Reference / Limen research

limen.cli

New runs use compressed round artifacts; existing JSONL runs remain readable. See the round storage format.

Contents

Provide the limen shell command for YAML validation, profiling, execution, validation selection, final assessment, manifest storage and project scaffolding.

Canonical docs

In the research projects

The Compiled project command_line scenario invokes the installed CLI, retains manifest lineage, runs and resumes an experiment, and verifies backup restoration. A CLI-built SFD belongs to Compiled Research.

What this package owns

Owns the preferred operator path for YAML manifests: scaffold, validate, profile, dry-run, run, resume, and manage committed manifests. Does not own manifest semantics, experiment execution internals, model behavior, or docs-site build behavior.

Key entry points

Entry point Use case Notes
limen.cli.main:cli Click command group exported by the package script Registered as the limen console script.
commands/validate.py YAML validation from the shell Uses limen.yaml validation.
commands/profile.py Parameter-space and runtime profiling for YAML manifests Fetches the configured data source and samples a covering array of permutations.
assessment.py Validation-selected Sensors/Cohorts and guarded final assessment Registers limen assess and limen reserve
commands/run.py Execute a YAML experiment Compiles YAML into a manifest and runs UEL with MSQ.
commands/resume.py Continue a saved YAML/MSQ run from a checkpoint Used through limen run --resume.
commands/init.py and commands/list_templates.py Bundled template discovery or scaffolding Reads limen/yaml/templates.
commands/commit.py and commands/ls.py Committed manifest storage Writes project-local manifests/committed state.
commands/new.py New project scaffold Clones the official project template.

Adjacent modules

  • limen.yaml parses, validates, compiles, stores, and profiles YAML manifests.
  • limen.experiment executes the compiled SFD through UEL.
  • limen.sfd supplies packaged decoder templates.

Quick orientation

Module or family Responsibility
main.py Click command group and command registration
git_utils.py git helper calls for project commands
text · README.md
cli/├── README.md├── __init__.py├── assessment.py├── commands/├── git_utils.py└── main.py

Things to know

  • limen profile always reports static search-space fields, then fetches the configured production data_source and runs a covering-array sample for timing and data-quality diagnostics. A data-load or sample failure is reported in the profile result.
  • The normal first-run loop is limen init, limen validate, limen profile, limen run --dry-run, then limen run.
  • limen run accepts both direct YAML files and committed manifest://sha256: URIs.
  • limen run writes a result directory containing the copied manifest, metadata.json, results.csv, and round_data.jsonl.gz.
  • limen commit writes the project manifest store before attempting the git commit.
  • Command code should stay thin; YAML semantics belong in limen.yaml and execution semantics belong in limen.experiment.

Limen / Python Research and Compiled ResearchReturn to the research book
Edit this sourceSource revision · 9bc8a630
Reference / keep your place
Find a passage