Incubyte-AI

A flexible and powerful chat library for building AI-powered conversational applications with contextual understanding and document retrieval capabilities.

✨ Features

• 🤖 AI-powered chat with deep context understanding
• 📄 Document retrieval from vector stores (RAG)
• 🔁 Streaming responses for real-time interaction
• 🔗 Seamless integration with LangChain and OpenAI
• 💾 Checkpoint saving for persistent conversations

📦 Installation

npm install @incubyte/ai

🚀 Usage

Basic Example

import { Chat } from "@incubyte/ai";

// Initialize Chat
const chat = new Chat();

// Send a message and get a response
const response = await chat.sendChat("tenant-id", "conversation-id", {
  query: "What is the best way to lose weight?",
});

console.log(response);

Streaming Example

import { Chat } from "@incubyte/ai";

// Initialize Chat
const chat = new Chat();

// Stream a chat conversation
const stream = await chat.sendChatStream("tenant-id", "conversation-id", {
  query: "Tell me about intermittent fasting",
});

// Process the stream
for await (const chunk of stream) {
  if (chunk.type === "message") {
    console.log(chunk.data.content);
  } else if (chunk.type === "context") {
    console.log("Retrieved context:", chunk.data);
  }
}

API Reference

Chat Class

The main class for interacting with the chat system.

Constructor

new Chat(checkpointSaverManager?: BaseCheckpointSaver)

• checkpointSaverManager (optional): An instance of BaseCheckpointSaver from @langchain/langgraph for saving chat state checkpoints.

Methods

sendChat

async sendChat(
  tenantId: string,
  conversationId: string,
  chatRequestDto: ChatRequestDto
): Promise<{ chatResponse: StateType<GraphStateAnnotation> }>

Sends a chat message and returns a structured response.

• tenantId: Identifier for the tenant or knowledge base

• conversationId: Unique identifier for the conversation

• chatRequestDto: Object with the following structure:

{
  query: string;
  metadataFilters?: Record<string, string | string[]>[];
}

sendChatStream

async sendChatStream(
  tenantId: string,
  conversationId: string,
  chatRequestDto: ChatRequestDto
): AsyncGenerator<StreamResponse>

Sends a chat message and returns a stream of response chunks.

• tenantId: Identifier for the tenant or knowledge base
• conversationId: Unique identifier for the conversation
• chatRequestDto: Same structure as above

✅ Requirements

• Node.js 16 or above

• TypeScript or JavaScript project compatibility

License

ISC License

• ISC License