@layer-ai/sdk

TypeScript/JavaScript SDK for Layer AI - Manage LLM requests with intelligent routing and fallbacks.

Installation

npm install @layer-ai/sdk
# or
pnpm add @layer-ai/sdk
# or
yarn add @layer-ai/sdk

Quick Start

import { Layer } from '@layer-ai/sdk';

const layer = new Layer({
  apiKey: process.env.LAYER_API_KEY,
  baseUrl: 'http://localhost:3001'
});

// Complete a request through a gate
const response = await layer.complete({
  gate: 'my-gate',
  prompt: 'Explain quantum computing in simple terms'
});

console.log(response.text);

Configuration

Constructor Options

const layer = new Layer({
  apiKey: string;        // Required: Your Layer API key
  baseUrl?: string;      // Optional: API base URL (default: http://localhost:3001)
  adminMode?: boolean;   // Optional: Enable mutation operations (default: false)
});

Admin Mode

By default, the SDK is read-only to prevent accidental mutations in production. To enable create/update/delete operations, set adminMode: true:

const layer = new Layer({
  apiKey: process.env.LAYER_API_KEY,
  adminMode: true  // Required for mutations
});

// Now you can create/update/delete gates and keys
await layer.gates.create({
  name: 'my-gate',
  model: 'gpt-4o',
  temperature: 0.7
});

Note: Admin mode is intended for setup scripts and infrastructure-as-code. For ongoing management, use the CLI or config files.

API Reference

Completions

layer.complete(params)

Send a completion request through a gate.

const response = await layer.complete({
  gate: 'my-gate',
  prompt: 'What is the capital of France?',

  // Optional overrides (if gate allows)
  model?: 'gpt-4o',
  temperature?: 0.8,
  maxTokens?: 500,
  topP?: 0.9
});

// Response shape
{
  text: string;
  model: string;
  usage: {
    promptTokens: number;
    completionTokens: number;
    totalTokens: number;
  };
  cost: number;
}

Gates

layer.gates.list()

List all gates.

const gates = await layer.gates.list();
// Returns: Gate[]

layer.gates.get(name)

Get a specific gate by name.

const gate = await layer.gates.get('my-gate');

layer.gates.create(data) ⚠️ Requires Admin Mode

Create a new gate. For detailed information about gate configuration fields, see the Configuration Guide.

await layer.gates.create({
  name: 'my-gate',
  model: 'gpt-4o',
  description: 'My custom gate',
  systemPrompt: 'You are a helpful assistant',
  temperature: 0.7,
  maxTokens: 1000,
  topP: 1.0,
  allowOverrides: true,  // or specific fields: ['temperature', 'maxTokens']
  routingStrategy: 'fallback',
  fallbackModels: ['claude-sonnet-4', 'gemini-2.0-flash-exp'],
  tags: ['production', 'general']
});

layer.gates.update(name, data) ⚠️ Requires Admin Mode

Update an existing gate.

await layer.gates.update('my-gate', {
  temperature: 0.8,
  maxTokens: 1500
});

layer.gates.delete(name) ⚠️ Requires Admin Mode

Delete a gate.

await layer.gates.delete('my-gate');

layer.gates.suggestions(gateName)

Get AI-powered model recommendations for a gate based on its task description.

const suggestions = await layer.gates.suggestions('my-gate');

console.log(suggestions.primary);       // "gpt-4o"
console.log(suggestions.alternatives);  // ["claude-sonnet-4", "gemini-2.5-flash"]
console.log(suggestions.reasoning);     // "Based on your task description..."

Smart routing analyzes your gate's description and user preferences (cost, latency, quality weights) to recommend the best models from the registry.

API Keys

layer.keys.list()

List all API keys.

const keys = await layer.keys.list();

layer.keys.create(name) ⚠️ Requires Admin Mode

Create a new API key.

const key = await layer.keys.create('my-app-key');
// Returns: { id: string, name: string, key: string, createdAt: string }
// Save the key - it's only shown once!

layer.keys.delete(id) ⚠️ Requires Admin Mode

Revoke an API key.

await layer.keys.delete('key-id');

Logs

layer.logs.list(options?)

List request logs.

const logs = await layer.logs.list({
  limit: 100,
  offset: 0,
  gateName: 'my-gate'  // Optional filter
});

TypeScript Support

The SDK is written in TypeScript and provides full type definitions:

import { Layer, Gate, CompletionRequest, CompletionResponse } from '@layer-ai/sdk';

const layer = new Layer({ apiKey: 'key' });

// All methods are fully typed
const response: CompletionResponse = await layer.complete({
  gate: 'my-gate',
  prompt: 'Hello'
});

Error Handling

try {
  const response = await layer.complete({
    gate: 'my-gate',
    prompt: 'Hello'
  });
} catch (error) {
  if (error instanceof Error) {
    console.error('Layer error:', error.message);
  }
}

Examples

For complete working examples, check out the layer-ai-examples repository:

• Content Generator - Next.js app with smart routing for content generation

Basic Completion

import { Layer } from '@layer-ai/sdk';

const layer = new Layer({
  apiKey: process.env.LAYER_API_KEY
});

const response = await layer.complete({
  gate: 'default',
  prompt: 'Explain machine learning'
});

console.log(response.text);

With Parameter Overrides

const response = await layer.complete({
  gate: 'my-gate',
  prompt: 'Write a haiku about coding',
  temperature: 0.9,  // More creative
  maxTokens: 100
});

Infrastructure as Code

import { Layer } from '@layer-ai/sdk';

const layer = new Layer({
  apiKey: process.env.LAYER_API_KEY,
  adminMode: true
});

// Create gates programmatically
await layer.gates.create({
  name: 'production-gate',
  model: 'gpt-4o',
  temperature: 0.7,
  routingStrategy: 'fallback',
  fallbackModels: ['claude-sonnet-4']
});

await layer.gates.create({
  name: 'dev-gate',
  model: 'gpt-4o-mini',
  temperature: 0.5
});

License

MIT