Installation

Choose the runtime model that matches your environment.

Recommendation

Use Docker for a first local or server deployment. Use Apptainer on HPC systems where Docker is not available. Use the non-containerized path only when you explicitly need to manage Python, Perl, and system dependencies yourself.

Path	Best for	Notes
Docker	Workstations, servers, reproducible local deployments	Easiest path when Docker is available
Apptainer	HPC systems and environments without Docker daemon access	Uses the same container image through Apptainer or Singularity
Non-containerized	Hosts where you want direct control over Python, Perl, and system dependencies	Most flexible, but requires more local dependency management

All installation modes use the same core idea: prepare the large external reference data once, then point bff-tools to that data.

Preflight Checklist

• Confirm that you have at least 200 GB of disk space for the full external data setup.
• Decide which reference genome your input uses before running genomic conversion.
• If you plan to use load or full, make sure MongoDB and mongosh are available.
• If you only want to validate metadata or inspect command behavior, start with the bundled test data first.

After Installation

Check that the command is visible from your chosen runtime:

bin/bff-tools --help

After installation, continue with the Quick Start.