gemini-3-flash",
    "claude-3-opus-*": "gemini-3-flash",
    "claude-haiku-*": "gemini-2.5-flash"
  }
}

Applying Presets

// From src/pages/ApiProxy.tsx:641
const handleApplyPresets = async () => {
  const selectedPresetData = presetOptions.find(p => p.id === selectedPreset);
  if (!selectedPresetData) return;
  
  const newConfig = {
    ...appConfig.proxy,
    custom_mapping: { 
      ...appConfig.proxy.custom_mapping, 
      ...selectedPresetData.mappings 
    }
  };
  
  await invoke('update_model_mapping', { config: newConfig });
};

Presets merge with existing custom mappings without overwriting unrelated entries.

Custom Presets

Save your current mappings as a reusable preset:

1. Configure Mappings

   Set up your desired model mappings in the router.

2. Open Preset Manager

   Click Save as Preset in the Model Router section.

3. Name Your Preset

   Enter a descriptive name and optional description.

4. Save

   Click Save Preset to add it to your collection.

// From src/pages/ApiProxy.tsx:606
const handleSaveCurrentAsPreset = () => {
  const newPreset: CustomPreset = {
    id: `custom_${Date.now()}`,
    name: newPresetName,
    description: t('proxy.router.custom_preset_desc'),
    mappings: { ...appConfig.proxy.custom_mapping }
  };
  
  const updatedPresets = [...customPresets, newPreset];
  localStorage.setItem('antigravity_custom_presets', JSON.stringify(updatedPresets));
};

Dynamic Model Forwarding

Automatic forwarding for deprecated models:

// From src-tauri/src/proxy/token_manager.rs:511
if let Some(rules) = account.get("quota")
    .and_then(|q| q.get("model_forwarding_rules"))
    .and_then(|r| r.as_object()) 
{
    for (k, v) in rules {
        if let Some(new_model) = v.as_str() {
            update_dynamic_forwarding_rules(
                k.to_string(),
                new_model.to_string()
            );
        }
    }
}

Dynamic forwarding rules are automatically synchronized from account quota data.

Tiered Routing

Intelligent routing based on account tiers:

// From src-tauri/src/proxy/token_manager.rs:1055
tokens_snapshot.sort_by(|a, b| {
    let tier_priority = |tier: &Option<String>| {
        let t = tier.as_deref().unwrap_or("").to_lowercase();
        if t.contains("ultra") { 0 }
        else if t.contains("pro") { 1 }
        else if t.contains("free") { 2 }
        else { 3 }
    };
    
    tier_priority(&a.subscription_tier)
        .cmp(&tier_priority(&b.subscription_tier))
});

The routing system prioritizes:

1. Ultra Accounts: Highest tier, best quota resets
2. Pro Accounts: Mid-tier with good quotas
3. Free Accounts: Basic tier, limited quotas

Background Request Detection

Automatic downgrading for background tasks:

// Detect background requests from tools like Claude CLI
// and route them to Flash models to preserve Pro/Ultra quota
const isBackgroundRequest = detectBackgroundTask(request);
if (isBackgroundRequest) {
  targetModel = "gemini-3-flash"; // Downgrade to Flash
}

Background task detection is based on request patterns and may not catch all cases.

Model Normalization

Standardize model identifiers:

// Example normalization rules
"gemini-3-pro-high" → "gemini-3.1-pro-high"
"gemini-3-pro-low" → "gemini-3.1-pro-low"
"claude-sonnet-4-5" → "claude-sonnet-4-6"

Quota Protection Integration

Model routing respects quota protection settings:

// From src-tauri/src/proxy/token_manager.rs:1184
let is_quota_protected = quota_protection_enabled
    && token.protected_models.contains(&normalized_target);

if is_quota_protected {
    // Skip this account for the requested model
    continue;
}

Resetting Mappings

Reset all custom mappings to defaults:

// From src/pages/ApiProxy.tsx:486
const executeResetMapping = async () => {
  const newConfig = {
    ...appConfig.proxy,
    custom_mapping: {}
  };
  
  await invoke('update_model_mapping', { config: newConfig });
  showToast(t('common.success'), 'success');
};

Resetting mappings will remove all custom and preset configurations.

Best Practices

1. Use Wildcards Wisely: Pattern gpt-4* catches all GPT-4 variants
2. Test Mappings: Verify mappings work with your specific tools
3. Save Presets: Save working configurations before experimenting
4. Monitor Quotas: Check quota usage after changing mappings
5. Document Custom Presets: Add descriptions to remember mapping purposes

Related Features

• API Proxy - Configure proxy settings
• Quota Monitoring - Track model usage
• Account Management - Manage accounts