API reference

Client

new FreeTheAI(options?) takes the configuration options.

Member	Signature
client.apiKey	The cleaned key in use
client.baseURL	The API address in use
client.search	search({ query, num_results? }): Promise<WebSearchResponse>
client.answer	answer({ query, num_results? }): Promise<WebAnswerResponse>

Resources

Resource	Methods
client.chat.completions	create(params): Promise<ChatCompletion>, or create({ ..., stream: true }): Promise<AsyncIterable<ChatCompletionChunk>>; streamText(params): AsyncGenerator<string>
client.models	list(): Promise<Model[]>
client.tasks	list({ limit? }), get(id), wait(id, { pollIntervalMs?, timeoutMs? }), content(id, { index? }): Promise<Uint8Array>
client.videos	create({ model, prompt?, input_reference?, seconds?, size? })
client.images	generate({ model, prompt, n?, size?, response_format?, image? })

Chat parameters are sent as they are, with the API's snake_case names: temperature, top_p, max_tokens, stop, seed, tools, tool_choice, response_format, reasoning_effort, and web_search_options. Add fallbackModels to a single call to override the client's list.

Media and search calls return the API's JSON. tasks.wait resolves with the finished task, or rejects with TaskFailedError.

Types

Every request and response type is exported: ChatCompletionCreateParams, ChatCompletion, ChatCompletionChunk, ChatMessage, ToolCall, FunctionTool, Usage, Model, MediaTask, VideoCreateParams, ImageGenerateParams, WebSearchResponse, WebAnswerResponse, and more.

Errors

Every error class is exported from the package root. See Errors for the full list and when each one is thrown.