Skip to main content

Installation

Python​

pip install vectlite

Pre-built wheels for macOS (x86_64, arm64), Linux (x86_64, aarch64), and Windows (x86_64). Requires Python 3.9+.

Install from source​

pip install git+https://github.com/mcsedition-hub/vectlite.git#subdirectory=bindings/python

Requires Rust and Cargo installed.

Optional extras​

The Python package ships with no hard runtime dependencies. Install the relevant SDK only for features you use:

  • LangChain integration -- pip install langchain-core
  • LlamaIndex integration -- pip install llama-index-core
  • Built-in embedders -- pip install openai / cohere / voyageai / fastembed / sentence-transformers (see the embedder factories in the Python API reference)
  • ONNX cross-encoder reranker -- pip install onnxruntime tokenizers huggingface-hub
  • Stemmer for the analyzer pipeline -- pip install PyStemmer
  • OpenTelemetry tracing -- pip install opentelemetry-api

Node.js​

npm install vectlite

Requires Node.js 18+. Pre-built binaries for macOS (x86_64, arm64), Linux (x86_64), and Windows (x86_64). Other platforms compile from source (requires Rust/Cargo).

Optional extras​

  • OpenTelemetry tracing -- npm install @opentelemetry/api

Rust​

Add to your Cargo.toml:

[dependencies]
vectlite-core = { git = "https://github.com/mcsedition-hub/vectlite.git" }

Swift (experimental)​

Experimental bindings built on a shared UniFFI layer. Ship VectLiteFFI.xcframework (built by the bindings/swift/build-xcframework.sh script in the repo) and add the Swift package wrapper to your project.

import VectLite

let db = try VectLite.open(path: "knowledge.vdb", dimension: 384)

See the bindings/swift folder in the GitHub repository for the package manifest and the XCFramework build script.

Kotlin / JVM (experimental)​

Experimental Kotlin/JVM bindings via UniFFI + JNA. Source lives in bindings/kotlin.

import org.mcsedition.vectlite.*

val db = VectLite.open("knowledge.vdb", dimension = 384)

Build the native library through the Gradle package in bindings/kotlin -- it compiles the generated UniFFI Kotlin source and loads the Rust FFI library through JNA at runtime.