model_aliasing.md

Model Aliasing Feature Testing

Feature Overview

The model aliasing feature allows users to create custom model aliases that reference hardcoded provider models. When creating a model, any missing information (name, pricing, capabilities) is automatically inferred from the hardcoded models.

How It Works

1. Model Reference Format: provider:model_id (e.g., openrouter:deepseek/deepseek-r1-0528-qwen3-8b:free)
2. Automatic Inference with Partial Support:
   • Name: If empty → uses hardcoded model name
   • Pricing: If not provided → uses hardcoded model pricing
     • Partial Support: If only input_token_price is provided → inherits output_token_price from hardcoded model
     • Partial Support: If only output_token_price is provided → inherits input_token_price from hardcoded model
   • Capabilities: If not provided → uses hardcoded model capabilities
     • Partial Support: If only some capabilities are provided → inherits missing capabilities from hardcoded model
     • Note: For boolean fields, explicit false values will override hardcoded true values

Example Usage

Create Model with Full Inference

{
  "model": {
    "id": "my-deepseek-alias",
    "model_reference": "openrouter:deepseek/deepseek-r1-0528-qwen3-8b:free",
    "provider_id": "openrouter", 
    "credential_id": "uuid-here",
    "credential_type": "openrouter"
  }
}

This will create a model with:

• Custom ID: my-deepseek-alias
• Inferred name: DeepSeek-R1-0528-Qwen3-8B
• Inferred pricing: { input: 0.0, output: 0.0 }
• Inferred capabilities: Full capabilities from hardcoded model

Create Model with Partial Override

{
  "model": {
    "id": "custom-deepseek",
    "name": "My Custom DeepSeek",
    "model_reference": "openrouter:deepseek/deepseek-r1-0528-qwen3-8b:free",
    "provider_id": "openrouter",
    "credential_id": "uuid-here", 
    "credential_type": "openrouter"
  }
}

This will create a model with:

• Custom ID: custom-deepseek
• Custom name: My Custom DeepSeek
• Inferred pricing and capabilities from hardcoded model

Create Model with Partial Pricing Override

{
  "model": {
    "id": "premium-deepseek",
    "model_reference": "openrouter:deepseek/deepseek-r1-0528-qwen3-8b:free",
    "provider_id": "openrouter",
    "credential_id": "uuid-here",
    "credential_type": "openrouter",
    "pricing": {
      "input_token_price": 0.001
    }
  }
}

This will create a model with:

• Custom ID: premium-deepseek
• Inferred name: DeepSeek-R1-0528-Qwen3-8B
• Partial pricing: input_token_price: 0.001, output_token_price: 0.0 (inherited)
• Inferred capabilities from hardcoded model

Create Model with Partial Capabilities Override

{
  "model": {
    "id": "limited-deepseek", 
    "model_reference": "openrouter:deepseek/deepseek-r1-0528-qwen3-8b:free",
    "provider_id": "openrouter",
    "credential_id": "uuid-here",
    "credential_type": "openrouter",
    "capabilities": {
      "max_input_tokens": 16384,
      "supports_vision": false
    }
  }
}

This will create a model with:

• Custom ID: limited-deepseek
• Inferred name and pricing from hardcoded model
• Partial capabilities: max_input_tokens: 16384, supports_vision: false, all other capabilities inherited from hardcoded model

Available Hardcoded Models

Currently available in internal/models/config.go:

1. openrouter:deepseek/deepseek-r1-0528-qwen3-8b:free

   • Name: "DeepSeek-R1-0528-Qwen3-8B"
   • Pricing: Free (0.0/0.0)
   • Capabilities: Full set with reasoning support

2. openrouter:meta-llama/llama-4-maverick-17b-128e-instruct:free

   • Name: "Llama-4-Maverick-17b-128e-instruct"
   • Pricing: Free (0.0/0.0)
   • Capabilities: Full set with reasoning support