Custom Models

Overview

OpenCaddis requires two model configs: default (chat/completions) and embeddings (vector embeddings). Beyond those, you can add any number of custom named model configs — each pointing to a different provider, model, or endpoint.

Custom configs are defined in the Settings UI and stored in fabr.json. Once saved, they appear in the Agent Configuration tab's model dropdown, so you can assign any model to any agent.

Supported Providers

Selecting a provider in the UI auto-fills the endpoint URI where applicable. Any provider that exposes an OpenAI-compatible API can be used by entering its endpoint manually.

Adding a Custom Model

1. Go to Settings > Model Configuration
2. Scroll to the Additional Models section
3. Click Add Model
4. Fill in the card:

   Field	Description	Example
   Name	Unique lowercase identifier. Alphanumeric + hyphens. Cannot be default or embeddings.	grok
   Provider	Select from dropdown. Auto-fills the endpoint.	Grok
   Endpoint URI	API endpoint URL for the provider.	https://api.x.ai/v1
   API Key	Provider API key. Masked by default with show/hide toggle.	xai-...
   Model	Model identifier.	grok-4-latest
   Timeout (s)	Request timeout in seconds.	60
   Max Tokens	Optional max output token limit.	4096

5. Click Save Configuration
6. Go to the Agent Config tab — the new model appears in the Models dropdown
7. Assign it to any agent

fabr.json Structure

Custom models are stored alongside default and embeddings in the ModelConfigurations array:

{
  "ModelConfigurations": [
    { "Name": "default", "Provider": "Azure", ... },
    { "Name": "embeddings", "Provider": "Azure", ... },
    {
      "Name": "grok",
      "Provider": "Grok",
      "Uri": "https://api.x.ai/v1",
      "Model": "grok-4-latest",
      "ApiKeyAlias": "grok-key",
      "TimeoutSeconds": 60
    },
    {
      "Name": "gemini",
      "Provider": "Gemini",
      "Uri": "https://generativelanguage.googleapis.com/v1beta/openai/",
      "Model": "gemini-2.5-flash",
      "ApiKeyAlias": "gemini-key",
      "TimeoutSeconds": 60
    }
  ],
  "ApiKeys": [
    { "Alias": "default-key", "Value": "..." },
    { "Alias": "grok-key", "Value": "..." },
    { "Alias": "gemini-key", "Value": "..." }
  ]
}

Assigning to Agents

Once a custom model is saved, it appears in the Agent Configuration tab's Models dropdown. Select it for any agent to set Args["ModelConfig"] on that agent's configuration.

At runtime, when the agent calls GetChatClient("grok"), FabrCore's ModelConfigController looks up the config by name from fabr.json and creates a client with the matching provider, endpoint, model, and API key.

// In your agent's OnInitialize()
var client = await GetChatClient("grok");  // resolves from fabr.json
_agent = new ChatClientAgent(client)
    .AsBuilder()
    .UseOpenTelemetry()
    .Build(ServiceProvider);

You can have different agents using different providers simultaneously — one agent on Azure OpenAI, another on Grok, another on Gemini — all running in the same OpenCaddis instance.

Import & Export

The Export and Import buttons on the Model Configuration tab handle additional models automatically. Custom model configs are included in the exported JSON and restored on import. This makes it easy to share or back up your multi-provider setup.

Example Configurations