API Reference

Complete API reference for multi-llm-ts.

Core Classes

LlmModel

The main abstraction that wraps an engine and model together. Recommended for most use cases.

const model = igniteModel('openai', chatModel, config)
await model.complete(messages)

LlmEngine

Low-level engine interface for advanced use cases. LlmModel wraps this internally.

const engine = igniteEngine('openai', config)
await engine.complete(messages, chatModel)

Message

Represents a message in the conversation with support for attachments.

const message = new Message('user', 'Hello!')
message.attach({ url: 'image.jpg', mimeType: 'image/jpeg' })

Plugin

Base class for creating custom tools/functions that the LLM can invoke.

class MyPlugin extends Plugin {
  async execute(context, parameters) {
    // Your tool logic
  }
}

Helper Functions

igniteModel()

Create an LlmModel instance:

function igniteModel(
  provider: string,
  model: ChatModel,
  config: EngineConfig
): LlmModel

igniteEngine()

Create an LlmEngine instance:

function igniteEngine(
  provider: string,
  config: EngineConfig
): LlmEngine

loadModels()

Load available models from a provider:

async function loadModels(
  provider: string,
  config: EngineConfig
): Promise<ModelsResponse>

Type Definitions

See Types for a complete reference of all TypeScript types, interfaces, and enums.

Provider IDs

Valid provider identifiers:

• 'openai' - OpenAI (also used for TogetherAI)
• 'anthropic' - Anthropic
• 'google' - Google
• 'ollama' - Ollama
• 'azure' - Azure AI
• 'cerebras' - Cerebras
• 'deepseek' - DeepSeek
• 'groq' - Groq
• 'meta' - Meta/Llama
• 'mistralai' - MistralAI
• 'openrouter' - OpenRouter
• 'xai' - xAI