Skip to content

singlestore-labs/elegance-sdk

Repository files navigation

SingleStore Elegance SDK

Attention: The code in this repository is intended for experimental use only and is not fully tested, documented, or supported by SingleStore. Visit the SingleStore Forums to ask questions about this repository.

The Elegance SDK is an SDK for quickly building real-time AI full-stack JavaScript applications using SingleStoreDB with support for MySQL and Kai (support for Mongo APIs) connection types and the OpenAI API. It provides a set of ready-made tools for implementing various types of functionality when you need to transact, analyze, and contextualize data in real-time or build real-time AI apps.

Features

  • Dot product search
  • Chat completions
  • File embeddings generation (csv, pdf)
  • SQL and aggregate queries
  • SQL and Kai (MongoDB) database connections support
  • Ready-to-use Node.js controllers and React.js hooks

Installation

npm install @singlestore/elegance-sdk

Usage

This guide will show you how to use the SDK in Express.js and React.js.

  1. Create a eleganceServerClient.ts file
  2. Import the createEleganceServerClient function from @singlestore/elegance-sdk/server
import { createEleganceServerClient } from "@singlestore/elegance-sdk/server";

export const eleganceServerClient = createEleganceServerClient("mysql", {
  connection: {
    host: "<DB_HOST>"
    user: "<DB_USER>"
    password: "<DB_PASSWORD>"
    database: "<DB_NAME>"
  },
  ai: {
    openai: {
      apiKey: "<OPENAI_API_KEY>"
    }
  }
});

In case if you don't want to use OpenAI, you can replace the existing AI SDK logic with customizers:

...
import {
  EmbeddingInput,
  CreateEmbeddingResult,
  CreateChatCompletionParams,
  CreateChatCompletionResult
} from "@singlestore/elegance-sdk/types";

...
ai: {
  customizers: {
   createEmbedding: async (input: EmbeddingInput): Promise<CreateEmbeddingResult> => {
     const embedding = await customFn(input);
     return embedding;
  },

   createChatCompletion: async (params: CreateChatCompletionParams): Promise<CreateChatCompletionResult> => {
    const chatCompletion = await customFn(params);
    return chatCompletion;
  }
 }
}
...
  1. Create a route handler for elegance/:route (using Express.js as an example).
import express from "express";
import type { Routes } from "@singlestore/elegance-sdk/server";
import { eleganceServerClient } from "@/services/eleganceServerClient";

export const eleganceRouter = express.Router();

eleganceRouter.post("/elegance/:route", async (req, res) => {
  try {
    const route = req.params.route as Routes;
    const result = await eleganceServerClient.handleRoute(route, req.body);
    return res.status(200).json(result);
  } catch (error: any) {
    return res.status(error.status).json(error);
  }
});
  1. Import the eleganceRouter into the ./server/index.ts
import express from "express";
import { eleganceRouter } from "./routes/elegance";

const app = express();

app.use(eleganceRouter);

app.listen(4000, () => {
  console.log(`Server started on port: ${4000}`);
});
  1. Run the server
  2. Create a eleganceClient.ts file
  3. Import the createEleganceClient function from @singlestore/elegance-sdk
import { createEleganceClient } from "@singlestore/elegance-sdk";

export const eleganceClient = createEleganceClient("mysql", {
  baseURL: "http://localhost:4000",
});
  1. Import the eleganceClient to your component
import { useEffect } from "react";
import { eleganceClient } from "@/services/eleganceClient";

export function ExampleComponent() {
  const query = eleganceClient.hooks.useQuery<{ name: string }[]>({
    initialValue: [],
    initialIsLoading: true,
  });
  const { execute: executeQuery } = query;

  useEffect(() => {
    executeQuery({ query: "SELECT name FROM table LIMIT 3" });
  }, [executeQuery]);

  if (query.isLoading) return "Loading...";

  return (
    <div>
      {query.value.map((item) => (
        <h4 key={item.name}>{item.name}</h4>
      ))}
    </div>
  );
}
  1. Run your client

Templates

You can find templates using the Elegance SDK here:

Example apps

You can find example apps using the Elegance SDK here:

API

createEleganceServerClient

Creates an EleganceServerClient instance for a server.

