ModelsProvider Connectors

The ModelsProvider subsystem manages LLM model configurations and availability.

If you are looking for how to configure custom models for a smythOS SDK project, take a look at this video.

Available Connectors

JSONModelsProvider

Role: JSON-based model configuration provider
Summary: Provides model configurations from JSON files. Manages model metadata, capabilities, and provider-specific settings.

Setting	Type	Required	Default	Description
models	string | TLLMModelsList	No	Built-in models	Directory path to models.json or models object
mode	string	No	"merge"	How to handle custom models: "merge" or "replace"

Example Configuration:

import { SRE } from '@smythos/sre';
//default : this is the default config used by the SDK
//it will look for models in the .smyth/models or ~/.smyth/models
SRE.init({
    ModelsProvider: {
        Connector: 'JSONModelsProvider',
    },
});

//you can customize the models folder
SRE.init({
    ModelsProvider: {
        Connector: 'JSONModelsProvider',
        Settings: {
            models: './config/models.json',
            mode: 'merge',
        },
    },
});
Copy

Note: If you are using SmythOS SDK, the SDK will initialize SRE for you with the default config, you only need to call SRE.init if you need to set custom configuration for the ModelsProvider connector.

Configuration Options:

Models Setting

• String path: Path to directory containing models .json files
• Object: Direct TLLMModelsList object with model definitions
• Undefined: Uses built-in model configurations

Mode Setting

• "merge": Combines custom models with built-in models
• "replace": Replaces built-in models entirely with custom ones

Use Cases:

• Custom model configurations
• Adding organization-specific models
• Environment-specific model settings
• Model metadata management

File Watching: When models is a directory path, the connector automatically watches for changes and reloads model configurations.

Custom Models Configuration

Model JSON Structure

Custom models are defined in JSON files with the following structure:

{
    "model-key": {
        "label": "Human readable name",
        "modelId": "actual-model-identifier",
        "provider": "ProviderName",
        "features": ["text", "image", "tools", "reasoning", "search"],
        "tags": ["New", "Personal"],
        "tokens": 128000,
        "completionTokens": 8192,
        "enabled": true,
        "credentials": "vault",
        "baseURL": "https://custom-api.example.com/v1"
    }
}
Copy

Model Properties

Property	Type	Required	Description
label	string	Yes	Human-readable display name
modelId	string	Yes	Actual model identifier used by the provider
provider	string	Yes	Provider name (OpenAI, Anthropic, etc.)
features	array	Yes	Supported features: text, image, tools, reasoning, search
tags	array	No	Categorization tags (e.g., "New", "Personal")
tokens	number	Yes	Maximum input tokens
completionTokens	number	Yes	Maximum output tokens
enabled	boolean	No	Whether the model is available (default: true)
credentials	string|object	Yes	Credential configuration (see below)
baseURL	string	No	Custom API endpoint URL
maxReasoningTokens	number	No	Max reasoning tokens (for reasoning models)
default	boolean	No	Mark as default model for the provider

You can use custom models json files to declare online models of known providers (OpenAI, Anthropic, etc.) or local models.

Credentials Configuration

Vault Credentials

Use "vault" to load credentials from the SRE vault:

{
    "my-model": {
        "credentials": "vault"
    }
}
Copy

Direct Credentials

Provide credentials directly as a JSON object:

{
    "my-model": {
        "credentials": {
            "apiKey": "your-api-key-here"
        }
    }
}
Copy

Declaring Local Models via OpenAI API

To use local models exposed through OpenAI-compatible APIs:

1. Set provider to "OpenAI"
2. Specify the local API endpoint in baseURL
3. Use appropriate credentials (often a placeholder API key)

{
    "openai/gpt-oss-20b": {
        "modelId": "openai/gpt-oss-20b",
        "provider": "OpenAI",
        "features": ["text", "tools"],
        "tokens": 4096,
        "completionTokens": 2048,
        "default": true,
        "credentials": "vault",
        "baseURL": "http://127.0.0.1:1234/v1"
    },
    "gemma-3n-e4b": {
        "modelId": "google/gemma-3n-e4b",
        "provider": "OpenAI",
        "features": ["text", "tools"],
        "tokens": 4096,
        "completionTokens": 2048,
        "default": true,
        "credentials": "vault",
        "baseURL": "http://127.0.0.1:1234/v1"
    }
}
Copy