salute

Introducing Salute: A Simplified Approach to Controlling Large Language Models

Salute is a JavaScript library designed to streamline interactions with Large Language Models (LLMs), such as those developed by OpenAI. Inspired by the simplicity of React and the powerful capabilities of Microsoft Guidance, Salute offers developers a way to manage complex LLM processes with ease and clarity.

Key Features

Salute stands out with several distinct features that make it an attractive choice for developers:

• React-like Composability: Salute employs a structure and syntax that will be familiar to anyone who has worked with React. This makes it easy to learn and integrate into existing workflows.
• Minimal Abstractions: The library is designed with simplicity in mind. It introduces only a few necessary abstractions to keep the codebase minimal and easily maintainable.
• Transparency: With Salute, what you see is what you get. There are no hidden processes or prompts—the operations are clear and visible.
• Low-Level Control: Developers have detailed control over how the LLM processes text, allowing for tailored and precise interactions.
• Familiar JavaScript Features: Salute leverages features like type-checking, linting, syntax highlighting, and auto-completion to provide a fast learning curve for those familiar with JavaScript.

How to Get Started

To get started with Salute, you can install it using your preferred package manager:

npm install salutejs

yarn add salutejs

pnpm add salutejs

Additionally, you'll need to set your OpenAI API key as an environment variable:

process.env.OPENAI_KEY = 'your-api-key'

Key Functionalities

Here's a glimpse into some of the core functionalities that Salute offers:

Simple Chat Completion

Salute allows you to define a sequence of messages using roles—system, user, and assistant. The gen function is pivotal in sending prompts to the LLM and capturing its responses.

import { gpt3, gen, assistant, system, user } from "salutejs";

// Define a new agent
const agent = gpt3(
  ({ params }) => [
    system`You are a helpful and terse assistant.`,
    user`I want a response to the following question: ${params.query}.`,
    assistant`${gen("answer")}`,
  ]
);

// Capture the result
const result = await agent({ query: `How can I be more productive?` });
console.log(result.answer);

Creating Chat Sequences

You can build complex interaction sequences by nesting components and sequencing multiple steps. Salute supports dynamic chat sequences that can produce comprehensive outputs from several interactions.

Array.map for Chat Sequences

Salute integrates seamlessly with JavaScript arrays, allowing you to map functions over an array to generate sequences dynamically. This feature makes it easy to iterate over prompts and manage multiple outputs efficiently.

Advanced Use

Salute also supports more sophisticated use cases, such as controlling the prompt context, handling multiple completions efficiently, and even facilitating conversations between virtual agents.

Configurability

While optimized for OpenAI models, Salute’s architecture is adaptable, designed to accommodate different AI models in the future. Developers can create custom configurations tailored to specific model requirements, offering flexibility in deployment.

Conclusion

Salute simplifies the process of interacting with LLMs by bringing a familiar React-like experience to the world of AI model control. It ensures that handling complex language models is as intuitive and straightforward as writing any other piece of JavaScript code, making it an excellent tool for developers looking to innovate with AI.