@darkdeploy/node-sdk

The official Node.js SDK for DarkDeploy — Enterprise Feature Flagging, Live A/B Experimentation, and Dynamic AI Prompt Configurations with Real-Time Server-Sent Events (SSE) Streaming.

Features

• ⚡ Zero-Latency In-Memory Evaluation: Evaluates flags and experiments locally in microseconds using @darkdeploy/core.
• 🔄 Real-Time Push Updates via SSE: Flag toggles, rollout changes, and prompt edits stream instantly without polling.
• 🛡️ Fault-Tolerant & Offline-First: Automatic reconnects with exponential backoff & jitter. Graceful fallbacks ensure your app never crashes.
• 🧪 Full A/B Experimentation: Automatic exposure logging and conversion metric tracking (client.track).
• 🤖 Dynamic AI Hyperparameters & Prompts: Update LLM foundation models, temperature, token limits, and system prompts on the fly without redeployment.

Installation

npm install @darkdeploy/node-sdk

Quickstart

import { DarkDeployClient } from "@darkdeploy/node-sdk";

const client = new DarkDeployClient({
  apiKey: "dd_dev_d934289e58292e134bcde54fbf153aa7",
  baseUrl: "https://darkdeploy-api.onrender.com", // Or your local server: http://localhost:4000
});

await client.initialize();
console.log("Connected to DarkDeploy Engine via SSE");

Usage Guide

1. Feature Flag Evaluation

Evaluate whether a feature is active for a specific user based on percentage rollouts or custom targeting rules:

const isEnabled = client.isEnabled("new-checkout-flow", {
  id: "usr_9812",
  attributes: {
    plan: "enterprise",
    country: "US",
  },
});

if (isEnabled) {
  // Render new 1-click checkout experience
}

Detailed evaluation with explanation:

const result = client.evaluate("new-checkout-flow", { id: "usr_9812" });
console.log(result.value);  // true or false
console.log(result.reason); // "RULE_MATCH" | "ROLLOUT" | "DEFAULT" | "OFF"

2. A/B Testing & Variant Tracking

Assign users to experiment variants and track conversion goals:

const experimentKey = "checkout-button-experiment";

// Automatically records an EXPOSURE event to the DarkDeploy telemetry pipeline
const variant = client.getVariant(experimentKey, { id: "usr_9812" });
console.log(`Variant assigned: ${variant?.key} (${variant?.name})`);

// When customer completes purchase, record the CONVERSION metric
await client.track(
  experimentKey,
  { id: "usr_9812" },
  "CONVERSION",
  49.99 // Optional revenue value
);

3. AI Configs & Dynamic LLM Prompts

Dynamically retrieve and hot-swap LLM models, hyperparameters, and prompts at runtime:

const aiConfig = client.getAIConfig(
  "customer-support-agent",
  { id: "usr_9812", attributes: { plan: "enterprise" } },
  {
    model: "gpt-4o",
    temperature: 0.7,
    maxTokens: 1024,
    systemPrompt: "You are a professional enterprise support agent.",
  }
);

console.log("Model:", aiConfig.model); // e.g. "claude-3-5-sonnet"
console.log("Temperature:", aiConfig.temperature); // e.g. 0.2
console.log("Prompt:", aiConfig.systemPrompt);

API Reference

| Method | Parameters | Return Type | Description | | :--- | :--- | :--- | :--- | | initialize() | None | Promise<void> | Fetches snapshots from API and establishes SSE stream. | | isEnabled() | flagKey, context, defaultValue? | boolean | Quick boolean flag evaluation. | | evaluate() | flagKey, context, defaultValue? | EvaluationResult | Full evaluation result including reason and matched rule. | | getVariant() | experimentKey, context | Variant \| null | Assigns variant and fires background EXPOSURE event. | | track() | experimentKey, context, type?, value? | Promise<void> | Records exposure or conversion telemetry event. | | getAIConfig() | flagKey, context, fallback? | AIConfig \| null | Evaluates dynamic LLM model configuration and system prompt. | | close() | None | void | Closes SSE connection and flushes local cache. |

Keywords

darkdeploy, feature flags, feature toggles, launchdarkly alternative, split testing, ab testing, experimentation, server sent events, sse streaming, dynamic prompts, llm config, ai flags, hot reload prompts, node sdk, typescript.

License

MIT