@singlestore/ai

SingleStoreAI

A module that enhances the @singlestore/client package with AI functionality, allowing you to integrate advanced AI features like embeddings and chat completions.

Table of Contents

• Installation
• Usage Examples
  • Create an Instance
  • Generate Embeddings
  • Create a Chat Completion
  • Stream Chat Completions
  • Develop a Chat Completion Tool
  • Custom Chat Completions
  • Custom Embeddings

Installation

To install the @singlestore/ai package, run the following command:

npm install @singlestore/ai

Usage Examples

Create an Instance

First, create an instance of the SingleStoreAI class using your OpenAI API key.

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

const ai = new SingleStoreAI({ openAIApiKey: "<OPENAI_API_KEY>" });

Generate Embeddings

Generate embeddings for a given input text using the create method.

const input = "Hi!";
const embeddings = await ai.embeddings.create(input);
console.log(embeddings);

Create a Chat Completion

Create a chat completion.

const prompt = "Hi, how are you?";
const chatCompletion = await ai.chatCompletions.create({
  prompt,
  model: "gpt-4o",
  systemRole: "You are a helpful assistant",
});
console.log(chatCompletion);

Stream Chat Completions

Stream chat completions to handle responses in real time.

const prompt = "Hi, how are you?";

const stream = await ai.chatCompletions.create({
  prompt,
  model: "gpt-4o",
  systemRole: "You are a helpful assistant",
  stream: true,
});

const onChunk: OnChatCompletionChunk = (chunk) => {
  console.log("onChunk:", chunk);
};

const chatCompletion = await ai.chatCompletions.handleStream(stream, onChunk);
console.log(chatCompletion);

Develop a Chat Completion Tool

Create a custom chat completion tool to handle specific tasks.

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

const findCityInfoTool = new ChatCompletionTool({
  name: "find_city_info",
  description: "Useful for finding and displaying information about a city.",
  params: z.object({ name: z.string().describe("The city name") }),
  call: async (params) => {
    const info = `${params.name} is known as a great city!`;
    return { name: "find_city_info", params, value: JSON.stringify(info) };
  },
});

const ai = new AI({
  openAIApiKey: "<OPENAI_API_KEY>",
  chatCompletionTools: [findCityInfoTool],
});

const chatCompletion = await ai.chatCompletions.create({ prompt: "Find info about Vancouver." });
console.log(chatCompletion);

Custom Chat Completions

Extend the ChatCompletions class to use a custom LLM for creating chat completions.

import { AI, type AnyChatCompletionTool, ChatCompletions } from "@singlestore/ai";

class CustomChatCompletions<
  TChatCompletionTool extends AnyChatCompletionTool[] | undefined,
> extends ChatCompletions<TChatCompletionTool> {
  constructor() {
    super();
  }

  getModels(): Promise<string[]> | string[] {
    // Your implementation
    return [];
  }

  async create<TStream extends boolean | undefined>(
    params: CreateChatCompletionParams<TStream, TChatCompletionTool>,
  ): Promise<CreateChatCompletionResult<TStream>> {
    // Your implementation
    return {} as CreateChatCompletionResult<TStream>;
  }
}

const ai = new AI({
  openAIApiKey: "<OPENAI_API_KEY>",
  chatCompletions: new CustomChatCompletions(),
});

Custom Embeddings

Create a custom embeddings class to use a custom LLM for creating embeddings.

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

class CustomEmbeddings extends Embeddings {
  constructor() {
    super();
  }

  getModels(): string[] {
    // Your implementation
    return [];
  }

  async create(input: string | string[], params?: CreateEmbeddingsParams): Promise<Embedding[]> {
    // Your implementation
    return [];
  }
}

const ai = new AI({
  openAIApiKey: "<OPENAI_API_KEY>",
  embeddings: new CustomEmbeddings(),
});