The library focuses on a structured, lightweight, modular design that lets developers easily assemble complex workflows from reusable components:

Prompts: Structured templating for managing sophisticated prompts with built-in Handlebars support.

Parsers: Components that transform raw LLM responses (strings) into structured data, supporting JSON, arrays, enum extraction, and custom parsing logic.

LLM Providers: Abstracted interfaces for various providers, including OpenAI, Anthropic, xAI, Google Gemini, AWS Bedrock, and Ollama, allowing seamless switching without changing implementation logic.

Executors: The LLM executor takes an llm, a prompt, optionally a parser, and wraps in a well-typed function. An LLM executor is a container that can be used to call an LLM with a pre-defined input and output; with additional values provided at the time of execution. An LLM executor's input and output types are determined by the prompt and parser respectively.

Utilities: Helpers for common tasks such as prompt debugging, caching, and managing conversation state.

Here's how you might create a structured executor:

--- import { createLlmExecutor, createChatPrompt, createParser, useLlm } from 'llm-exe';

const llm = useLlm("openai.gpt-4o-mini");

const prompt = createChatPrompt<{input: string}>('Translate the phrase "{{input}}" into French, Spanish, and German, returning the results as an unordered markdown list.');

const parser = createParser('listToArray');

const translateExecutor = createLlmExecutor({ llm, prompt, parser });

// result is typed as string[]

// the input is also well-typed!

const result = await translateExecutor.execute({ input: "Hello, world!" });

console.log(result);

// Outputs: ["Bonjour le monde!", "¡Hola, mundo!", "Hallo, Welt!"] ---

Check out the documentation and more examples here: llm-exe.com. I'd love your feedback or contributions!