@singlestore/ai

@singlestore/ai

@singlestore/ai is a module of the @singlestore/client package that provides AI capabilities, including chat completions and embeddings, integrated with OpenAI models. Additionally, it offers flexibility for users to integrate their own AI models and tools.

Features

• Chat Completions: Leverage OpenAI's chat models to generate completions, manage tools, and handle streams of chat data. Alternatively, integrate your own AI models.
• Embeddings: Generate text embeddings using OpenAI's embedding models to power various AI-driven functionalities, or use your custom embedding models.

Installation

To install the @singlestore/ai module, use npm:

npm install @singlestore/client

Usage

Basic Setup

First, initialize the AI module with your OpenAI API key or your custom AI instance:

import { AI } from "@singlestore/ai";

const ai = new AI({
  openAIApiKey: "your-openai-api-key", // or your custom AI instance
});

Custom AI Models and Tools

You can provide your own AI models and tools instead of using OpenAI. For example:

import { AI, ChatCompletionTool } from "@singlestore/ai";

const customTool = new ChatCompletionTool({
  name: "CustomTool",
  description: "A custom tool for chat completion",
  params: {},
  call: async (params) => {
    // Custom tool logic
  },
});

const ai = new AI({
  chatCompletionTools: [customTool],
  embeddings: new CustomEmbeddings(), // Your custom embeddings class
});

Chat Completions

Create chat completions with integrated tools and streaming capabilities:

const completion = await ai.chatCompletions.create({
  prompt: "Tell me a joke.",
  systemRole: "You are a helpful assistant.",
  stream: false,
});

console.log(completion.content);

Embeddings

Generate embeddings for your text:

const embeddings = await ai.embeddings.create("This is a sample text.");
console.log(embeddings);

API Reference

The following section describes the main classes and methods available in the @singlestore/ai module.

AI

The main class that provides access to chat completions and embeddings.

Constructor

new AI(config: AIConfig)

Parameters:

• config.openAIApiKey (string): Your OpenAI API key.
• config.chatCompletionTools (ChatCompletionTool[] | undefined): An optional array of tools to be used in chat completions.
• config.embeddings (Embeddings | undefined): An optional custom embeddings instance.

ChatCompletions

Handles creating chat completions and managing the stream of chat data.

Methods:

async create(params: OpenAICreateChatCompletionParams)

Creates a chat completion with optional support for tools and streaming.

Parameters:

• prompt (string | undefined): The initial user prompt.
• systemRole (string | undefined): The system role for the AI.
• messages (ChatMessage[] | undefined): An array of messages in the chat.
• stream (boolean | undefined): If true, returns an async generator for streaming completions.
• tools (ChatCompletionTool[] | undefined): Tools to be used during chat completion.
• toolCallHandlers (ToolCallHandlers | undefined): Handlers for tool calls.
• toolCallResultHandlers (ToolCallResultHandlers | undefined): Handlers for tool call results.

Returns:

• Promise<ChatCompletion | AsyncGenerator<ChatCompletionStream>>: A promise that resolves to either a single chat completion or a stream of chat completions.

initTools(tools: ChatCompletionTool[])

Initializes tools to be used in chat completions.

Parameters:

• tools (ChatCompletionTool[]): An array of tools.

Embeddings

Handles generating text embeddings using OpenAI models.

Methods:

async create(input: string | string[], params?: OpenAICreateEmbeddingsParams)

Generates embeddings for the provided input text.

Parameters:

• input (string | string[]): The input text or array of texts to generate embeddings for.
• params (OpenAICreateEmbeddingsParams | undefined): Optional parameters for creating embeddings.

Returns:

• Promise<Embedding[]>: A promise that resolves to an array of embedding vectors generated by OpenAI.

Data Types

ChatCompletion

An object representing a chat completion.

Properties:

• content (string): The content of the chat completion.

ChatCompletionStream

An async generator that yields chunks of chat completions.

Embedding

An object representing a text embedding.

Properties:

• embedding (number[]): An array of numbers representing the embedding vector.

ChatMessage

An object representing a message in the chat.

Properties:

• role (string): The role of the message sender (e.g., 'system', 'user', 'assistant').
• content (string): The content of the message.

ChatCompletionTool

An object representing a tool used in chat completions.

Properties:

• name (string): The name of the tool.
• description (string): A description of what the tool does.
• params (object | undefined): Parameters required by the tool.
• call (function): The function that executes the tool.

OpenAICreateChatCompletionParams

Parameters used for creating a chat completion.

Properties:

• model (string): The model to use for the chat completion (e.g., 'gpt-4').
• temperature (number | undefined): Sampling temperature for randomness in outputs.
• prompt (string | undefined): The initial user prompt.
• systemRole (string | undefined): The system role for the AI.
• messages (ChatMessage[] | undefined): An array of messages in the chat.
• stream (boolean | undefined): If true, returns an async generator for streaming completions.
• tools (ChatCompletionTool[] | undefined): Tools to be used during chat completion.

OpenAICreateEmbeddingsParams

Parameters used for creating embeddings.

Properties:

• model (string): The model to use for generating embeddings (e.g., 'text-embedding-ada-002').
• input (string | string[]): The input text or array of texts to generate embeddings for.