Install on macOS Apple Silicon

On Apple Silicon, ParseHawk runs the API, worker, Web UI, and Phoenix in Docker. The model server runs natively on the host through vLLM Metal so it can use the Mac’s GPU and unified memory.

Requirements

• An Apple Silicon Mac
• 16 GB unified memory minimum; 32 GB or more for larger context windows
• Docker Desktop
• uv
• Xcode Command Line Tools

ParseHawk is verified on M3 Pro machines with 18 GB and 36 GB unified memory.

Install prerequisites

Install the Apple developer tools if they are not already present:

xcode-select --install

Start Docker Desktop and wait until its engine is ready. Then confirm the command-line prerequisites:

docker info
uv --version
xcode-select -p

Install the CLI

git clone https://github.com/parsehawk/parsehawk.git
cd parsehawk
uv tool install --editable .

Reinstall after dependency or entry-point changes:

uv tool install --editable . --force

Start the stack

parsehawk start

On the first run, ParseHawk provisions its pinned vLLM Metal environment under ~/.parsehawk/runtimes/, downloads the configured model, and warms the runtime. This can take several minutes. Later starts reuse the environment and model cache.

Verify all services:

parsehawk status
parsehawk doctor

Then open http://127.0.0.1:5173 or continue to your first extraction.

Tune for available memory

ParseHawk selects conservative runtime settings from the detected memory tier. Override them only when you know the workload fits:

PARSEHAWK_VLLM_MAX_MODEL_LEN=16384 parsehawk restart

If the runtime is killed or the machine becomes unresponsive, lower the context length before changing other settings. See deployment and hardware for the trade-offs.