🚀 Introducing lollms_client_js: The Ultimate Tool for Interacting with Large Language Models! 🤖

Ever wished you could unleash the power of large language models with just a few lines of code? Look no further! With our new lollms_client_js library, you can easily generate text using various parameters and configurations. Intrigued? Let's dive in!

See it in action:

🚀 Installation

Get started by installing the library with npm:

npm install lollms_client_js

📝 Usage

The library provides functions to interact with the lollms server. On the server side, you need to install and load a certain number of personas, then you can use this library to generate text either using your own conditionning, or by summoning one of the personalities mounted on the server side.

For this moment, two functions are exposed:

generateText : Generates text out of a prompt. If you specify a personality id (between 0 and nb mounted personalities) then you can directly send a user prompt. If you use -1 as personality id, then you need to format your prompt as you need. You can use simple completion such as sending "Once apon a time " and the AI wil ldo the completion, or yo ucan use one of our advised prompt formats. The advised prompt format is : !@>system: here you place your conditioning !@>user: here you put the user prompt !@>ai: You can also use instruct mode: !@>instruct: here you put the instruction !@>process: Lollms uses the !@> to detect role change, lollms also uses the model own eos token to detect end of sentense.
listMountedPersonalities : Lists all mounted personalities. The index of each personality can be used in generateText

generateText

To use the generation without personality, import the generateText function from the library and begin generating text with the desired parameters:

import { generateText } from 'lollms_client_js';

// Prompt
const prompt = "Once upon a time";
const generatedText = await generateText(prompt);
console.log(generatedText);

📚 Parameters

The generateText function supports the following parameters:

prompt (string, required): Initial input text prompt for text generation
host (string, optional, default: "http://localhost:9600"): Host URL of the lollms server
model_name (string, optional): Name of the model to use for text generation
personality (number, optional, default: -1): Personality index for the model
n_predict (number, optional, default: 1024): Maximum number of tokens to generate
stream (boolean, optional, default: true): Stream generated text or not
temperature (number, optional, default: 1.0): Sampling temperature for text generation
top_k (number, optional, default: 50): Top-k sampling parameter for text generation
top_p (number, optional, default: 0.95): Top-p sampling parameter for text generation
repeat_penalty (number, optional, default: 0.8): Repeat penalty for text generation
repeat_last_n (number, optional, default: 40): Number of tokens to consider for repeat penalty
seed (number, optional): Random seed for text generation
n_threads (number, optional, default: 8): Number of threads to use for text generation

listMountedPersonalities

This function retrieves a list of all the personalities currently mounted on the lollms server. It's particularly useful for users who wish to explore the different personas available for text generation. By using this function, you can select a specific personality index to generate text that aligns with a particular tone, style, or domain knowledge.

Usage

To use this function, simply call it from the lollms_client_js library. The function will return an array of personalities, each represented by an object containing at least an ID and a name. This allows you to easily identify and select the appropriate personality for your text generation needs.

import { listMountedPersonalities } from 'lollms_client_js';

async function showPersonalities() {
  const personalities = await listMountedPersonalities();
  console.log(personalities);
}

showPersonalities();

Output

The function returns a promise that resolves to an array of objects. Each object represents a personality, containing the following properties:

id (number): The unique identifier of the personality.
name (string): The name of the personality.

Example output:

[
  { "id": 0, "name": "Creative Writer" },
  { "id": 1, "name": "Tech Enthusiast" },
  { "id": 2, "name": "Science Fiction Guru" }
]

💻 Example

Check out and adapt the example below for a seamless start:

import { generateText } from 'lollms_client_js';

// Prompt
const prompt = "Once upon a time";

// Custom Configuration
const configurations = {
  model_name: 'myModel',
  personality: 1,
  n_predict: 256,
  temperature: 0.85,
  top_k: 20,
  repeat_penalty: 0.7,
};

// Generate Text
const generatedText = await generateText(prompt, configurations);
console.log(generatedText);

🙌 Contributing

We welcome and appreciate your contributions! Share any issues or ideas you have in a new issue.

📄 License

This project is licensed under the Apache-2.0 License.

⚠️ Important Note

Please make sure the lollms server is configured to accept CORS requests from the server serving the lollms_client_js client. Add the host address to the allowed_origins list in the configs/local_configs.yaml file.

For example, if the origin is https://mydomain.com:95620, add it there. This step ensures a smooth lollms operation.

More about me and my work

Hey there! I am ParisNeo, a seasoned research engineer with a deep passion for Artificial Intelligence, Robotics, and Space. I enjoy creating extraordinary AI models, so feel free to join me on my Twitter @ParisNeo_AI, Discord, Sub-Reddit, and Instagram @spacenerduino

ParisNeo/lollms_client_js