🦉 Tawny 🦉

🚀 Tawny is an extremely low latency, public facing pubsub. It uses protobuf for end-to-end communication (from browser to server and server to browser) using grpc-web protocol. It allows you to build an awesome user experience with realtime features.

Core features

Tawny was designed with performance/scaling/reliability in mind. It provides multiple keys feature:

• Publish/Subscribe, publish a message from anywhere and get your users notified instantly.
• Presence, know who is connected or not in a channel, perfect for building presense feature like google docs.
• State, you can share a state with your presence, awesome for "is typing" feature.
• Binary data, thanks to profobuf/grpc, your payload doesn't need to be json. It can be every binary file
• Private channel, you can control who has access to a channel.
• Complete admin library/cli, manage your Tawny server using a powerful admin api.
• Monitoring, monitor your server easily with our promotheus (and more to come) metrics exporter.

HTTP2/Protobuff advantage over HTTP1/json

• Network bandwidth Profobuf allows the app to use less bandwidth as demonstrated there.auth0 blog
• Usage of http2 Http2 is the new version of the famous http protocol, faster, more reliable, allows server push and much more. cloudfare blog
• Binary data You can send raw binary data with tawny using protobuff.

Getting started

Visit the documentation

Feature roadmap

This is our feature roadmap. If you want to ask for a new feature, please open an issue

v0.1 Alpha => On going

• Push service (Public facing publish/subscrib)
• Presence service (Presence on channel/topic with state)
• Admin service
• Typescript web npm package
• Typescript server npm package
• Golang module
• Automatic TLS termination for https
• GRPC secure mode
• Docker image
• Protected topic (message can only be pushed from authenticated admin request)

Backlog

• Secure channel
• Cluster mode
• Admin ui
• Clients (java/swift/php/c#/python)

Performance

With 1 core / 1gb of ram on Digital ocean, Tawny is able to process up to 50k messages seconds within < 5 ms latency.

Community

Join us on gitter

Show your support

Give a ⭐️ if you support the project!

_This README was generated with ❤️ by [readme-md-generator]