Skip to content
Sunbitrn
Preparing for launch

Technology

The engineering behind the product.

Sunbitrn is a native mobile app backed by edge services and a relational database. This page describes what is in the current build, and keeps what is planned clearly apart.

Architecture

Three layers, one account.

The app talks to services at the edge, which read and write a relational database. Real-time features use a persistent connection alongside ordinary requests.

  1. Mobile app

    React Native with Expo

    • Feed, messaging, channels, marketplace
    • Local cache for offline use
  2. Edge services

    Cloudflare Workers

    • Content and news endpoints
    • Turnstile on card claims
  3. Data

    PostgreSQL via Supabase

    • Accounts and communities
    • Server-side validation

This describes the mobile client in the current development build. Server-side components are summarised only where the client confirms them.

In the current build

What exists today.

Native mobile client
A React Native app built with Expo. Android is the first platform being prepared for launch; other platforms have not been announced.
Edge services on Cloudflare Workers
Content endpoints, including the aggregator behind the news channels, run on Cloudflare Workers. The app retries with a timeout and backoff and falls back gracefully when an endpoint is slow.
Relational data on PostgreSQL
Account and community data lives in PostgreSQL, accessed through Supabase.
Real-time connections
Messaging and presence use WebSockets. A device holds one connection, sends a heartbeat every 30 seconds, closes the connection when the app goes to the background, and reconnects itself if the server stops acknowledging.
Honest feed signals
A post counts as seen only after it has been at least half visible for one second, and as read after three. Events are batched, carry unique IDs so retries cannot double count, and use a capped queue so tracking can never slow the interface. The server validates every event again.
Offline resilience
Conversations and feeds are cached on the device and shown when the network is gone. Optimistic updates, such as a comment or reaction, roll back cleanly if sending fails. Automated tests cover this behaviour.
Account protection
Two-step verification at sign-in, a list of signed-in devices that can be signed out remotely, credentials kept in the device’s secure store, and a Cloudflare Turnstile challenge on Community Card claims to resist automated abuse.

Infrastructure

Built to run at the edge.

Latency and abuse are the two things a social product feels first. Sunbitrn puts its content services close to people and challenges automated traffic before it reaches the database.

Content endpoints run on Cloudflare Workers, so responses are produced near the person asking for them rather than in a single region.

Community Card claims are protected by Cloudflare Turnstile, which keeps scripted accounts from draining limited supplies.

This website is a static build prepared for Cloudflare Pages, with its security headers and content policy defined in the project rather than left to defaults.

Direction

Where machine learning could help.

No AI features are included in the current app. These are directions Sunbitrn intends to evaluate, each grounded in a signal or queue that already exists. They are plans, not capabilities, and carry no dates.

Feed relevance
The feed already records honest impression and dwell signals. A learned ranking model is a natural next step to evaluate against them.
Moderation triage
Reports arrive with a stated reason and reach owner queues. Models that help order and group those reports could shorten response time, with people still making the decision.
Search and discovery
Search spans people, usernames, channels and cards. Better understanding of what someone is looking for could improve how communities are found.

Reviewing Sunbitrn technically?

Tell us what you would like to see and we will walk you through the product and its architecture.