Minds - Build AI Powered Workflows!

A JS library (Typescript) that makes it easy to build your workflows and app backends with large language models (LLMs) like OpenAI, Azure OpenAI, Cohere and AlephAlpha.

This library handles all the complex prompt engineering so you can focus on building amazing things like context power chat, question answering, natural language search in minutes. Define Javascript functions that AIs can use. For example the AI can lookup your database, call an API or search the web while answering a business question.

We totally believe that AI will soon replace your entire app backend. We truly live in amazing times. Please join our Discord so we can help you build your idea.

npm i @dosco/minds

Why use Minds

• Sensible Defaults
• Simpler and smaller than the alternatives
• Minimal to zero dependencies
• Focus on useful real-world workflows
• Single interface: OpenAI, Azure OpenAI, Cohere and more

Example Workflows

| Example | Description | | --------------------- | -------------------------------------------------------------- | | ask-questions.js | AI uses Google search to find the correct answer | | product-search.js | Customers can as product related questions in natural language | | customer-support.js | Extract valuable details from customer communications | | marketing.js | Use AI to generate short effective marketing sms messages | | transcribe-podcast.js | Transcribe multiple podcast channels into text | | chat-assistant.js | AI chat bot capable of intellegent conversations | | get-summary.js | AI to generate a short summary of a large block of text | | ai-vs-ai.js | OpenAI has a friendly chat with Cohere |

Answer Customer Product Questions

Build business or personal workflows where the AI calls your APIs to fetch data and use that data to solve problems or help customers find answers.

const productDB = [
  { name: 'Macbook Pro', description: 'M2, 32GB', in_stock: 4321 },
  { name: 'Macbook Pro', description: 'M2, 96GB', in_stock: 2 },
];

const productSearch = (text) => {
  return JSON.stringify(productDB.filter((v) => text.includes(v.name)));
};

const actions = [
  {
    name: 'Product Search',
    description: 'Used to search up a products information by its name',
    action: productSearch,
  },
];

const prompt = new QuestionAnswerPrompt(actions);

// Customer support email
const customerQuery = `Do you guys have 5 Macbook Pro's M2 with 96GB RAM and 3 iPads in stock?`;

await prompt.generate(ai, query);

>  Answer for customer: We have 2 Macbook Pro's M2 with 96GB RAM and 0 iPads in stock.

Extract Details From Business Messages

Inbound messages and emails from customers are a core part of all business workflows. AI makes it easy to extract valuable details and classify messages so you can build workflows that make your processes more efficient.

const entities = [
  { name: BusinessInfo.ProductName },
  { name: BusinessInfo.IssueDescription },
  { name: BusinessInfo.IssueSummary },
  { name: BusinessInfo.PaymentMethod, classes: ['Cash', 'Credit Card'] },
];

const prompt = new ExtractInfoPrompt(entities);

const customerMessage = `
I am writing to report an issue with my recent order #12345. I received the package yesterday, but 
unfortunately, the product that I paid for with cash (XYZ Smartwatch) is not functioning properly. 
When I tried to turn it on, the screen remained blank, and I couldn't get it to respond to any of 
the buttons.

Jane Doe`;

const res = await prompt.generate(ai, customerMessage);

Extracted Details From Customer Message:
{
  'Product Name' => [ 'XYZ Smartwatch' ],
  'Issue Description' => [ 'Screen remained blank and unable to respond to any buttons' ],
  'Issue Summary' => [ 'Product is not functioning properly' ],
  'Payment method' => [ 'Cash' ]
}

Promotional Messaging

Use AI to generate effective promotional content for emails, sms or other channels.

const product = {
  name: 'Acme Toilet Cleaning',
  description: '24/7 Commercial and residential restroom cleaning services',
};

const to = {
  name: 'Jerry Doe',
  title: 'Head of facilities and operations',
  company: 'Blue Yonder Inc.',
};

const prompt = new MessagePrompt({ type: MessageType.Text }, product, to);

const context = `
1. Under 160 characters
2. Prompts recipients to book an call
3. Employs emojis and friendly language
`;

const res = await prompt.generate(ai, context);
console.log(res.value());

Hey Jerry! 🤗 Acme Toilet Cleaning is offering 24/7 commercial and residential restroom cleaning services. Let us take care of your restroom needs so you don't have to. 🚽 Book a call today and let's get started! 📞

AI Smart Assistant

Build an AI powered assistant that maintains context as you converse with it asking if various questions.

import {
  Cohere,
  AlephAlpha,
  OpenAI,
  Memory,
  AssistantPrompt,
} from '@dosco/minds';

// const ai = new Cohere(process.env.COHERE_APIKEY)
// const ai = new AlephAlpha(process.env.AALPHA_APIKEY)

const ai = new OpenAI(process.env.OPENAI_APIKEY);

const prompt = new AssistantPrompt();

await prompt.generate(ai, `How far is the sun from the moon?`);
await prompt.generate(ai, `And from mars?`);
await prompt.generate(ai, `Will it ever end?`);

❯ node chat-assistant.js
AI: How far is the sun from the moon?
> The sun is about 384,400 kilometers away from the moon.

