feature-flags

Feature Flags & A/B Testing

Tactical patterns for flag evaluation, progressive rollouts, controlled experiments, and flag lifecycle — the code-level companion to infrastructure-level canary deployments.

When to Activate

• Adding feature flag support to a new or existing service
• Designing a percentage-based or ring-based rollout
• Setting up A/B experiments or multivariate tests
• Choosing between LaunchDarkly, Unleash, and OpenFeature
• Implementing sticky bucketing or mutual exclusion across experiments
• Writing unit or integration tests for flag-gated code paths
• Auditing, cleaning up, or governing stale flags

Flag Types & Evaluation Context

Four flag types cover all use cases:

Evaluation Context

The context is the set of attributes the flag platform uses to apply targeting rules. Always include a stable user identifier.

// TypeScript — OpenFeature
import { OpenFeature, type EvaluationContext } from "@openfeature/server-sdk";

const ctx: EvaluationContext = {
  targetingKey: user.id,        // required: stable, not session-scoped
  email: user.email,
  orgId: user.orgId,
  plan: user.plan,              // "free" | "pro" | "enterprise"
  country: request.geo.country,
  appVersion: "2.4.1",
  betaTester: user.betaTester,  // custom boolean attribute
};

const client = OpenFeature.getClient("payments");
const enabled  = await client.getBooleanValue("new-checkout", false, ctx);
const variant  = await client.getStringValue("checkout-layout", "control", ctx);
const timeout  = await client.getNumberValue("api-timeout-ms", 5000, ctx);
const config   = await client.getObjectValue("rate-limit-config", { rpm: 100 }, ctx);