Experimental Rust bindings + safe-ish wrapper for Luxonis DepthAI-Core v3.

• High-level crate: depthai-rs (Rust API)
• Low-level crate: depthai-sys (builds DepthAI-Core and exposes an FFI surface via autocxx)

• depthai-sys
  • Builds DepthAI-Core and its dependencies (in target/dai-build/).
  • Compiles a small C++ wrapper (depthai-sys/wrapper/wrapper.cpp) and generates Rust bindings using autocxx.
• depthai-rs
  • Safe(-er) Rust wrapper types like Device, Pipeline, typed camera helpers, and a generic node API.

depthai-sys/            # FFI crate (build script + wrapper)
	build.rs           # clones/builds DepthAI-Core (Linux) or downloads prebuilt (Windows)
	wrapper/           # C ABI functions used by Rust
src/                 # Rust API (`Device`, `Pipeline`, nodes, camera helpers)
examples/            # runnable examples
tests/               # tests (some are ignored unless you enable hardware testing)

• Linux: primarily Debian/Ubuntu-like systems (today).
• Windows: intended to use prebuilt DepthAI-Core artifacts.

If you’re on another distro/OS, it may still work, but you may need to adjust packages and toolchain paths.

Install build tooling used by autocxx + CMake builds:

sudo apt -y install \
	clang libclang-dev \
	cmake ninja-build pkg-config \
	python3 \
	autoconf automake autoconf-archive libtool \
	libudev-dev libssl-dev \
	nasm \
	libdw-dev libelf-dev

Optional (not required for core builds, but useful for local OpenCV tooling):

sudo apt -y install libopencv-dev

Some DepthAI devices require udev rules so you can access them without running as root. If you hit permission errors (or see the device only under sudo), consult the official DepthAI/DepthAI-Core docs for the recommended udev rules for your device.

Install:

• LLVM/Clang (for autocxx/libclang)
• Visual Studio Build Tools (C++ workload)
• CMake (if not already installed)

Example (PowerShell):

winget install -e --id LLVM.LLVM

• rerun: Enables Rerun visualization support with RerunHostNode. Adds Tokio runtime and web viewer dependencies.

• hit: Hardware Integration Tests - enable with cargo test --features hit when you have a physical device connected.

To build without the rerun feature:

cargo build --no-default-features

From the repo root:

cargo build

Notes:

• The first build can take a while because DepthAI-Core is fetched/built and dependencies are prepared.
• Build artifacts for native code live under target/dai-build/.

# Basic examples
cargo run --example pipeline_creation
cargo run --example camera
cargo run --example composite_node

# Host node examples
cargo run --example host_node
cargo run --example threaded_host_node

# Rerun visualization examples (requires rerun feature)
cargo run --example rerun_host_node --features rerun
cargo run --example rgbd_rerun --features rerun

This section is generated from the native DepthAI-Core C++ examples vendored in this repo under target/dai-build/depthai-core/examples/cpp.

• ✅ in the Supported column means depthai-rs currently wraps enough of that feature/node to build and run at least one equivalent pipeline.
• A blank cell means it’s not yet wrapped/exposed in the Rust API (even if DepthAI-Core supports it).

depthai-sys exposes a few environment variables that affect native builds:

• DEPTHAI_CORE_ROOT: override the DepthAI-Core checkout directory.
• DEPTHAI_SYS_LINK_SHARED=1: prefer linking against libdepthai-core.so (otherwise static is preferred).
• DEPTHAI_OPENCV_SUPPORT=1: enable DepthAI-Core OpenCV support (if available).
• DEPTHAI_DYNAMIC_CALIBRATION_SUPPORT=1: toggle DepthAI-Core dynamic calibration support.
• DEPTHAI_ENABLE_EVENTS_MANAGER=1: toggle DepthAI-Core events manager.

This usually means another process already owns the device connection.

• Close other DepthAI apps (including Python scripts) and try again.
• Prefer Pipeline::with_device(&device) so you don’t accidentally open two connections.

Make sure clang and libclang-dev are installed on Linux, and that LLVM is installed on Windows.

By default, the build prefers static linking where possible. If you opt into shared linking (DEPTHAI_SYS_LINK_SHARED=1) you may need to ensure the runtime loader can find the shared libraries.

There is a hit feature flag intended for hardware integration testing:

cargo test --features hit

See LICENSE.