Reproducibility & Environments

Make results repeatable by fixing parameters, paths, tools, and store versions.

What BPM gives you

  • Fixed params: precedence and project.yaml storage avoid hidden changes.
  • Stable paths: render.into keeps outputs predictable across runs and hosts.
  • Store versioning: the active BRS records version and commit in stores.yaml.

Your part (inside templates/hooks)

  • Pin tools: use modules/conda/container with explicit versions in pre_run hooks.
  • Capture metadata: write tool versions and command lines to files next to outputs.
  • Publish pointers: expose key outputs via publish so downstream steps don’t guess paths.

Simple patterns 

# hooks/env.py
def main(ctx):
    # Example: activate conda or module; keep it simple in your environment
    # os.environ["PATH"] = f"/opt/toolX-1.2/bin:" + os.environ["PATH"]
    return None

# template_config.yaml
hooks:
  pre_run:
    - hooks.env:main

Recording context

  • Include a small RUN_INFO.txt from run.sh with key versions and ctx.params.
  • Consider adding a publish key like run_info pointing to that file.

When to pin the store

  • For regulated or shared work, use a tagged release of your BRS and avoid auto‑updates.
  • bpm resource update refreshes the cache; record the resulting version/commit in change logs when you update.