A type-safe, realtime collaborative Graph Database in a CRDT

TypeScript graph database goes realtime — devs split between AI dreams and JS screams

TLDR: An alpha TypeScript graph database that syncs live across tabs and speaks AI-friendly queries is here. Commenters are split between excitement about “agent brains” and worries about JavaScript performance, chain-heavy APIs, tool overload, and how schema changes sync—making it a flashy but contentious vision for collaborative, AI-ready data.

A new TypeScript-based graph database just dropped, promising live, Google-Docs-style collaboration and AI-friendly queries, and the comments are already a battlefield. The pitch: define your data with simple schemas, query it with Gremlin/Cypher (think map-reading for data), store it in a browser-friendly Yjs CRDT (a “no-fights” syncing system), and even let an LLM (a large language model) ask questions. There’s a slick airline-routes demo, and it runs anywhere TypeScript does.

The hottest take? Agent brains. One commenter imagines this as the memory palace for AI assistants, comparing it to ants finding sugar and reorganizing when it runs out. On the other side, skeptics bark: Why build a database in JavaScript/TypeScript? Performance alarm bells ring, with folks warning of “scale traps” and hard-to-shard nightmares. Meanwhile, a veteran dev rolls their eyes at the retro chain-everything API, calling it a pain to test and a throwback to 2009.

Fans praise the Cypher-over-Gremlin angle because LLMs can already speak Cypher, but the nerdery gets real when someone asks: what happens during schema changes in a synced doc—conflicts or silent drops? And then there’s the meme crowd: “Four languages in a trench coat”—Gremlin, Cypher, Yjs, Zod—while the Datalog faithful ask why not just use one tool to rule them all. It’s alpha, it’s bold, and it’s stirring up maximum comment chaos.

Key Points

  • A TypeScript-first graph database offers full type safety for schemas, queries, traversals, and mutations.
  • It supports Gremlin-like traversals and Cypher-like queries, plus full-text search and a pluggable storage layer.
  • Schemas can be defined with Zod (and other Standard Schema libraries), flowing types through compile and runtime.
  • Switching to a Yjs-backed storage (YGraph) enables offline-first, realtime collaborative sync via CRDTs.
  • Distributed via npm with no native dependencies; used in production at codemix but labeled alpha-quality.

Hottest takes

“Cypher-over-Gremlin is a smart call — LLMs can write Cypher” — brianbcarter
“It seems like this could be a performance trap” — 2ndorderthought
“You can do all of this in datalog” — AlotOfReading
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.