@jaypie/llm

Large language model abstraction with multi-provider support.

Overview

@jaypie/llm provides a unified interface for working with large language models, including:

• Multi-provider support (OpenAI, Anthropic)
• Tool calling and function execution
• Structured output generation
• Conversation history management
• Multi-turn reasoning

Installation

npm install @jaypie/llm

Key Features

LLM Interface

Unified interface for multiple LLM providers:

import { Llm } from "@jaypie/llm";

const llm = new Llm("gpt-4o");
const response = await llm.send("What is the capital of France?");

Tool Calling

Execute functions during LLM conversations:

import { Llm, toolkit } from "@jaypie/llm";

const llm = new Llm("gpt-4o");
const response = await llm.operate("What's the weather in Paris?", {
  tools: [toolkit.weather],
  turns: 12,
});

Structured Outputs

Generate structured data with natural schema syntax:

const response = await llm.send("Extract user information", {
  format: { name: String, age: Number, email: String },
});

Built-in Tools

Pre-configured tools for common tasks:

• time - Get current time
• weather - Get weather information
• random - Generate random numbers
• roll - Roll dice

API Documentation

API documentation will be generated from TypeScript definitions.

Related Packages

• @jaypie/core - Core utilities
• @jaypie/aws - AWS integrations
• @jaypie/testkit - Testing utilities