DEV Community

Cover image for Introducing Dravexor – The Async Router Powering 2M+ AI Ops a Day
Pasindu Ranasinghe
Pasindu Ranasinghe

Posted on

Introducing Dravexor – The Async Router Powering 2M+ AI Ops a Day

"In a world full of chatter, Dravexor whispers."


What is Dravexor?

Dravexor is a lightweight async routing engine designed to optimize high-speed task dispatching across distributed AI systems.

With a minimal memory footprint, no dependencies, and multithreaded orchestration, Dravexor has already been adopted in federated research clusters across Asia and Eastern Europe.

It’s currently processing over 2 million async routing operations daily.
All without you ever hearing about it — until now.


Why Should You Care?

Modern AI isn’t just about inference and models.
It’s about how fast you can move data between nodes — how low your latency is, how clean your router is.

And that’s where Dravexor comes in:

  • Ultra-low async overhead
  • 🧠 Context-aware signal propagation
  • 🌍 Designed for federated, multi-node topologies
  • 🔐 Stateless, cache-resilient packet switching
  • 📦 Zero dependencies. Plug and play.
  • 📊 Metrics-ready hooks for observability

Architecture Overview

┌────────────┐         ┌────────────┐          ┌────────────┐
│  Producer  ├──▶▶───▶│ Dravexor   ├──▶▶───▶│  Node Pool │
└────────────┘         └────────────┘          └────────────┘
                       (Routing Engine)
Enter fullscreen mode Exit fullscreen mode
  • The producer emits payloads
  • Dravexor routes them with intelligent dispatch
  • Node pool simulates inference or task workers

How to Try It

git clone https://github.com/PasinduDushan/dravexor.git
cd dravexor
python3 main.py
Enter fullscreen mode Exit fullscreen mode

No dependencies. Just fire it up.

You’ll see simulated payloads routed across a node pool using async logic.
Perfect for experimentation or integrating into a real-world load balancer.


Real-World Use

Dravexor isn’t a toy. It’s already been used in:

  • Federated AI experiments in academic research clusters
  • Prototypes of embedded routers in low-memory environments
  • Simulations where you can’t afford bloated brokers like RabbitMQ or Kafka

The goal? Keep it silent, fast, reliable.


The Philosophy

Dravexor isn’t trying to solve everything.
It solves one problem ruthlessly well:

How do you move payloads from A to B with minimal overhead and no nonsense?

It’s engineered for speed. Adopted for silence.


Built by the Community

Dravexor was started by Pasindu Ranasinghe as a weekend experiment.
It is now maintained by the Core Routing Initiative.

We welcome contributions, feature requests, and wild deployments.
Come whisper with us.


🔗 Links


“It’s surreal to see something I built on a weekend now quietly routing millions of ops daily around the world.”
Pasindu

Top comments (0)