GenosDB - The serverless graph database with real-time P2P sync
by•
What's new since GenosDB's first launch: cellular-mesh sync that breaks WebRTC's full-mesh limit, so real-time stays smooth as the room grows; an optional Fallback Server that joins as just another peer for persistence and 24/7 uptime; and a zero-trust security model — signed, peer-verified operations with role-based access, per-node ACLs, and rule-based governance. Runs in the browser: real-time, offline-first, serverless by default. Free for personal and commercial use.
Replies
Hi PH 👋 I'm Esteban, the maker of GenosDB.
GenosDB is a graph database that runs entirely in the browser — real-time P2P sync, offline-first, serverless by default. Since its first launch it's grown a lot:
• Cellular-mesh sync — WebRTC full-mesh doesn't scale, so GenosDB now partitions peers into cells linked by elected bridges. Connections per peer stay flat as the room grows (measured live: 4 connections where a full mesh needed 14).
• Optional Fallback Server — spin up a node that joins the network as just another peer, giving you persistence and 24/7 availability whenever you want it. Your infra, your data, zero lock-in.
• Zero-Trust security — every operation is cryptographically signed and verified by each peer, with no central authority. On top of that: role-based access, per-node ACLs, and rule-based governance (promotion rules are GenosDB queries).
You asked last time for more docs and real examples — both shipped (links below).
Upfront on licensing: GenosDB is free for personal and commercial use — integrate it, ship it, even resell it in your app, no fees. The source code is proprietary.
Install it, open two tabs, and watch your data sync in real time. I'd love to hear what you think.
The OPFS storage engine choice is genuinely clever for a decentralized setup, avoiding the usual IndexedDB headaches for performance-critical graph traversals. Curious how you handle conflict resolution when two peers edit the same node offline.
@glayeg2g Thanks — that was exactly the reasoning behind OPFS: skip the IndexedDB pain and get a real file handle for the graph.
On conflicts: every write carries a Hybrid Logical Clock (HLC) timestamp. When two offline peers edit the same node and then reconnect, the sync protocol reconciles deterministically — last-write-wins by HLC, which gives a total order every peer agrees on, so they all converge to the same state with no central coordinator. The catch-up itself rides a delta oplog (with a full-state fallback) so peers only exchange what changed. Happy to go deeper into that if it's useful.