Vector is a lightweight state and data manager with built-in replication, parallel work distribution and stealing, as well as a novel classification algorithm for prioritization and adaptive smoothing via higher-order derivatives. It inlines time and serializes data into compact memory buffers.
It is useful for developers who want ultra-fast entity state management, automated network batching, and responsive parallel load balancing without writing boilerplate synchronization code.
Key Features
Ultra-Low Latency (~5.5µs Reads / ~5.9µs Writes): Writes and Reads get as low as 5.5 - 5.9µs latency over 10k iterations, averaging 7 to 8.9µs with the slowest being 9µs
Parallel Distribution:Your game is lagging? well vector is able to automatically detect delays within your vectors(tables) to automatically distribute work. No need for manual gates. It also knows when the delay is normal via history and EMA.
Replication: Vector makes batches and sends them to the client to be synced. Using the data client side is as simple as simply grabbing it from vector field via local script.
Novel Classification: Unlike standard classifiers that get polluted by one-time loops or spikes, Vector tracks the velocity, acceleration, and rate of change (jerk) of your data writes. It automatically sorts fast variables from slow ones with zero manual configuration—giving you peak classifications even under heavy workload shifts.
Listeners: set up a listener for any variable to detect changes, Whether it would be a change or a target value. Currently there are 3 types:
OnWrite - Callback on first write
Event- Callback on replication
Poll - actively listen via heartbeat.
Rollback: Save critical data previous values (must include Critical = true) to be able to rollback data