Why we built it: Graphs are quickly becoming foundational in AI workflows—GraphRAG, hybrid RAG, knowledge graphs, data lineage, agent memory, etc. But graph query languages have been fragmented for years (Cypher, SPARQL, Gremlin), which hurts portability and locks users in. With ISO GQL finally standardized, we wanted a small, embeddable, fully open implementation that anyone can build on.
What’s in GraphLite today: * Full ISO GQL 2024 support * Rust implementation memory-safe * ACID transactions, 435+ tests * No server, embeddable—Like “SQLite for graphs” * Python + Java bindings * Developed with AI-assisted coding (all PRs human-reviewed, tested, and documented)
Why We're posting here: It’s been a a few days since launch, and the community response has been energizing—100+ stars, early PRs, feature requests, and bug fixes. We have also published the Rust crates and tightened up the docs and internals.
Our long-term vision: A fully embeddable GraphRAG / HybridRAG engine built on an open standard rather than proprietary DSLs.
We’d love your feedback: * What would you want from an embedded graph database? * For those building graph-based AI workflows… what’s missing from existing tools?
We genuinely want to hear from you — design critiques, technical concerns, challenges, wishlists, all of it. So please share your thoughts with us on github, discord, or linkedIn at links below.
And if you’re interested in contributing: There’s a ton of surface area: performance layers, bindings, storage engines, examples, benchmarks, GraphRAG tooling, etc. We have good-first-issues tagged, and we’re very responsive on PRs. Early contributors can shape the project in meaningful ways.
GitHub: https://github.com/GraphLite-AI/GraphLite Discord: https://discord.gg/qMhREHB7 LinkedIn: https://www.linkedin.com/showcase/graphlite-ai/posts/?feedView=all&viewAsMember=true