Installation
Pick the library for your stack. All three are peers — same engine, same semantics.
Python
Requires Python 3.10+. Core depends only on numpy.
cd python
pip install -e ".[pandas]"
Extras: [pandas] for the DataFrame convenience layer, [rt] for the native
RT-J backend (sentence-transformers + huggingface_hub), [dev] for pytest.
Java
Requires Java 17+. Gradle modules under group dev.relativedb:
relativedb-core— engine, schema, PQL parser, retriever SPIrelativedb-rt— optional JNA binding to the native RT-J engine
cd java
./gradlew build
Rust
Cargo workspace; the crate depends only on chrono and libloading.
cd rust
cargo build
Native model engine (optional)
The RT-J model backend needs the C++ library librt_c:
cd cpp
cmake -B build -S . && cmake --build build -j
All libraries auto-discover cpp/build/librt_c.{dylib,so}, or set
RELATIVEDB_RT_LIB. Everything else works without it — the default backend is
model-free.