Skip to content
track

Local Prerequisites

This page is about your local machine: the one where you run track, capture tasks, and open the browser UI.

Release installer prerequisites

Section titled “Release installer prerequisites”

The normal install path uses the released trackup bootstrap script. That script downloads the shared backend assets from the GitHub release and builds the track CLI from the tagged source release with cargo install.

Make sure these tools exist locally before you continue:

  • git
  • curl
  • rustc
  • cargo
  • jq
  • tar
  • cmake
  • clang
  • docker compose
  • one C compiler: cc, gcc, or clang
  • one C++ compiler: c++, g++, or clang++

If you want a quick verification pass, this is a reasonable checklist:

Terminal window
git --version
curl --version
rustc --version
cargo --version
jq --version
tar --version
cmake --version
clang --version
docker compose version
command -v cc || command -v gcc || command -v clang
command -v c++ || command -v g++ || command -v clang++

trackup also verifies release checksums before it installs anything. It can use either sha256sum or shasum, and most Linux and macOS systems already ship one of those by default.

On Linux x86_64, trackup can also install a CUDA-accelerated CLI build. That path requires a local CUDA toolkit installation with nvcc available.

Local capture model

Section titled “Local capture model”

track uses a local GGUF model for task capture.

For the normal setup path, you do not need to prepare a model manually. The first task capture downloads the default model into ~/.track/models automatically if it is not already cached.

If you know in advance that you want a different model, keep the GGUF file at a stable path and override it later through track configure --model-path ... or the configuration file. That is an advanced setup, not something you need before the first run.