llm-chain

A TypeScript library for easily interacting with various Large Language Model providers through a unified interface.

Installation

npm install llm-chain

Features

• Unified interface for multiple LLM providers:
  • OpenAI
  • Groq
  • Gemini (Google)
  • Anthropic
• Type-safe API with TypeScript
• Support for streaming responses
• Built-in validation using Zod
• Easy to extend for new providers
• Automatic model validation and token limits
• Provider-specific optimizations
• Support for multiple deployment options (Direct API, AWS Bedrock, Google Vertex)

Usage

OpenAI

import { LLMClient } from "llm-chain";

// Create a client with OpenAI
const client = LLMClient.createOpenAI("your-openai-api-key");

// Simple completion
const response = await client.complete("What is the capital of France?");
console.log(response);

// Chat completion with more options
const chatResponse = await client.chatCompletion({
  model: "gpt-4o-mini",
  messages: [
    { role: "system", content: "You are a helpful assistant." },
    { role: "user", content: "What is the capital of France?" },
  ],
  temperature: 0.7,
});

console.log(chatResponse.message.content);

Groq

import { LLMClient } from "llm-chain";

// Create a client with Groq
const client = LLMClient.createGroq("your-groq-api-key");

// Chat completion with Mixtral
const chatResponse = await client.chatCompletion({
  model: "mixtral-8x7b-32768",
  messages: [
    { role: "system", content: "You are a helpful assistant." },
    { role: "user", content: "What is the capital of France?" },
  ],
  temperature: 0.7,
});

console.log(chatResponse.message.content);

Gemini

import { LLMClient } from "llm-chain";

// Create a client with Gemini
const client = LLMClient.createGemini("your-gemini-api-key");

// Chat completion with Gemini Pro
const chatResponse = await client.chatCompletion({
  model: "gemini-2.0-flash",
  messages: [{ role: "user", content: "What is the capital of France?" }],
  temperature: 0.7,
});

console.log(chatResponse.message.content);

Anthropic

import { LLMClient } from "llm-chain";

// Create a client with Anthropic
const client = LLMClient.createAnthropic("your-anthropic-api-key");

// Chat completion with Claude
const chatResponse = await client.chatCompletion({
  model: "claude-3.5-sonnet-latest",
  messages: [{ role: "user", content: "What is the capital of France?" }],
  temperature: 0.7,
});

console.log(chatResponse.message.content);

Streaming Responses

await client.streamChatCompletion(
  {
    model: "mixtral-8x7b-32768", // or any other supported model
    messages: [{ role: "user", content: "Write a story about a cat." }],
  },
  chunk => {
    process.stdout.write(chunk);
  }
);

Supported Models

OpenAI Models

• Default model: gpt-4o-mini
• Supports all OpenAI chat models

Groq Models

Production Models

• mixtral-8x7b-32768 (Mistral, 32k context)
• llama-3.3-70b-versatile (Meta, 128k context)
• llama-3.1-8b-instant (Meta, 128k context)
• Various Whisper models for audio processing

Preview Models

• llama3-groq-70b-8192-tool-use-preview (Groq)
• llama3-groq-8b-8192-tool-use-preview (Groq)
• Various LLaMA-3 preview models

Gemini Models

Gemini 2.0

• gemini-2.0-flash-exp (Experimental, 1M input tokens)

Gemini 1.5

• gemini-1.5-flash (32k context, versatile)
• gemini-1.5-flash-8b (32k context, high volume)
• gemini-1.5-pro (32k context, complex reasoning)

Utility Models

• text-embedding-004 (Embeddings)
• aqa (Question answering)

Anthropic Models

Claude 3 Models

• claude-3-5-sonnet-20241022 (Latest Sonnet)
• claude-3-5-haiku-20241022 (Latest Haiku)
• claude-3-opus-20240229 (Most powerful)
• claude-3-sonnet-20240229 (Balanced)
• claude-3-haiku-20240307 (Fast)
• claude-2.1 (Legacy)

Advanced Features

Custom Provider

import { LLMProvider, LLMClient } from "llm-chain";

class CustomProvider implements LLMProvider {
  // Implement the LLMProvider interface
  // ...
}

const client = new LLMClient(new CustomProvider(), "your-model-name");

Anthropic Deployment Options

// Direct API
const client = LLMClient.createAnthropic("your-api-key");

// AWS Bedrock
const client = LLMClient.createAnthropic("your-api-key", "bedrock", {
  aws_access_key: "your-access-key",
  aws_secret_key: "your-secret-key",
  aws_region: "us-east-1",
});

// Google Vertex
const client = LLMClient.createAnthropic("your-api-key", "vertex", {
  project_id: "your-project-id",
  region: "us-central1",
});

Error Handling

The library provides built-in error handling and validation:

• Model availability checks
• Token limit validation
• Context window size validation
• Streaming capability checks
• Provider-specific error handling

License

MIT