Parameters:

  • connectionType: "kai" | "mysql"
  •   config: {
        connection: KaiConnectionConfig | MySQLConnectionConfig;
        ai?: {
          openai?: OpenAIConfig;
          customizers?: {
            createEmbedding?: (input: EmbeddingInput) => Promise<CreateEmbeddingResult>;
            createChatCompletion?: (params: CreateChatCompletionParams) => Promise<CreateChatCompletionResult>;
        } // You can use your own functions to create an embedding or a chat completion.
      };
    }

Returns: eleganceServerClient

eleganceServerClient

Server client that includes a database connection, controllers and AI client. It's used on the server to handle requests from the client and execute logic.

eleganceServerClient.connection

MySQL or MongoDB connection to interact with a database

eleganceServerClient.handleRoute

Accepts a route and executes the controller for that route.

Parameters:

  • route: string - controller route name
  • body: object - controller body

eleganceServerClient.controllers.insertOne<T>

Inserts one record.

Parameters:

Kai

body: {
    db?: string;
    collection: string;
    generateId?: boolean;
    value: MongoOptionalUnlessRequiredId<T>;
    options?: MongoInsertOneOptions;
}

MySQL

body: {
    db?: string;
    collection: string;
    generateId?: boolean;
    value: T;
}

Returns: T

eleganceServerClient.controllers.insertMany<Array<T>>

Inserts many records.

Parameters:

Kai

body: {
    db?: string;
    collection: string;
    values: Array<MongoOptionalUnlessRequiredId<T[number]>>;
    generateId?: boolean;
    options?: MongoBulkWriteOptions;
}

MySQL

body: {
    db?: string;
    collection: string;
    generateId?: boolean;
    values: Array<T>;
}

Returns: Array<T>

eleganceServerClient.controllers.updateMany<Array<T>>

Updates many records.

Parameters:

Kai

body: {
    db?: string;
    collection: string;
    filter: MongoFilter<T[number]>;
    update: MongoUpdateFilter<T[number]>;
    options?: MongoUpdateOptions;
    updatedFilter?: MongoFilter<T[number]>;
}

MySQL

body: {
    db?: string;
    collection: string;
    where: MySQLWhere;
    set: MySQLSet;
    updatedWhere: MySQLWhere;
}

Returns: Array<T>

eleganceServerClient.controllers.deleteMany<T>

Deletes many records.

Parameters:

Kai

body: {
    db?: string;
    collection: string;
    filter: MongoFilter<T>;
    options?: MongoDeleteOptions;
}

MySQL

body: {
    db?: string;
    collection: string;
    where: MySQLWhere;
}

Returns: { message: string }

eleganceServerClient.controllers.findOne<T>

Gets one record.

Parameters:

Kai

body: {
    db?: string;
    collection: string;
    filter?: MongoFilter<T>;
    options?: MongoFindOptions;
}

MySQL

body: {
    db?: string;
    collection: string;
    columns?: string[];
    where?: MySQLWhere;
    extra?: string;
}

Returns: T

eleganceServerClient.controllers.findMany<Array<T>>

Gets many records.

Parameters: Kai

body: {
    db?: string;
    collection: string;
    filter?: MongoFilter<T[number]>;
    options?: MongoFindOptions;
}

MySQL

body: {
    db?: string;
    collection: string;
    columns?: string[];
    where?: MySQLWhere;
    skip?: number;
    limit?: number;
    extra?: string;
}

Returns: Array<object>

eleganceServerClient.controllers.query<Array<T>>

Executes MySQL or aggregate query.

Parameters:

Kai

body: {
    db?: string;
    collection: string;
    query: object[];
    options?: MongoAggregateOptions;
}

MySQL

body: {
  query: string;
}

Returns: Array<T>

eleganceServerClient.controllers.createEmbedding

Creates embedding.

Parameters:

body: {
  input: string | string[] | object | object[];
}

Returns: Embedding

eleganceServerClient.controllers.createFileEmbeddings

Accepts a CSV or PDF file, splits it into chunks and creates embeddings.

Parameters:

body: {
  dataURL: string;
  textField?: string;
  embeddingField?: string;
  chunkSize?: number;
}

Returns: Array<{ text: string; embedding: Embedding; }>

eleganceServerClient.controllers.createAndInsertFileEmbeddings

Accepts a CSV or PDF file, splits it into chunks, creates embeddings, and inserts them into a database.

