Types

TypeScript interfaces used throughout the library.

`Message`

A single message in a conversation.

interface Message {
  role: 'system' | 'user' | 'assistant'
  content: string
}

Property	Type	Description
`role`	`'system' \| 'user' \| 'assistant'`	The role of the message sender
`content`	`string`	The text content of the message

Example:

const messages: Message[] = [
  { role: 'system', content: 'Be helpful.' },
  { role: 'user', content: 'Hello!' },
  { role: 'assistant', content: 'Hi there!' }
]

`ChatResponse`

Response from a chat completion request.

interface ChatResponse {
  content: string
  usage: TokenUsage
  model: string
  id: string
  finishReason: 'stop' | 'length' | 'content_filter' | null
  latency: number
}

Property	Type	Description
`content`	`string`	The generated text response
`usage`	`TokenUsage`	Token usage statistics
`model`	`string`	The model that generated the response
`id`	`string`	Unique identifier for this completion
`finishReason`	`'stop' \| 'length' \| 'content_filter' \| null`	Why generation stopped
`latency`	`number`	Request latency in milliseconds

Finish Reasons:

'stop' — Natural completion
'length' — Hit max tokens limit
'content_filter' — Response was filtered
null — Unknown or not provided

`TokenUsage`

Token usage statistics for a request.

interface TokenUsage {
  promptTokens: number
  completionTokens: number
  totalTokens: number
}

Property	Type	Description
`promptTokens`	`number`	Tokens in the prompt/input
`completionTokens`	`number`	Tokens in the response/output
`totalTokens`	`number`	Total tokens used

`GenerationParams`

Parameters for controlling text generation.

interface GenerationParams {
  temperature?: number
  maxTokens?: number
  topP?: number
  stop?: string[]
}

Property	Type	Description
`temperature`	`number`	Sampling temperature (0-2). Higher = more random
`maxTokens`	`number`	Maximum tokens to generate
`topP`	`number`	Nucleus sampling threshold (0-1)
`stop`	`string[]`	Sequences that stop generation

`ClientOptions`

Configuration for creating an LLM client.

interface ClientOptions {
  apiKey: string
  model: string
  baseUrl?: string
  defaultParams?: GenerationParams
  referer?: string
  title?: string
}

Property	Type	Required	Description
`apiKey`	`string`	Yes	API key for authentication
`model`	`string`	Yes	Model identifier (e.g., `'anthropic/claude-sonnet-4'`)
`baseUrl`	`string`	No	API endpoint. Default: OpenRouter
`defaultParams`	`GenerationParams`	No	Default generation parameters
`referer`	`string`	No	HTTP-Referer header for OpenRouter
`title`	`string`	No	X-Title header for OpenRouter

`ChatOptions`

Options for a single chat request.

interface ChatOptions extends GenerationParams {
  model?: string
  signal?: AbortSignal
  retry?: boolean
  maxRetries?: number
}

Property	Type	Description
`model`	`string`	Override model for this request
`signal`	`AbortSignal`	Signal to cancel the request
`retry`	`boolean`	Enable retries. Default: `true`
`maxRetries`	`number`	Max retry attempts. Default: `3`
`temperature`	`number`	Sampling temperature (0-2)
`maxTokens`	`number`	Maximum tokens to generate
`topP`	`number`	Nucleus sampling threshold
`stop`	`string[]`	Stop sequences

`ConversationOptions`

Options for creating a conversation.

interface ConversationOptions {
  system?: string
  initialMessages?: Message[]
}

Property	Type	Description
`system`	`string`	System prompt to guide model behavior
`initialMessages`	`Message[]`	Pre-populated message history

`ModelInfo`

Information about a model's capabilities and pricing.

interface ModelInfo {
  contextLength: number
  pricing: {
    prompt: number
    completion: number
  }
}

Property	Type	Description
`contextLength`	`number`	Maximum tokens the model can process
`pricing.prompt`	`number`	Cost per million prompt tokens (USD)
`pricing.completion`	`number`	Cost per million completion tokens (USD)

`LLMClient`

The main client interface.

interface LLMClient {
  chat(messages: Message[], options?: ChatOptions): Promise<ChatResponse>
  stream(messages: Message[], options?: ChatOptions): AsyncIterable<string>
  createConversation(options?: ConversationOptions): Conversation
  getModel(): string
  setModel(model: string): void
  estimateChat(messages: Message[]): { prompt: number; available: number }
}

`Conversation`

A stateful conversation interface.

interface Conversation {
  send(content: string, options?: ChatOptions): Promise<string>
  sendStream(content: string, options?: ChatOptions): AsyncIterable<string>
  history: Message[]
  addMessage(role: 'user' | 'assistant', content: string): void
  clear(): void
  clearAll(): void
}