By ByteBot
JetBrains has open-sourced YouTrackDB, a graph database two years in the making — built by the engineer who rewrote OrientDB’s storage engine, from the codebase of a project SAP acquired, failed to maintain, and effectively killed. It landed on Hacker News this week with 140+ points, Apache 2.0 licensed. The story here is not just another database release. It is what happens when enterprise acquires open source, the community scatters, and one engineer rebuilds quietly in the background.
How OrientDB Died
OrientDB started in 2010 as a Java graph database with real ambitions: object-oriented data models, Gremlin queries, ACID transactions. It built a production user base. Then CallidusCloud acquired the company in 2017. SAP bought CallidusCloud in 2018 for $2.4 billion. What followed was slow-motion abandonment: commercial support dropped in September 2021, full maintenance ended in December 2023.
The community split in two directions. Luca Garulli — the founder — left SAP and started ArcadeDB from scratch, applying a decade of lessons to a clean codebase. Andrii Lomakin, the engineer who rebuilt OrientDB’s storage layer into the production “plocal” engine between 2012 and 2014, went a different way. He forked the original OrientDB codebase in December 2024 and built YouTrackDB for JetBrains. The GitHub issue documenting SAP’s support drop is worth reading if you want a time capsule of what production-grade open-source abandonment looks like in practice.
What YouTrackDB Brings
The technical pitch centers on an embedded-first design. YouTrackDB ships as a shaded uber-jar — no separate server, no network overhead, no dependency conflicts with your existing classpath. Relationships are stored as native pointers, which delivers O(1) link traversal without computing JOINs at runtime. ACID transactions with snapshot isolation are on by default, eliminating dirty reads and phantom reads. For queries, you get YQL (a SQL dialect with graph extensions), Gremlin via the Apache TinkerPop API, and a hybrid match() step that accepts GQL-style pattern matching inside a Gremlin query.
The licensing situation matters here as much as the features. Neo4j removed embedded deployment in version 5.x, which narrowed your options considerably if you needed a graph database inside a JVM process. ArangoDB and Memgraph carry BSL 1.1 licenses with commercial restrictions. ArcadeDB is Apache 2.0 and excellent, though its focus leans toward multi-model performance rather than object-oriented application mapping. YouTrackDB targets a specific niche: embedded, Apache 2.0, OO-first, JVM-native — backed by a company with real commercial software in production. Compare this to other database-as-a-service approaches like ExtendDB’s PostgreSQL-backed DynamoDB compatibility layer: both are solving the “escape the vendor” problem, just in different directions.
The Honest Production Status
It is worth being precise here. JetBrains uses YouTrackDB internally. However, YouTrack — the issue tracker — still runs on Xodus, JetBrains’ key-value store. The migration to YouTrackDB is the stated long-term plan, not the current reality. The version number (0.5.0-SNAPSHOT) reflects where this project sits in its development arc. Moreover, the HN discussion includes a project participant confirming it is not yet in active production use in YouTrack itself.
That distinction matters. Nevertheless, “backed by JetBrains” is meaningfully different from “interesting solo-maintainer project.” JetBrains has commercial incentive to keep this going. Lomakin built the storage engine that ran OrientDB in production for years. The risk profile here is lower than average for an open-source database at version 0.5.
What to Watch
Three groups have an immediate reason to look at this. First, Java and Kotlin developers building applications that need graph traversal without running a separate server process — the embedded story fills a real gap that Neo4j’s licensing decisions created. Second, the OrientDB diaspora: engineers who built production systems before SAP’s 2021 exit and have been carrying technical debt since. YouTrackDB’s OO model and storage lineage will feel familiar. Third, developers building AI agents who need local graph memory. Graph databases are the emerging substrate for agent knowledge stores, and YouTrackDB’s own roadmap already includes vector index support for AI and RAG use cases.
YouTrackDB is not version 1.0. It is not yet the database powering the product it is named after. But it is the most credible continuation of the OrientDB lineage with active corporate backing, a genuine technical niche, and a license that will not surprise you in year two. The repository is live, the Apache 2.0 license is real, and the embedded deployment story works today. If graph databases with object-oriented semantics are anywhere on your roadmap, this is the moment to take a serious look.
