FeatureGate

About this project

• •A feature flag check sits in the hot path of every request, so it cannot be a network call. FeatureGate evaluates locally and pushes changes out over SSE.
• •Segment-based targeting on top of a rule evaluation engine that uses consistent hashing for percentage rollouts, so a user who is in the 10% stays in the 10% as the rollout grows. Flag changes propagate to connected clients over SSE in under 500ms.
• •The Node.js SDK (packages/sdk-node in the monorepo) caches flags in memory and evaluates locally, putting a check under a millisecond and removing the per-request network call entirely. The developer experience is deliberately LaunchDarkly-compatible so migrating costs nothing.
• •Runs on an Oracle Cloud VM behind Nginx under PM2. GitHub Actions runs lint, typecheck, build, and tests on every push; a separate deploy workflow rsyncs the release to the VM over SSH and reloads the process.

Key Features

• •Consistent hashing so percentage rollouts stay stable
• •Flag changes propagate over SSE in under 500ms
• •Node SDK evaluates in memory, under a millisecond
• •LaunchDarkly-compatible developer experience
• •Oracle Cloud VM behind Nginx, managed by PM2
• •GitHub Actions CI plus rsync-over-SSH deploys

Technologies Used