limen.cli

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

Canonical docs

• Command-Line Interface
• Experiment Manifest
• Universal Experiment Loop

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.
commands/run.py	Execute a YAML experiment	Compiles YAML into a manifest and runs UEL.
commands/resume.py	Continue a checkpointed artifact-backed run	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

cli/
|-- main.py       # Click command group and command registration
|-- git_utils.py  # git helper calls for project commands
`-- commands/
    |-- validate.py
    |-- profile.py
    |-- run.py
    |-- resume.py
    |-- init.py
    |-- list_templates.py
    |-- commit.py
    |-- ls.py
    `-- new.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.
• 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.

Read next

• Command-Line Interface
• Experiment Manifest