← all repositories
objectbox/objectbox-java

SQLite with a green conscience, now doing vectors

A NoSQL database for Android and JVM that recently grew vector search to keep AI workloads on-device.

4.6k stars Java RAG · SearchData Tooling
objectbox-java
Velocity · 7d
+1.3
★ / day
Trend
steady
star history

What it does ObjectBox is an embedded NoSQL database for Java and Kotlin apps. You annotate POJOs with @Entity, stash them in typed Box containers, and skip SQL entirely. It runs on Android 5+ and JVM 8+ across Linux, macOS, and Windows.

The interesting bit The project claims to be the “first on-device vector database” in this space — embedding storage and similarity search bolted onto the same CRUD engine your app already uses. That lets you do RAG or generative-AI retrieval locally without shipping data to a server or running a separate vector store.

Key highlights

  • Vector search built in, not bolted on via extension
  • Native object relations without manual foreign-key juggling
  • Targets low CPU and power draw; the README mentions sustainability as a first-class concern
  • Gradle plugin with version-catalog support, plus a Maven path for JVM projects
  • Apache 2.0 licensed

Caveats

  • Maven setup is JVM-only; Android developers are Gradle-bound
  • Benchmark claims vs. SQLite and Realm are stated but not detailed in the README
  • The “first on-device vector database” claim is specific to the Java/JVM ecosystem, not universal

Verdict Worth a look if you’re building Android or JVM apps that need local AI retrieval and would rather not stack SQLite + a separate vector library. Less compelling if you’re already happy with Room or need heavy cross-platform C++ integration.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.