@motioneffector/llm

A TypeScript client for LLM APIs with OpenRouter support, streaming responses, conversation management, and automatic retries.

Features

• OpenRouter Integration - Access 200+ models through a unified API
• Streaming Support - Real-time response streaming with async iterators
• Conversation Management - Stateful conversations with automatic history tracking
• Automatic Retries - Smart retry logic with exponential backoff
• Token Estimation - Estimate prompt tokens before sending requests
• Type Safety - Full TypeScript definitions with no any types
• Model Information - Built-in pricing and context length data
• Abort Support - Cancel requests using AbortController signals

Read the full manual →

Quick Start

import { createLLMClient } from '@motioneffector/llm'

// Create a client
const client = createLLMClient({
  apiKey: process.env.OPENROUTER_KEY,
  model: 'anthropic/claude-sonnet-4'
})

// Send a chat completion request
const response = await client.chat([
  { role: 'user', content: 'Explain quantum computing in simple terms' }
])

console.log(response.content)
console.log(`Used ${response.usage.totalTokens} tokens in ${response.latency}ms`)

Testing & Validation

• Comprehensive test suite - 265 unit tests covering core functionality
• Fuzz tested - Randomized input testing to catch edge cases
• Strict TypeScript - Full type coverage with no any types
• Zero dependencies - No supply chain risk

License

MIT © motioneffector