GPT Mind Core

The Mind library is a JavaScript library for building conversational agents that can generate responses based on a set of pre-defined rules. It allows you to define a state map, bindings between agents, and a set of agents that can generate responses based on input statements.

Installation

You can install the Mind library using npm:

npm install @gpt-mind/core

Usage

To use the Mind library, you need to create an instance of the Mind class and pass it a configuration object. The configuration object should have the following properties:

• stateMap: An object representing the current state of the conversation.
• bindings: An object mapping agents to other agents that they are bound to.
• agents: An object containing the definitions of the agents.
• start: An object representing the initial state of the conversation.
• sources: An object mapping agents to functions that generate responses based on the current state of the conversation.

Here is an example of how to use the Mind library:

import { Mind } from 'mind-library';
imoport OpenAI from 'openai-api';

const openAI = new OpenAI('sk-...');

function complete(query, options) {
  return openAI.complete({
    engine: 'davinci',
    prompt: query,
    maxTokens: 100,
    temperature: 0.9,
    topP: 1,
    presencePenalty: 0,
    frequencyPenalty: 0,
    bestOf: 1,
    n: 1,
    stream: false,
    stop: ['\n', '###'],
    ...options
  });
}

const config = {
  stateMap: {},
  bindings: {
    "input": ["agent1"],
    "agent1": ['agent2'],
    "agent2": ['agent3'],
    "agent3": ['output', 'agent4'],
    "agent4": ["internal"],
    "internal": ['canVisualize', 'think'],
    "canVisualize": ["visualize"],
    "visualize": ["internal"],
    "output": []
  },
  agents: {
    "agent1": "generate a description of the statement \"{{input}}\"",
    "agent2": "generate a response to the statement \"{{agent1}}\"",
    "agent3": "generate a response given statements \"{{agent1}}\"\nand statements \"{{agent2}}\"",
    "agent4": "generate a response free of social mind to the statement \"{{agent1,agent2}}\"",
    "canVisualize": "is there visualizable imagery in the statement \"{{agent1,agent2}}\"",
    "visualize": "generate a detailed description of the imagery in the statement \"{{agent1,agent2}}\"",
  },
  start: {},
  sources: {
    "agent1": (agent, query, options) => complete(query, options),
    "agent2": (agent, query, options) => complete(query, options),
    "agent3": (agent, query, options) => complete(query, options),
    "agent4": (agent, query, options) => complete(query, options),
    "canVisualize": (agent, query, options) => complete(query, options),
    "visualize": (agent, query, options) => Automatic111.txt2img(query, "", { steps: 20 }),
    "think": (agent, query, options) => complete(query, options),
  }
};

const mind = new Mind(config);

mind.on('output', (event, output) => {
  console.log(output);
});

mind.input('agent1', 'Hello');

In this example, we create an instance of the Mind class with a configuration object that defines a set of agents that can generate responses based on input statements. We then register an event listener for the output event, which is triggered when an agent generates a response. Finally, we call the input method of the mind instance to start the conversation with the agent1 agent.

API

Mind(config: Object)

The Mind class constructor takes a configuration object as an argument. The configuration object should have the following properties:

stateMap: An object representing the current state of the conversation. bindings: An object mapping agents to other agents that they are bound to. agents: An object containing the definitions of the agents. start: An object representing the initial state of the conversation. sources: An object mapping agents to functions that generate responses based on the current state of the conversation.

on(event: string, callback: (event: string, output: string) => void)

Registers an event listener for the specified event. The callback function is called when the event is triggered, and is passed two arguments: the name of the event, and the output generated by the agent that triggered the event.

once(event: string, callback: (event: string, output: string) => void)

Registers an event listener for the specified event that will only be called once. The callback function is called when the event is triggered, and is passed two arguments: the name of the event, and the output generated by the agent that triggered the event.

off(event: string, callback: (event: string, output: string) => void)

Removes the specified event listener.

emit(event: string, output: any)

Triggers the specified event with the given output.

input(agent: string, query: string)

Sets the input for the specified agent and triggers the conversation.

run(agent: string)

Runs the specified agent and its bound agents.

complete(agent: any, query: string, options: any)

Generates a response for the specified agent with the given query and options.

Contributing

Contributions are welcome!

Please read the [contributing guidelines](