Parameters:

body: {
    db?: string;
    collection: string;
    dataURL: string;
    textField?: string;
    embeddingField?: string;
    chunkSize?: number;
}

Returns: Array<{ text: string; embedding: Embedding; }>

eleganceServerClient.controllers.createChatCompletion

Accepts a prompt and creates chat completion.

Parameters:

body: {
  model?: ChatCompletionCreateParamsNonStreaming["model"];
  prompt?: string;
  systemRole?: string;
  messages?: ChatCompletionMessage[];
  temperature?: number;
  maxTokens?: number;
}

Returns:

string | null;

eleganceServerClient.controllers.searchChatCompletion

Accepts a prompt, performs dot product search, and creates chat completion for the found records.

Parameters:

body: {
    db?: string;
    collection: string;
    model?: ChatCompletionCreateParamsNonStreaming["model"];
    prompt: string;
    systemRole?: string;
    messages?: ChatCompletionMessage[];
    temperature?: number;
    maxTokens?: number;
    textField?: string;
    embeddingField?: string;
    minSimilarity?: number;
    maxContextLength?: number;
}

Returns:

{
  content: string;
  context: string;
}

eleganceServerClient.controllers.dotProductSearch

Performs dot product search in the collection based on the query.

Parameters:

body: {
    db?: string;
    collection: string;
    query: string;
    queryEmbedding?: Embedding;
    embeddingField: string;
    limit?: number;
    minSimilarity?: number;
    includeEmbedding?: boolean;
}

Returns: Array<any>

eleganceServerClient.ai.createEmbedding

Creates embedding

Parameters:

  • input: string | Array<string> | object | Array<object>

Returns: Embedding

eleganceServerClient.ai.embeddingToBuffer

Converts an embedding into a buffer that is then inserted into the database (for Kai).

Parameters:

  • embedding: Embedding

Returns: Buffer

eleganceServerClient.ai.textToEmbeddings

Converts text into embeddings by splitting the text into chunks.

Parameters:

  • text: string
  •   options?: {
        chunkSize?: number;
        textField?: string;
        embeddingField?: string;
    }

eleganceServerClient.ai.dataURLtoEmbeddings

Converts a DataURL (csv, pdf) into an embedding by splitting the text into chunks.

Parameters:

  • dataURL: string
  •   options?: {
        chunkSize?: number;
        textField?: string;
        embeddingField?: string;
    }

Returns: Array<{ text: string; embedding: Embedding }>

eleganceServerClient.ai.createChatCompletion

Creates a chat completion.

Parameters:

  •   params: {
        model?: ChatCompletionCreateParamsNonStreaming["model"];
        prompt?: string;
        systemRole?: string;
        messages?: ChatCompletionMessage[];
        temperature?: number;
        maxTokens?: number;
    }

Returns: string

createEleganceClient

Creates an EleganceClient instance for a client.

Parameters:

  • connectionType: "kai" | "mysql"
  • config: {
      baseURL: string;
    }

Returns: eleganceServerClient

eleganceClient

Client that includes requests and hooks. It is used to make requests to the server.

eleganceClient.requests

Clean functions to make requests to the server. They can be used anywhere within a project and are handled like typical async functions. The parameters for each request correspond to the controller parameters by name. To familiarize with the parameters, refer to the eleganceServerClient.controllers.<requestName> section.

eleganceClient.hooks

Ready-to-use React.js hooks with state handlers that use requests. Each hook has the following type:

<R = any,>(options?: {
  initialValue?: R;
  initialIsLoading?: boolean;
  onError?: (error: DefaultError) => void;
}) => {
  value: R | undefined;
  error: DefaultError | undefined;
  isLoading: boolean;
  setValue: react.Dispatch<react.SetStateAction<Awaited<R> | undefined>>;
  setError: react.Dispatch<react.SetStateAction<DefaultError | undefined>>;
  setIsLoading: react.Dispatch<react.SetStateAction<boolean>>;
  execute: (body: object) => Promise<Awaited<R> | undefined>;
};

License

Apache 2.0

© 2023 SingleStore

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published