AI: And from mars?
> The sun is about 384,400 kilometers away from Mars as well.

AI: will it ever end?
> The sun will eventually end, but not for billions of years.

cd examples
npm i
node chat-assistant.js

Why use Minds?

Large language models (LLMs) are getting really powerful and have reached a point where they can work as the backend for your entire product. However since its all cutting edge technology you have to manage a lot of complexity from using the right prompts, models, etc. Our goal is to package all this complexity into a well maintained easy to use library that can work with all the LLMs out there.

How to use this library?

1. Pick an AI to work with

// Use Cohere AI
const ai = new Cohere(process.env.COHERE_APIKEY);

// Use OpenAI
const ai = new OpenAI(process.env.OPENAI_APIKEY);

2. Pick a memory for storing context (optional)

// Can be sub classed to build you own memory backends
// like one that uses redis or a database
const mem = new Memory();

3. Pick a prompt based on your usecase

// For building conversational chat based AI workflows
const prompt = new AssistantPrompt();

// For building prompts that return structed data and can
// can search the web, or lookup your database to
// answer questions
const prompt = new ZPrompt(zodSchema, actions);

// For building question answer workflows where the
// AI can search the web, or lookup your database to
// answer questions
const prompt = new QuestionAnswerPrompt(actions);

4. Engage the AI

// Query the AI
const res = await prompt.generate(
  ai,
  `Do we have the product the email referes to in stock? 
  Email: I'm looking for a Macbook Pro M2 With 96GB RAM.`
);

// Get the response
console.log('>', res.value());

{
  "data": [
    {
      "name": "Macbook Pro",
      "units": 2,
      "desc": "M2, 96GB RAM"
    }
  ],
  "response": "We have 2 Macbook Pro M2 96GB RAM in stock."
}

What are actions?

Actions are functions that you define and the API can then call these functions to fetch data it needs to come up with the final answer. All you have to do is pass a list of actions to a prompt that supports actions like the QuestionAnswerPrompt.

Embeddings are also supported. If you use a function with two arguments then the embeddings for the text (first argument) will be passedi in the second argument. Embeddings can be used with a vector search engine to find similiar things. For example to fetch similiar text paragraphs when coming up with a correct answer.

// Action without embeddings
const productSearch(text) {
  return result
}

// Action with embeddings
const productSearch(text, embeds) {
  return result
}

You must specify a name and a clear description matching what the action does. The action itself is a function on the action object that you define.

const actions = [
  {
    name: 'Product Search',
    description: 'Used to search up a products information by its name',
    action: productSearch,
  },
  {
    name: 'Customer Issues Search',
    description: 'Used to search the customer support database for solutions',
    action: wolframAlphaSearch,
  },
];

Finally pass the action list to the prompt.

const prompt = new QuestionAnswerPrompt(actions);

Or use the ZPrompt to get the response as a JS object

const CustomerResponse = z.object({
  data: z
    .array(
      z.object({
        name: z.string().describe('product name'),
        units: z.number().describe('units in stock'),
        desc: z.string().max(15).describe('product description'),
      })
    )
    .describe('inventory information'),
  response: z.string().max(50).describe('customer response'),
});

const prompt = new ZPrompt(CustomerResponse, actions);

Reach out

We're happy to help reach out if you have questions or join the Discord

twitter/dosco

Prompt Engineering

There is a bit of magic to getting an LLM (AI) to do your bidding. For fans of boarding school going wizards it's sort of like learning spells that trigger patterns deep inside the latent space of the model. This is the new and exciting field of Prompt Engineering and this library is how we turn these spells into code to help make building with AI a more democratic endevour. Join us we're just getting started.

Featured AI Art

MidJourney Art by AmitDeshmukh

Prompt: kids walking on a dirt road through a futuristic village, field of daisies on both sides, futuristic clothing, some traditional homes, cumulus clouds, bright mid morning sunlight, birds and drone