Skip to content
Latestv1.0.63

useAiChat Data Responsive Stream ​

@yh-ui/hooks provides extremely powerful Headless Hooks Abstraction for AI workflows. The useAiChat and useAiConversations data engines completely separate visual rendering from pure data modeling.

In a traditional setup, handling AI model REST API requires maintaining Token streams, handling Stop/Abort capabilities, managing loading states, and keeping conversations synchronized.

Now, simply inject a custom request functional interceptor to useAiChat, and unlock the full-link state cycle immediately.

Basic Integration ​

Use simple Promises to take over the isolation state without manually doing messages.push.

Minimal Engine

Native Event Stream and Instant Abort ​

Most large language models utilize text/event-stream for lower Time-To-First-Byte (TTFB). Just turn your request engine into an AsyncGenerator (or yield streams). The Hook core automatically converts chunks into brilliant & silky-smooth continuous typewriter outputs!

It safely halts all subsequent fetches entirely by invoking the native browser AbortController from stop().

Data Streaming with Native Abort

Conversation Sidebar Navigation (useAiConversations) ​

A complete ChatGPT-like client requires managing side menus. The useAiConversations hook quickly sets up a managed conversation list for sidebar state.

Conversation Management Hook

API Reference ​

useAiChat Options ​

PropertyDescriptionType
initialMessagesArray of default chat recordsAiChatMessage[]
requestRequest adapter with signature (message, history, abortSignal) => AsyncGenerator | Promise<string | Response>Function
idGeneratorReplaces the default random ID generator() => string
parserSSE / stream chunk parser for different vendor formatsStreamChunkParser
typewriterEnable or disable the built-in typewriter effectboolean
charsPerFrameNumber of characters rendered per frame in typewriter modenumber
systemPromptSystem prompt automatically prepended to request historystring
onErrorError callback(err) => void
onFinishCallback fired when the assistant message finishes(message) => void

useAiChat Returns ​

The state hook unpacks all data and methods automatically:

PropertyDescriptionType
messagesReturns state representation bound directly to conversations.Ref<AiChatMessage[]>
isGeneratingWhether the assistant is currently generating.Ref<boolean>
isSendingSemantic alias of isGenerating.ComputedRef<boolean>
sendMessageMain trigger appending custom queries to AI(content: string) => Promise<void>
stopStop the current generation and settle message state.() => void
clearClears local history tracking window() => void
removeMessageSingle deletion handler(id: string) => void
updateMessageUpdate a specific message in place(id: string, patch: Partial<AiChatMessage>) => void

Released under the MIT License.