Configuration

Constructor options


const lg = new LaunchGate(config: LaunchGateConfig);

`LaunchGateConfig`

Property	Type	Default	Description
`apiKey`	`string`	required	Your LaunchGate API key (starts with `lg_live_` or `lg_test_`)
`baseUrl`	`string`	`"https://api.launchgate.ai"`	API base URL
`timeout`	`number`	`30000`	Request timeout in milliseconds

Examples

Minimal


const lg = new LaunchGate({
  apiKey: process.env.LAUNCHGATE_API_KEY!,
});

Full configuration


const lg = new LaunchGate({
  apiKey: process.env.LAUNCHGATE_API_KEY!,
  baseUrl: "https://api.launchgate.ai",
  timeout: 60_000, // 60 seconds for suites with many LLM judge cases
});

Local development


const lg = new LaunchGate({
  apiKey: "lg_test_local_dev_key",
  baseUrl: "http://localhost:4001",
});

Retry behaviour

The SDK automatically retries failed requests with exponential backoff:

Max retries: 3
Backoff: Exponential (1s, 2s, 4s)
Retried: Network errors, 5xx server errors, 429 rate limit errors
Not retried: 401 auth errors, 402 payment errors, 400 validation errors

Graceful degradation

If the LaunchGate API is unreachable after all retries, the SDK returns a result with status: "skipped" instead of throwing. This prevents eval failures from blocking your application:


const result = await lg.run("my-suite", { output: "..." });
 
if (result.status === "skipped") {
  // LaunchGate was unreachable — proceed with caution
  console.warn("Eval skipped — LaunchGate API unavailable");
}