CelebiChrono v1.0.0b4

Published 2026-10-10

Commands and help

The three entry points now have distinct roles: celebi opens the interactive shell, celebi-cli exposes the complete non-interactive interface, and celebi-git contains Git-specific operations.
Commands remain flat but are presented in task-oriented groups covering projects, navigation, objects, configuration, execution, data, runners, inspection, files, and maintenance.
Canonical names are more explicit, including project-path, set-parameter, set-environment, set-memory-limit, runner-environments, and view-url. Earlier spellings remain available as hidden compatibility aliases.
The interactive shell now supports all three help forms: helpme, help <command>, and <command> --help.
Previously missing CLI wrappers are now exposed, including the multi-input and multi-object commands, cache settings, result checks, documentation helpers, file display/export, impression maintenance, and bookkeeping helpers.
The obsolete jobs command is removed; use status, history, and engine-logs instead.

submit handles initial execution and safe re-submission. Active workflows are deduplicated, completed results are reused, and failed, dissonant, or stopped workflows can be replaced with their predecessor recorded.
submit --with-unknown refreshes an uncertain SSH launch and creates a replacement only after failure is confirmed.
status --wait synchronizes with Yuki’s workflow refresh lock, queries the runner, and streams changed snapshots until the object reaches a terminal state. --poll-interval / -i controls the delay.
Navigation and project commands now fail cleanly outside a Celebi project.

A new IHEP HepJob backend submits both the workflow coordinator and individual Snakemake rules through hep_sub, tracks and cancels child jobs, maps Celebi memory limits, and keeps runtime state on shared storage.
Delayed SSH launches remain unknown and pollable until later evidence confirms success or failure.
Failed/stopped remote workflows are fenced from republishing stale cache data; completed cache entries are published atomically.
Dependency failures retain structured blocked_by snapshots, and failed workflow stageout can be collected explicitly as unofficial diagnostic output.
Yuki’s packages are reorganized into focused jobs, workflows, runners, execution, storage, services, and utility modules.
Conda-based RabbitMQ setup is more reliable and enables Celery-compatible behavior.

The README, Read the Docs command reference, bilingual website guides, and Celebi skills now use the same command names, categories, and help model.