This React hook creates a chat instance used to interact with the LLM. The result object contains functions and state enabling you to send and receive messages and monitor the state of the chat.
The useStructuredChat hook provides functionality for structured chats. Structured chats are used when you want to use the LLM to generate structured data according to a defined schema. This is particularly useful for: - Generating typed data structures - Creating form responses - Building UI components - Extracting information into a specific format
function useStructuredChat<Schema extends s.HashbrownType , Tools extends Chat.AnyTool , Output extends s.Infer <Schema> = s.Infer <Schema>>(options: UseStructuredChatOptions <Schema, Tools, Output>): UseStructuredChatResult <Output, Tools>;
function useStructuredChat<Schema extends s.HashbrownType, Tools extends Chat.AnyTool, Output extends s.Infer<Schema> = s.Infer<Schema>>(options: UseStructuredChatOptions<Schema, Tools, Output>): UseStructuredChatResult<Output, Tools>;@paramoptionsUseStructuredChatOptions<Schema, Tools, Output>@typeSchemas.HashbrownType@typeToolsChat.AnyTool@typeOutputs.Infer<Schema>@returns {UseStructuredChatResult} An object containing chat state and functions to interact with the chat.
UseStructuredChatResult<Output, Tools>Examples
In this example, the LLM will respond with a JSON object containing the translations of the input message into English, Spanish, and French.
const { messages, sendMessage } = useStructuredChat({
model: 'gpt-4o',
system: 'You are a helpful translator that provides accurate translations.',
schema: s.object('Translations', {
english: s.string('English translation'),
spanish: s.string('Spanish translation'),
french: s.string('French translation')
}),
});